perm filename E78.OUT[LET,JMC]1 blob
sn#386670 filedate 1978-10-10 generic text, type C, neo UTF8
COMMENT ⊗ VALID 00333 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00034 00002
C00035 00003 ∂03-Jul-78 2333 CLT fr
C00036 00004 ∂04-Jul-78 0040 RAK A new winner
C00037 00005 ∂04-Jul-78 1511 JLH Pascal
C00038 00006 ∂04-Jul-78 1836 EJG Phone message
C00039 00007 ∂04-Jul-78 2313 LLW Party and Vertical Lifter
C00041 00008 ∂04-Jul-78 2327 MRC quote from JAB (who is going to BBN)
C00042 00009 ∂04-Jul-78 2350 LLW Technological Exotica Discussion
C00043 00010 ∂05-Jul-78 0057 MRC LOTS.DMP[1,MRC]
C00044 00011 ∂05-Jul-78 1045 Creary at SUMEX-AIM ARPA Proposal Revision
C00046 00012 ∂05-Jul-78 1115 LCW S1 Meeting
C00047 00013 ∂05-Jul-78 1302 REF
C00048 00014 ∂05-Jul-78 1515 RAK A modest proposal
C00051 00015 ∂05-Jul-78 1538 BTH s1 meeting
C00052 00016 ∂05-Jul-78 1602 DCL Data disk use restriction
C00053 00017 ∂05-Jul-78 2027 DCL
C00054 00018 ∂06-Jul-78 0819 Creary at SUMEX-AIM Problem Solver
C00056 00019 ∂06-Jul-78 0902 LCW S1 CS Meeting
C00057 00020 ∂06-Jul-78 1255 DWW Change in meeting place
C00058 00021 ∂06-Jul-78 1332 ZM leave of absence
C00059 00022 ∂06-Jul-78 1947 MRC dialnet
C00060 00023 ∂06-Jul-78 2144 LCW Semi-Weekly Technical Meetings
C00061 00024 ∂06-Jul-78 2211 100 : m addenda to dialnet renewal
C00068 00025 ∂07-Jul-78 1036 Tajnai at SUMEX-AIM SCIP funds
C00070 00026 ∂07-Jul-78 1308 BPM NY Times wire
C00071 00027 ∂07-Jul-78 1422 RPG Talk at U of I
C00072 00028 ∂07-Jul-78 1534 Creary at SUMEX-AIM Paper Search
C00073 00029 ∂07-Jul-78 1551 RAK
C00077 00030 ∂07-Jul-78 1636 GLS Dynamic floating point
C00078 00031 ∂07-Jul-78 1640 GLS MACSYMA BFLOAT package
C00079 00032 . Message from 100 on TTY161 0720: J. Weizenbaum via MIT-ML
C00080 00033 ∂08-Jul-78 1343 JOSEPH at MIT-ML (Joseph Weizenbaum)
C00081 00034 ∂08-Jul-78 1440 100 : Creary via AMES-TIP#50 Password
C00082 00035 ∂08-Jul-78 1450 JOSEPH at MIT-ML (Joseph Weizenbaum)
C00085 00036 ∂08-Jul-78 1919 MRC dialnet
C00086 00037 ∂08-Jul-78 2230 ELLEN at MIT-MC (V. Ellen Lewis) Forwarded request
C00088 00038 ∂09-Jul-78 0159 Feigenbaum at SUMEX-AIM thoughts on Meltzer and Journal Editorship
C00091 00039 Marvin and Ed:
C00099 00040 ∂09-Jul-78 1939 SSO at SU-AI (Susan Owicki) Modal logic seminars
C00101 00041 ∂09-Jul-78 2200 Feigenbaum at SUMEX-AIM copy of my message to russell after he alerted me to trouble
C00104 00042 ∂10-Jul-78 0030 MRC IAZ needs an intro to PUB
C00106 00043 ∂10-Jul-78 1305 PAT
C00107 00044 ∂10-Jul-78 1323 GLS Iterative top level for Interpreter
C00108 00045 ∂10-Jul-78 1553 GLS Spotty iteration
C00109 00046 ∂10-Jul-78 2252 MINSKY at MIT-AI (Marvin Minsky)
C00114 00047 Replies
C00136 00048 ∂11-Jul-78 0048 MRC dialnet
C00137 00049 ∂11-Jul-78 0151 JMC
C00139 00050 ∂11-Jul-78 1103 HVA Continuation of Your Appointment
C00140 00051 ∂11-Jul-78 1353 TOB
C00143 00052 ∂11-Jul-78 1612 PLS Hayes' paper
C00144 00053 ∂12-Jul-78 1003 Tajnai at SUMEX-AIM Principal Dissertation Advisor
C00145 00054 ∂12-Jul-78 1308 JB WISE MEN.
C00147 00055 ∂12-Jul-78 1417 RWW meta stuff from A. Bundy
C00148 00056 ∂13-Jul-78 0226 KEN at MIT-AI (Kenneth Kahn) Your paper ''Ascribing Mental Qualities to Machines''
C00156 00057 ∂13-Jul-78 0120 LGC via AMES-TIP#57 Problem Solver
C00159 00058 ∂13-Jul-78 1259 MRC Dialnet
C00160 00059 ∂13-Jul-78 1255 DCL Stanford -1 project
C00161 00060 ∂14-Jul-78 0431 DCL
C00163 00061 ∂14-Jul-78 0824 Feigenbaum at SUMEX-AIM AI Qual Exam for 1978-79 academic year
C00165 00062 ∂14-Jul-78 1250 RWW
C00167 00063 ∂14-Jul-78 1533 CARLSON at USC-ISI Re: Request for meeting
C00169 00064 ∂15-Jul-78 0617 DEK dinner tonite
C00170 00065 ∂15-Jul-78 1502 JMC*
C00171 00066 ∂15-Jul-78 1627 RAK Computer usage JULY 1-14--S1 takes off
C00173 00067 ∂15-Jul-78 2028 MRC Dialnet proposal
C00175 00068 ∂16-Jul-78 0152 MRC Dialnet protocol documentation
C00176 00069 ∂16-Jul-78 1613 MRC runaway REMUX
C00177 00070 ∂16-Jul-78 2257 ARK S1 Meeting
C00178 00071 ∂17-Jul-78 0201 HWC REMUX
C00183 00072 ∂17-Jul-78 0216 RAK REMUX--last comment from me, promise
C00185 00073 ∂17-Jul-78 0230 MRC forwarding mail without sender's consent
C00187 00074 ∂17-Jul-78 1433 JB ORALS.
C00188 00075 ∂17-Jul-78 1550 LGC via AMES-TIP#20 Problem Solver
C00191 00076 ∂17-Jul-78 1648 LCW S1FORT Meeting
C00192 00077 ∂18-Jul-78 0310 MRC PROTOC.XGP[DLN,MRC]
C00193 00078 ∂18-Jul-78 1402 LES
C00195 00079 ∂18-Jul-78 1615 Hart at SRI-KL (Peter Hart) Invitation
C00198 00080 ∂18-Jul-78 1646 LCW S1 CS Meeting Times
C00199 00081 ∂18-Jul-78 2001 MRC mailed without comment
C00207 00082 ∂19-Jul-78 1201 Creary at SUMEX-AIM Appointment Request
C00208 00083 ∂19-Jul-78 1306 DCL
C00209 00084 ∂19-Jul-78 1641 MRC merge of opcode and channel?
C00210 00085 ∂19-Jul-78 2136 MRC ARPAnet directory
C00211 00086 ∂20-Jul-78 1204 DCL Arpa Proposal
C00212 00087 ∂20-Jul-78 1356 MRC
C00214 00088 ∂21-Jul-78 1041 TOB music 222
C00216 00089 ∂21-Jul-78 1226 WINOGRAD at PARC-MAXC2 Splenitude
C00218 00090 ∂21-Jul-78 1231 Blum at SUMEX-AIM New discussion group on medical computing applications
C00220 00091 ∂22-Jul-78 0310 DCL
C00223 00092 ∂22-Jul-78 0556 MRC Datacomputer availability
C00227 00093 ∂22-Jul-78 1624 LGC via AMES-TIP#50 Oral presentation
C00228 00094 ∂22-Jul-78 1649 LGC via AMES-TIP#50 Message Addendum
C00230 00095 ∂22-Jul-78 1728 MRC APRAnet directory
C00231 00096 ∂22-Jul-78 2352 LGC via AMES-TIP#54 ARPA Appendix
C00234 00097 ∂23-Jul-78 0421 LLW Dinner
C00235 00098 ∂23-Jul-78 1246 LGC via AMES-TIP#53 Problem Solver ARPA Appendix
C00236 00099 ∂23-Jul-78 1938 TOB
C00237 00100 ∂23-Jul-78 2319 LES Appendix to ARPA proposal
C00238 00101 ∂24-Jul-78 0000 JMC*
C00239 00102 ∂24-Jul-78 0310 LGC via AMES-TIP#52
C00241 00103 ∂24-Jul-78 1000 JMC*
C00243 00104 ∂24-Jul-78 1854 Creary at SUMEX-AIM Phone Numbers
C00244 00105 ∂25-Jul-78 1008 PEG Rides to/from SAIL
C00245 00106 ∂25-Jul-78 1305 DCL via SU-TIP#10 Delete list
C00246 00107 ∂25-Jul-78 1532 ME COPY/DIRED within E
C00248 00108 ∂26-Jul-78 0646 FB via SU-TIP#1 big micros
C00249 00109 ∂26-Jul-78 1035 CLT
C00250 00110 ∂26-Jul-78 1410 DAN
C00251 00111 ∂26-Jul-78 1754 MRC autologout
C00253 00112 ∂26-Jul-78 2231 MRC file transfers
C00254 00113 ∂26-Jul-78 2358 MRC Is LOTS now urgent
C00255 00114 ∂27-Jul-78 0843 REG via AMES-TIP#54 If you see this in time, XS
C00256 00115 ∂27-Jul-78 1013 REG CANCEL MY PREVIOUS
C00257 00116 ∂27-Jul-78 2324 Davis at SUMEX-AIM car
C00258 00117 ∂28-Jul-78 0040 PEG Dull lunches at SAIL
C00260 00118 ∂28-Jul-78 1311 DCL via SU-TIP#10 Phone lines and delete list
C00261 00119 ∂28-Jul-78 1946 Wiederhold at SUMEX-AIM datamedias
C00262 00120 ∂28-Jul-78 2324 PMF FSIM
C00263 00121 ∂29-Jul-78 0043 RWW
C00264 00122 ∂29-Jul-78 0052 RWW
C00265 00123 ∂29-Jul-78 0203 MRC Dialnet modem at LOTS
C00266 00124 ∂29-Jul-78 1407 JMC NSF proposal for program-prover cum proof-checker for LISP course
C00268 00125 ∂30-Jul-78 1441 ARK S1 Meeting
C00269 00126 ∂30-Jul-78 1504 MRC dialer at LOTS
C00270 00127 ∂30-Jul-78 1511 MRC modem
C00271 00128 ∂30-Jul-78 1625 MRC LOSING VADIC
C00272 00129 ∂30-Jul-78 1643 DCL
C00275 00130 ∂30-Jul-78 2214 Boyer at SRI-KL (Bob Boyer) Proposal for Course Improvement
C00277 00131 ∂31-Jul-78 1018 Engelmore at SUMEX-AIM Siglunch for Thursday, August 3
C00279 00132 ∂31-Jul-78 1024 RWW woody
C00280 00133 ∂31-Jul-78 1035 Tajnai at SUMEX-AIM cs206
C00281 00134 ∂31-Jul-78 1045 Boyer at SRI-KL (Bob Boyer)
C00282 00135 ∂31-Jul-78 1153 CLT
C00285 00136 ∂31-Jul-78 1246 Rsmith at SUMEX-AIM RUTGERS
C00287 00137 ∂31-Jul-78 1327 MRC losing Dialnet modem
C00291 00138 ∂01-Aug-78 1148 TED VADIC MODEMS
C00292 00139 ∂01-Aug-78 1253 LCW PASCAL Header Page
C00293 00140 ∂01-Aug-78 1300 MRC modems
C00294 00141 ∂01-Aug-78 1905 REM at MIT-AI (Robert Elton Maas) Interval-Refinement Stochastic-Mapping
C00296 00142 ∂01-Aug-78 2046 REM via AMES-TIP#66
C00297 00143 ∂01-Aug-78 2146 PMF PDP-10 pascal misfeature
C00298 00144 ∂02-Aug-78 1611 YAW possible visit
C00300 00145 ∂02-Aug-78 1854 DCL Meeting
C00301 00146 ∂02-Aug-78 1859 DCL
C00302 00147 ∂02-Aug-78 2305 CLT
C00305 00148 ∂02-Aug-78 2327 CLT
C00306 00149 ∂03-Aug-78 0435 REM at SU-AI at MIT-MC (Robert Elton Maas) More good data-compression with INTCOM/IRSM
C00313 00150 ∂03-Aug-78 1054 PAT call from Laske
C00314 00151 ∂03-Aug-78 1119 CLT
C00315 00152 ∂03-Aug-78 1209 REM via AMES-TIP#20 First draft of intended paper (public, CACM or IEEEx etc.)
C00317 00153 ∂03-Aug-78 1252 MRC DLNSER documentation
C00318 00154 ∂03-Aug-78 1624 TW via PARC-MAXC2
C00319 00155 ∂03-Aug-78 2151 MRC SPOOL
C00321 00156 ∂03-Aug-78 2309 MLB New FSIM
C00322 00157 ∂04-Aug-78 1037 Tajnai at SUMEX-AIM Syllabus for Comprehensive
C00325 00158 ∂04-Aug-78 1906 Wiederhold at SUMEX-AIM Minutes from three weeks ago.
C00328 00159 ∂04-Aug-78 2253 LLW Forest Baskett-SAIL Affiliation
C00331 00160 ∂06-Aug-78 2213 REM via SU-TIP#1 INTCOM WORKING WITH LEFT-CONTEXT ADAPTIVE ALGORITHM
C00341 00161 ∂06-Aug-78 2307 REM via SU-TIP#1 More on working version of INTCOM/CONTEXT
C00343 00162 ∂07-Aug-78 0752 MARG at PARC-MAXC hi
C00344 00163 ∂07-Aug-78 0937 Nilsson at SRI-KL (Nils Nilsson) A Word About the Future
C00346 00164 ∂07-Aug-78 1028 REM via AMES-TIP#20 More on new version of INTCOM
C00348 00165 ∂07-Aug-78 1101 REM via AMES-TIP#20 More good results on INTCOM/CONTEXT
C00349 00166 ∂07-Aug-78 1557 DCL
C00351 00167 ∂09-Aug-78 1040 Robinson at SRI-KL (Ann Robinson) Your trip to Washington for Scientists for Shcharansky
C00353 00168 ∂11-Aug-78 2302 RWG re dcl:
C00359 00169 ∂14-Aug-78 1835 Feigenbaum at SUMEX-AIM Re: AIL Usage
C00361 00170 ∂15-Aug-78 0819 MLB
C00362 00171 ∂15-Aug-78 1120 Levinthal at SUMEX-AIM 8/8/1978 Memo
C00363 00172 ∂16-Aug-78 1351 JED AI Qual Reading List
C00365 00173 ∂16-Aug-78 1715 Wiederhold at SUMEX-AIM S1 meeting
C00366 00174 ∂16-Aug-78 1738 WINOGRAD at PARC-MAXC2 Re: AI Qual Reading List
C00368 00175 ∂16-Aug-78 1734 LLW Call for Proposals
C00370 00176 ∂17-Aug-78 0737 RSMITH at RUTGERS RUTGERS MACHINE
C00371 00177 ∂17-Aug-78 0851 Wiederhold at SUMEX-AIM Re: Call for Proposals
C00372 00178 ∂17-Aug-78 0902 Wiederhold at SUMEX-AIM (Response to message)
C00373 00179 ∂17-Aug-78 0942 Wiederhold at SUMEX-AIM (Response to message)
C00374 00180 ∂17-Aug-78 1116 ARK Reminder: S1 meeting
C00375 00181 ∂17-Aug-78 2246 MRC Dialnet
C00379 00182 ∂18-Aug-78 0914 Nilsson at SRI-KL (Nils Nilsson) AI Workshop
C00383 00183 ∂18-Aug-78 1328 BTH NEW ARCHITECTURE MANUAL
C00384 00184 ∂21-Aug-78 1257 David Levy at MIT-AI (David A. Steele)
C00385 00185 ∂22-Aug-78 1352 Tajnai at SUMEX-AIM New PHD Students - support
C00386 00186 ∂23-Aug-78 1500 LCW Sites' FY79 Proposal
C00387 00187 ∂23-Aug-78 2139 LCW via SU-TIP#11 Disk Areas
C00391 00188 ∂23-Aug-78 2224 LCW via SU-TIP#11 AREAS
C00392 00189 ∂24-Aug-78 0017 S1 *,S1 Files
C00393 00190 ∂24-Aug-78 1307 MSS updating distribution files
C00394 00191 ∂24-Aug-78 1554 PAT phone call from Bob Albrecht
C00395 00192 ∂26-Aug-78 2247 LCW Distribution Lists
C00396 00193 ∂27-Aug-78 1616 Feigenbaum at SUMEX-AIM retransmission of my July 9 message re Meltzer
C00400 00194 ∂27-Aug-78 1616 Feigenbaum at SUMEX-AIM what to do about Meltzer's letter of June
C00403 00195 ∂27-Aug-78 1617 Feigenbaum at SUMEX-AIM Re: various
C00404 00196 ∂28-Aug-78 0733 RSMITH at RUTGERS our TOPS20 system...
C00405 00197 ∂28-Aug-78 2242 DCL
C00406 00198 ∂28-Aug-78 2250 Kiessig at Rand-Unix Dialnet project
C00408 00199 ∂29-Aug-78 1448 HEDRICK at RUTGERS account at Rutgers
C00409 00200 ∂30-Aug-78 1127 Tajnai at SUMEX-AIM Max Shiffman
C00410 00201 ∂31-Aug-78 1408 PBK
C00411 00202 ∂31-Aug-78 1606 LCW Linkage Conventions
C00412 00203 ∂01-Sep-78 0001 JMC*
C00413 00204 ∂01-Sep-78 1619 BALZER at USC-ISIB STRAWMAN PROPOSAL TO HELP FOCUS DISCUSSION AT L.A. MEETING
C00420 00205 ∂03-Sep-78 1053 JLH Pascal improvement project
C00422 00206 ∂04-Sep-78 0017 MRC
C00423 00207 ∂04-Sep-78 0028 LLW Proposals and Requests-for-No-Cost-Extensions
C00428 00208 ∂05-Sep-78 2250 Wiederhold at SUMEX-AIM S1 meetings
C00452 00209 ∂06-Sep-78 0952 Nilsson at SRI-KL (Nils Nilsson) Report on the SRI AI meeting
C00457 00210 ∂07-Sep-78 0934 Sacerdoti at SRI-KL (Earl Sacerdoti) ''white paper'' on distributed robotics
C00488 00211 ∂08-Sep-78 0007 LCW Contracts
C00489 00212 ∂08-Sep-78 0010 REM WORKING INTCOM VERSION 2
C00492 00213 ∂08-Sep-78 0233 REM Addenda on new Interval-Refinement compression program
C00494 00214 ∂09-Sep-78 0341 REM More results from my new data-compression program
C00496 00215 ∂10-Sep-78 2247 REG via SU-TIP LOTS- If you're back in timee
C00498 00216 ∂11-Sep-78 0457 REM Now is the time...
C00499 00217 ∂11-Sep-78 1410 MARK SHERMAN at CMU-10A (C410MS40) DFTP Allocation on the Datacomputer
C00501 00218 ∂12-Sep-78 0850 MINSKY at MIT-AI (Marvin Minsky) klaut still lives
C00507 00219 ∂12-Sep-78 0924 REM Enhancements to my crunch program
C00509 00220 ∂12-Sep-78 1430 JOSEPH at MIT-ML (Joseph Weizenbaum)
C00511 00221 ∂12-Sep-78 1528 BALZER at USC-ISIB SUMMARY OF L.A. MEETING ON POSSIBLE A.I. INITIATIVE
C00521 00222 ∂12-Sep-78 1735 MANN at USC-ISIB afterthoughts on AI initiative ISI meeting
C00525 00223 ∂13-Sep-78 0413 REM via SU-TIP Crunch program available to users on limited basis
C00528 00224 ∂13-Sep-78 0959 PAT Doshita
C00529 00225 ∂13-Sep-78 1155 BALZER at USC-ISIB MESSAGE LIST FOR A.I. INITIATIVE GROUP
C00532 00226 ∂13-Sep-78 1234 MRC
C00534 00227 ∂13-Sep-78 1445 HEDRICK at RUTGERS (Response to message)
C00535 00228 ∂13-Sep-78 1452 HEDRICK at RUTGERS sloppy message reading
C00536 00229 ∂13-Sep-78 1822 DPB
C00537 00230 ∂13-Sep-78 2333 MARG at PARC-MAXC (Response to message)
C00538 00231 ∂14-Sep-78 0758 JMC robot
C00539 00232 ∂14-Sep-78 0909 JMC
C00540 00233 ∂14-Sep-78 0911 Feigenbaum at SUMEX-AIM (Response to message)
C00541 00234 ∂14-Sep-78 1120 JRA
C00543 00235 ∂14-Sep-78 1651 Tajnai at SUMEX-AIM Update on New Ph.D. Support
C00552 00236 ∂15-Sep-78 0000 JMC*
C00553 00237 ∂15-Sep-78 0040 IAZ Dialnet file transfers at SAIL
C00556 00238 ∂15-Sep-78 0735 JRA
C00557 00239 ∂15-Sep-78 1358 RDR via AMES-TIP ComputerWorld network issue To: MRC at
C00558 00240 ∂15-Sep-78 1655 REF
C00559 00241 ∂15-Sep-78 2149 DEK f(x,y,z)←if x≤y then y else f(f(x-1,y,z),f(y-1,z,x),f(z-1,x,y))
C00561 00242 ∂16-Sep-78 0119 DPB colloquium
C00562 00243 ∂16-Sep-78 0502 MRC FTP data transfer rate
C00563 00244 ∂16-Sep-78 1656 MARG at PARC-MAXC dinner
C00564 00245 ∂16-Sep-78 1737 REM via SU-TIP Abstract for paper I plan to submit to COMM-ACM or similar. Feedback?
C00567 00246 ∂16-Sep-78 2259 LLW DARPA and S-1 Systems
C00570 00247 ∂16-Sep-78 2311 LLW
C00571 00248 ∂17-Sep-78 1222 DEK ctak
C00572 00249 ∂17-Sep-78 1527 Allen Newell at CMU-10A (A310AN02) Re AIJ editor
C00574 00250 ∂18-Sep-78 0120 REM via SU-TIP More IC2 results
C00575 00251 ∂18-Sep-78 0739 Feigenbaum at SUMEX-AIM Re: Re AIJ editor
C00576 00252 ∂18-Sep-78 1225 REP via SRI-KL TAing 206
C00577 00253 ∂18-Sep-78 1322 CLT book
C00578 00254 ∂18-Sep-78 1942 MRC Klatu
C00579 00255 ∂18-Sep-78 1946 MRC Klatu show
C00580 00256 ∂18-Sep-78 2241 IAZ FTP
C00581 00257 ∂19-Sep-78 0004 MINSKY at MIT-AI (Marvin Minsky) (Response to message)
C00621 00258 ∂19-Sep-78 0903 MINSKY at MIT-AI (Marvin Minsky)
C00622 00259 ∂20-Sep-78 1139 DPB Colloq.
C00623 00260 ∂20-Sep-78 1602 100 : pat flight to detroit
C00624 00261 ∂21-Sep-78 1014 Tajnai at SUMEX-AIM Queenie/LOTS
C00625 00262 ∂21-Sep-78 1514 REP
C00626 00263 ∂22-Sep-78 1611 JBR
C00627 00264 ∂23-Sep-78 1006 REM via AMES-TIP More IC2 results...
C00629 00265 ∂23-Sep-78 1439 Moore at SRI-KL (J Moore) A Mechanical Proof of a Metatheoretic Result
C00635 00266 ∂23-Sep-78 1604 TOB collaboration
C00636 00267 ∂23-Sep-78 2146 CARL at MIT-AI (Carl Hewitt)
C00641 00268 ∂23-Sep-78 2219 MRM shoes
C00642 00269 ∂24-Sep-78 1016 REM
C00644 00270 ∂24-Sep-78 1557 RPG Downtime
C00645 00271 ∂25-Sep-78 0947 SSO Modal logic/concurrent programming seminar
C00647 00272 ∂25-Sep-78 0953 DAVID at USC-ISIB Information request
C00649 00273 ∂25-Sep-78 1215 ARK S-1 Meeting
C00650 00274 ∂25-Sep-78 1345 PN wednesday meeting
C00651 00275 ∂25-Sep-78 1513 Tajnai at SUMEX-AIM FACULTY MEETING, TUESDAY, SEPTEMBER 26, 2:30 p.m., POLYA 204
C00652 00276 ∂25-Sep-78 1601 MRC Dialnet at LOTS
C00655 00277 ∂25-Sep-78 1640 SSO correction of seminar date
C00656 00278 ∂25-Sep-78 1659 Mcgoveran at SUMEX-AIM Faculty meeting
C00657 00279 ∂25-Sep-78 1954 DSB
C00658 00280 ∂26-Sep-78 0918 Mcgoveran at SUMEX-AIM Today's Faculty Meeting
C00659 00281 ∂26-Sep-78 0928 JRA lisp
C00660 00282 ∂26-Sep-78 0949 FB via SU-TIP S-1 management and funds
C00662 00283 ∂26-Sep-78 1042 RWW
C00663 00284 ∂26-Sep-78 1123 LES Half-cocked response
C00664 00285 ∂26-Sep-78 1151 PAT stoyan paper
C00665 00286 ∂26-Sep-78 1315 ARK S-1 Meeting Rescheduled
C00666 00287 ∂26-Sep-78 1411 CH oak tree mazda
C00667 00288 ∂26-Sep-78 1647 REM at MIT-MC (Robert Elton Maas) Design for a Home-Information-Terminal (HOTER) service
C00681 00289 ∂26-Sep-78 1707 FB via SU-TIP S-1 contract and funds
C00682 00290 ∂26-Sep-78 1827 MRC Dialnet at SAIL
C00683 00291 ∂26-Sep-78 1855 MRC change to Dialnet programming at SAIL
C00684 00292 ∂27-Sep-78 1126 MARG at PARC-MAXC ( Forwarded Mail )
C00686 00293 ∂27-Sep-78 1906 Mcgoveran at SUMEX-AIM Informal Meeting of Faculty, Sen Res Assoc, Res Assoc
C00688 00294 ∂27-Sep-78 1957 SSO Visit and seminar
C00689 00295 ∂28-Sep-78 0126 MRC Tops-20 Dialnet
C00691 00296 ∂28-Sep-78 0838 AFB
C00694 00297 ∂28-Sep-78 2027 CLT 206 notes
C00695 00298 ∂28-Sep-78 2247 MRC
C00696 00299 ∂29-Sep-78 0055 REP NEW 206 CLASSROOM
C00697 00300 ∂29-Sep-78 0808 RWW shoes
C00698 00301 ∂29-Sep-78 1549 CLT
C00699 00302 ∂29-Sep-78 1647 Feigenbaum at SUMEX-AIM dinner sunday night with the Michies
C00700 00303 ∂30-Sep-78 1010 Jrobinson at SRI-KL (Jane Robinson) brunch
C00702 00304 ∂30-Sep-78 1726 AFB
C00703 00305 ∂30-Sep-78 1935 JMC note for LISP bulletin
C00704 00306 ∂30-Sep-78 2301 PG tech-note-for-lisp-bulletin
C00707 00307 ∂01-Oct-78 1627 MRC Dialnet bug
C00708 00308 ∂01-Oct-78 1729 RPG Require
C00709 00309 ∂02-Oct-78 1530 SSO Seminar for Thursday, Oct. 5 at 2:45. ERL 237
C00713 00310 ∂02-Oct-78 1533 SSO Lunch
C00714 00311 ∂03-Oct-78 0557 ZM car
C00715 00312 ∂03-Oct-78 0858 SSO TIME CHANGE
C00716 00313 ∂03-Oct-78 1622 CH phone call
C00717 00314 ∂05-Oct-78 0814 JRA lisp conf
C00718 00315 ∂05-Oct-78 0923 CH DIALNET USE
C00719 00316 ∂05-Oct-78 1010 Feigenbaum at SUMEX-AIM copy of message to Carlson
C00721 00317 ∂05-Oct-78 1045 BPM Recognition of human faces
C00722 00318 ∂05-Oct-78 1100 MRC*
C00723 00319 ∂05-Oct-78 2252 MRC VADIC modem for GSB
C00725 00320 ∂06-Oct-78 1100 MRC*
C00726 00321 ∂06-Oct-78 1349 MRC via SU-TIP Dialnet
C00727 00322 ∂06-Oct-78 1855 MRC FYI
C00731 00323 ∂06-Oct-78 2230 PEG S1 meeting notes-- new location
C00732 00324 ∂08-Oct-78 0523 RWG at MIT-MC (Bill Gosper)
C00734 00325 ∂08-Oct-78 1731 LLW Jeff Rubin
C00738 00326 ∂09-Oct-78 1649 Lederberg at SUMEX-AIM Corrected directory entry SAIL]: people.dat[per,csd]
C00740 00327 ∂09-Oct-78 2308 LGC via AMES-TIP This Afternoon's Discussion
C00741 00328 ∂10-Oct-78 0827 CARLSON at USC-ISI Real-Time Symbolic Computing Workshop
C00742 00329 ∂10-Oct-78 1156 Mcgoveran at SUMEX-AIM Computer Science Visiting Advisory Committee
C00745 00330 ∂10-Oct-78 1557 DPB reminder re cs200
C00748 00331 ∂10-Oct-78 1610 PAT my schedule
C00749 00332 ∂10-Oct-78 1953 MRC Dialnet metering
C00751 00333 ∂10-Oct-78 2217 Feigenbaum at SUMEX-AIM carlson response to my msg about jmc
C00753 ENDMK
C⊗;
∂03-Jul-78 2333 CLT fr
you should look at the paragraph beginning on line 132 of page 3
and the paragraph beginning on line 118 of page 4.
∂04-Jul-78 0040 RAK A new winner
To: LES, JMC, DCL
You may be interested to know that Music is no longer the group with the
most accounts (28, not counting Music 220) after the list of Ph.D. students.
Yes, there are 30 people with
assorted S1 accounts. Of these, 12 would probably have accounts for other
reasons. But 18 are original. I decided to check because they are getting
far more obvious. But then we have enough resources for everybody.....
Dick
∂04-Jul-78 1511 JLH Pascal
To: GIO, JMC, SSO
CC: JLH
Wednesday 11-12 seems to suit most people , DSL conference room.
∂04-Jul-78 1836 EJG Phone message
Your wife just called. She said to tell you that she is on her way home,
and hopes to be home by 8:00. She is in Oakdale (?) now.
∂04-Jul-78 2313 LLW Party and Vertical Lifter
To: JMC
CC: LLW
John, I was very glad that you could come to the party this past Saturday,
and I again want to apologize for being somewhat unavailable for rational,
reasonably well-connected discourse during much of it, especially near the
end. I hope to eliminate these shortcomings on subsequent occasions, when
I hope you'll also be able to attend.
Dr. Rod Hyde, my Group's Advanced Design & Systems Studies Section Leader
(and the guy with whom I design fusion starships and other exotica from
time to time), was the one to whom you talked briefly re Hans Moravec's
rotating cable vertical lifter calculations on Saturday PM. He indicated
to me later that evening that he was looking forward to a detailed
discussion of Hans' thinking and calculations with him. Rod was ranked as
one of the most capable graduates ever of MIT's Aero-&-Astro Dept. when he
got his degree there 2 years ago, and I think Hans might find a discussion
with him to be quite productive.
Lowell
It would be interesting to get together for dinner some time before
I go on my trip at the end of the month and discuss technological exotica.
∂04-Jul-78 2327 MRC quote from JAB (who is going to BBN)
"I think [BBN] would be more interested [in Dialnet] than DEC. I don't
think too much of DEC's networking efforts."
He is going to BBN on the 14th. I am going to look into this further
once he's there.
∂04-Jul-78 2350 LLW Technological Exotica Discussion
To: JMC
CC: LLW
I would be delighted to discuss technological exotica with you and anyone
else you consider appropriate, this month or subsequently. Rod and I plan
to throw another party as soon as we get moved into our new location
(presently scheduled for about the third week of this month), and could
format it so as to provide some relatively quite time for discussion;
alternatively, we could meet for dinner at Louie's some evening, or any
place else that seems preferable to you.
Lowell
∂05-Jul-78 0057 MRC LOTS.DMP[1,MRC]
I have hacked it sufficiently so that it will get less buffer overruns.
It does this by exploiting a feature of the system to give it abnormally
high run priority, meaning that if the system is heavily loaded, running
it will make it worse. But at least the output is tolerable.
∂05-Jul-78 1045 Creary at SUMEX-AIM ARPA Proposal Revision
Date: 5 Jul 1978 1046-PDT
From: Creary at SUMEX-AIM
Subject: ARPA Proposal Revision
To: JMC at SU-AI
It now seems clear to me that the final sentence in the description of me
should NOT say that I will "begin development of the proposed problem solver",
but rather that I will "work on epistemological problems of AI and on the
proposed feasibility study for a problem solver." Corresponding changes
should be made in the discussion of the problem solver itself. Also, the first
sentence about me should contain the sequence " ... epistemology, semantics,
logic, ..." instead of "... logic, semantics, epistemology, ..." (This latter
change was made last night, but I'm not sure the file was properly saved.)
Can these changes still be made before the proposal goes out? If further
discussion is needed, I can stop by later today. -- Lew Creary
-------
∂05-Jul-78 1115 LCW S1 Meeting
To: S1.DIS[S1D,PMF]:
The S1 CS Group meeting is Thursday. I propose that it be in the small
conference room at SAIL at 1330.
Curt
∂05-Jul-78 1302 REF
∂05-Jul-78 1242 LES
∂04-Jul-78 1808 REF
SPELL admits the non-word "primative"
This isn't in the dictionary that I gave to REG. I haven't figured out
where his is.
∂05-Jul-78 1515 RAK A modest proposal
To: LES, BH, DCL, JMC
Considering the results of the system meeting, I make the following
proposal for autologout and ena liv:
1. The autologout time should be MAX(15,30-(2*length queue))
2. The "subject to autologout" message, probably reworded, should be
unconditionally displayed on the screen after 10 minutes of idle time.
3. If there is anyone in the queue, the LIV option should be ignored for
data disc channels. (Notice that the "subject to...." message should
still be printed, regardless of LIV status. A user may want to have
an additional option saying "DON'T WARN ME" if there is critical screen
output--this may be relevant even now).
Dick
∂05-Jul-78 1538 BTH s1 meeting
To: S1.DIS[S1D,PMF]:
∂05-Jul-78 1115 LCW S1 Meeting
To: S1.DIS[S1D,PMF]:
The S1 CS Group meeting is Thursday. I propose that it be in the small
conference room at SAIL at 1330.
Curt
-----
I would like to propose that the meetings stay at Serra House for at least
this week. Changes could then be openly discussed.
Brent
∂05-Jul-78 1602 DCL Data disk use restriction
To: JMC, LES
I suggest there should be an EXCLUDE LIST of programmer names. LOGIN will
not let anyone on the list login between 1400 and 1700 Monday to Friday.
Who goes on the EXCLUDE LIST?
1.All Ph.D and M.S. students, Visitors and Guests, who are not affiliated with
a Lab. project. (This will save 2 or 3 channnels as of this msg.)
2. Each project leader gets to choose his top TEN men. Everyone else on his
project goes on theEXCLUDE LIST.
OTHERWISE BUSINESS AS USUAL WITH AUTOLOGOUT ETC.
∂05-Jul-78 2027 DCL
To: JMC, LES
∂05-Jul-78 2024 JMC
To: LES
CC: DCL
General idea of DCL proposal has merit.
REPLY:
BH has pointed out that there is already an EXCLUDE LIST in LOGIN,
so implementation should not be a big deal.
∂06-Jul-78 0819 Creary at SUMEX-AIM Problem Solver
Date: 6 Jul 1978 0809-PDT
From: Creary at SUMEX-AIM
Subject: Problem Solver
To: JMC at SU-AI
I have read the PUB file for the ARPA proposal, and it's just fine -- it
resolves very nicely the concerns that I expressed yesterday. I detected just
two minor glitches: 1) the section head for the problem solver is quoted,
while none of the other section heads are, and 2) [Cartwright 1977] is referred
to in the text of p. 5-2, but does not appear in the bibliography on p. 8-1.
(Maybe the second one was caught during pubbing.)
By the way, my reference yesterday to a "feasibility study" of the
problem solver does not imply an armchair approach to the project on my part.
One of the main things I have learned at HPP is the value of working programs
as a source of intellectual stimulation and development ideas. My present
inclination is to start by getting a version of Nilsson's new deduction system
running, and using that as an experimental springboard for further developments.
-- Lew
-------
∂06-Jul-78 0902 LCW S1 CS Meeting
To: S1
There has been only a little resistance and much support for meeting at
SAIL in the small conference room, so let's do it. 1330 today.
Curt
∂06-Jul-78 1255 DWW Change in meeting place
To: S1
∂06-Jul-78 0902 LCW S1 CS Meeting
To: S1
There has been only a little resistance and much support for meeting at
SAIL in the small conference room, so let's do it. 1330 today.
Curt
[At SAIL? Boy, am I glad I'm not coming to S-1 meetings any more.]
∂06-Jul-78 1332 ZM leave of absence
To: Feigenbaum at SUMEX-AIM
CC: JMC
John suggests that the three of us meet to decide the exact dates of my leave of
absence. How about meeting over lunch tomorrow? Zohar
∂06-Jul-78 1947 MRC dialnet
To: JMC, LES
well, dialnet doesn't seem to crash the system anymore. however, output
doesn't seem to do anything.
∂06-Jul-78 2144 LCW Semi-Weekly Technical Meetings
To: S1
Semi-weekly technical meetings of certain S-1 CS subgroups will commence
tomorrow (Friday). These meetings will be held Tuesdays and Fridays at
the following times at SAIL:
S1FORT, LINK 1PM
SOPA 2PM
MANUAL 3PM
These will be in addition to the regular Thursday bi-weekly administrative
S-1 CS meetings.
Curt
∂06-Jul-78 2211 100 : m addenda to dialnet renewal
you may not agree with all of this, but here's some things from my
viewpoint:
[1] we have demonstrated that 1200 baud is too fast for the "input side"
of many operating systems' terminal service. consequently a large amount
of buffer overflow has been occured. of course, the dialnet protocols
are robust enough to account for and correct this error condition (in fact,
it is our belief that dialnet guarantees essentially error-free data
communications; we have not yet performed an error analysis but we believe
the chance of an undetected/uncorrected error is small enough to be
discountable).
[2] it was partially because of the buffer overrun problem that dialnet
was implemented in the operating system at the ai lab (other reasons were
peculiar to the ai lab's operating system). the ai lab implemention,
which has been completed and is now in the final debugging phase, solves
the buffering overrun problem by taking device interrupts from the modem
at the lowest level, circumventing the terminal service's processing, and
by maintaining its own buffers, which are loaded and emptied at processor
interrupt level.
this implementation level is very efficient, however, it extracts the
price that any operating system work takes in system downtime and
reliability while the programming is being done.
[3] the lots implementation, which will be on a more powerful operating
system than the ai lab's, will attempt in its initial form to be a user
mode process requiring no operating system modifications. the programming
staff at lots has advised us of the buffering problems there, and have
fact have stated that it will be worse than what was encountered at the
ai lab. however, tops-20's terminal service is largely handled by a
pdp11 front end, and our hope is that the "dirty work" can be done in the
pdp11 front end instead of the pdp10 main processor.
[4] our consultations with people who are interested in dialnet seem to
indicate that most are interested in the file transfer aspect of dialnet;
ie, they wish to have a reliable method to transfer files from one
machine to another within the same organization, and are only secondarily
interested (if at all) in talking with another organization's machines.
while this doesn't affect the design of our protocols, it has affected
some of our thinking as to what directions we should aim at in more
detail; in particular, greater attention has been given to the design
of the file transfer protocols than was originally planned, in order to
make file transfer of greater utility to a larger audience.
-- m
ps: i don't think there really has to be much asked for in the forthcoming
years. even if we ask for more modems or something probably the biggest
expendicture will be salaries. i am not really sure what our budget looks
like right now but i suspect that we are going to have a surplus when the
year ends, so perhaps it might be politically advisable to ask for less
(i seem to remember the original proposal asked for 80k(?)) on the grounds
that the work is being done at a lesser expense than had been envisioned...
********
1. I don't know how many of your points are worth mentioning in the renewal
proposal even in so far as I agree with them.
2. With Stanford's cut, we may not be far under budget.
3. I never expected to use operating systems' terminal service and
always supposed that Dialnet would require system modifications.
I also suppose that once Dialnet becomes attractive, manufacturers
would begin to implement it. It shouldn't require separate implementation
for each computer.
4. File transfer was always considered important.
5. While the immediate uses are point-to-point communication, so were
the immediate uses of the telephone. It will take a while after
Dialnet becomes popular, before it becomes common for people who
suddenly find themselves sharing an interest, will suddenly discover
that both of their installations have already implemented Dialnet.
∂07-Jul-78 1036 Tajnai at SUMEX-AIM SCIP funds
Date: 7 Jul 1978 1037-PDT
From: Tajnai at SUMEX-AIM
Subject: SCIP funds
To: faculty.list:
Everyone cooperated so beautifully Autumn - Spring in cutting down
SCIP expenses that we now find we have a healthy balance.
Money is available for worth-while projects. If you have a CS
student who needs SCIP money for 293 or 390 project, please write
a note and refer student to me. Let me know how much money and what
access he/she needs, i.e.
Batch
Interactive Services
keypunching
No. of tracks
We have between $1000 and $1500.
Netmsg is fine in lieu of note.
Carolyn
-------
∂07-Jul-78 1308 BPM NY Times wire
To: ND
CC: TED, LES, JMC
∂07-Jul-78 1149 ND NYT wire
is anyone planning to repair the NYT wire service line?
[BPM: Yes, it's been dead for 10 days now. Couldn't TED be convinced to
look at it?]
∂07-Jul-78 1422 RPG Talk at U of I
To: JMC, TOB, CCG
I will be giving an overview talk on work at the AI Lab
at the Co-ordinated Science Lab in Urbana in 2 weeks. If you could
get your students etc to produce a couple of slides each and a short
burb I'd be happy to espouse whatever point of view you like at the talk.
However, the sooner the better so that I can absorb the material
before I go.
-rpg-
∂07-Jul-78 1534 Creary at SUMEX-AIM Paper Search
Date: 7 Jul 1978 1535-PDT
From: Creary at SUMEX-AIM
Subject: Paper Search
To: JMC at SU-AI
I'm having a hard time locating a copy of the paper indicated below; the
C. S. library doesn't have it, nor does anyone else over here. Do you know
where I might find a copy for xeroxing?
Kowalski, R., "Logic for Problem Solving," University of Edinburgh School of
Artificial Intelligence Memo No. 75 (1974).
-------
∂07-Jul-78 1551 RAK
To: GROUP.DIS[VCG,DCL]: mbrown at PARC-MAXC
***********************************************************************
JOINT SEMINAR
CONCURRENT SYSTEMS
and
VERIFICATION GROUP
TUESDAY 11th JULY
PLACE: ERL 237
TIME: 2:30 p.m.
TITLE: Specification and historical development
of a fast balanced tree merge algorithm.
SPEAKER: Mark Brown
Xerox Research Labs.
ABSTRACT
Bob Tarjan and I recently devised an algorithm for merging sorted lists
represented as height-balanced (AVL) trees. The algorithm is interesting
primarily because its worst-case running time is optimal to within a constant
factor.
Our algorithm is not significantly more complicated than the trivial algorithm
which inserts the items from the smaller list one-by-one into the larger list,
using the standard height-balanced tree insertion procedure. Nevertheless, the
program's text is long and its correctness does not seem obvious. It might be an
interesting program to verify formally, since the program's specification should
be very much smaller than the code itself.
In this talk I will describe the merging algorithm, including the considerations
which influenced its design, and will try to give convincing (but informal)
arguments for its correctness. My hope is that someone in the audience will
become inspired to convince the PASCAL verifier of the algorithm's correctness
(in a more formal way), or to find a flaw in it.
Note: If you want to look at the algorithm before the seminar, the original
code is in M.SAI[VCG,RAK]. A Pascal version of the merging portion only is
in M.PAS[VCG,RAK]
∂07-Jul-78 1636 GLS Dynamic floating point
To: JMC
CC: GLS
Yes, I would be very interested in any ideas along these lines.
Does it differ in major respects from the BFLOAT package
in MACSYMA (which is somewhat clumsy to use)?
Sorry, I'll have to read about that. Please give me a pointer.
∂07-Jul-78 1640 GLS MACSYMA BFLOAT package
To: JMC
CC: GLS
[a] Fateman, Richard J. "The MACSYMA 'Big-Floating-Point'
Arithmetic System". Proc. 1976 ACM Symposium on Symbolic
and ALgebraic Computation (SYMSAC 76), 209-213.
[b] MACSYMA Reference Manual (Version 9), Dec. 1977.
I have copies of these if you want to borrow them.
-- GLS
.;; Message from 100 on TTY161 0720: J. Weizenbaum via MIT-ML
The letter is alleged to have been written t\ by the biologist
Z.A.Medvedev. In any case, please tell Lewis Robinson of the Livermore
Labs that Pat Winston n\and Richard B. Adler (MIT EE/CS Prof.) gave
permission to use their names. Pls acknowledge as this will save
me a long distance phone call to Robinson. Suggest you, John, get
in touch with Fredkin and Marvin via mit-ai to si\olicit their help.
You might ve\\even get J>B>Wiesner to lend his name if YOU tried.
Best - JW
∂08-Jul-78 1343 JOSEPH at MIT-ML (Joseph Weizenbaum)
Date: 8 JUL 1978 1645-EDT
From: JOSEPH at MIT-ML (Joseph Weizenbaum)
To: jmc at SU-AI
Sorry, John, msg got stepped on. Please retransmit.
Sorry, John, your msg got stepped on. Pls retransmit.
,
∂08-Jul-78 1440 100 : Creary via AMES-TIP#50 Password
I will appreciate it very much if you will log in as me (LGC) and set up the
password PHILAIER for me, so I can use it to log in over the ARPANET this
afternoon. Thanks.
∂08-Jul-78 1450 JOSEPH at MIT-ML (Joseph Weizenbaum)
Date: 8 JUL 1978 1751-EDT
From: JOSEPH at MIT-ML (Joseph Weizenbaum)
To: jmc at SU-AI
Read your msg.
Robinson (whom I don't know either) called me yesterday to tell
me that you and he and others are organizing a visit to the U.S.A.
(San Fran.) for Mrs. Scharansky to coincide with her husband's trial.
He asked for my support (given, of course) and that of others --
but not of organizations. Wants a broad base of scientists from many
disciplines and many institutions. Stunned to learn you don't know
(of ?) him. He certainly prominently mentioned your name.
I agree on your thesis re disinformation. Quite right.
The best course with respect to helping political prisoners (anywhere -
a la Amnesty Int.) is to follow one's humane instincts and not try
to be politically subtle. One may occasionally fall into some sort of
trap, i.e., lend one's name to something that turns out to not have been
worthy of support. But in the main, I think one helps save lives.
Robinson is PR director for Livermore Labs -- as I understand it.
His home phone is (415) 254-2537. Would still appreciate your calling
him re Winston and Adler.
Will you communicate with Ed Fredkin and Marvin? Your influence
with THEM is certainly greater than mine.
best Joseph
He must be the person who called me some weeks ago and who I agreed could
use my name on some statement. I'll call him.
∂08-Jul-78 1919 MRC dialnet
i have succeeded in transferring a file over dialnet. unfortunately, doing
so also crashes the system. i hope to fix that tommorrow.
-- m
∂08-Jul-78 2230 ELLEN at MIT-MC (V. Ellen Lewis) Forwarded request
Date: 9 JUL 1978 0132-EDT
From: ELLEN at MIT-MC (V. Ellen Lewis)
Subject: Forwarded request
To: Dianne at MIT-ML
CC: JMC at SU-AI
Date: 8 Jul 1978 1631-PDT
From: JMC at SU-AI (John McCarthy)
To: ellen at MIT-MC
How do I can version 9 of the Macsyma Reference Manual?
-------
Please send him a MACSYMA Care package.
∂09-Jul-78 0159 Feigenbaum at SUMEX-AIM thoughts on Meltzer and Journal Editorship
Date: 9 Jul 1978 0151-PDT
From: Feigenbaum at SUMEX-AIM
Subject: thoughts on Meltzer and Journal Editorship
To: jmc at SAIL, newell at CMUA
Dear John and Al,
Please forgive the delay in my response on the issue of Meltzer's
continuing as Journal Editor. I forgot that I had forgotten to send
it to you!
I am inclined to agree with John's suggestion that Bernard continue
for two more years, but I feel that the period should be LIMITED to
two years. Why
1. Bernard has done a fine job. Since he has implied that he wants to
continue in part at least because of the editorial stipend, the Board
can reward Bernard for his service by asking him to continue for a
while.
2. I do not, as a matter of principle, believe in an individual
being in any one position of power and responsibility too long
(e.g. deans, department chairmen, ARPA office directors,etc.)
Bernard has been Journal Editor a long time! Is it not time for him to
let some younger person have a chance (I don't however have a candidate
in mind).
3. Bernard's active knowledge of AI will decay rapidly after he retires
because he will not be in the stream of research, students, etc.
Raphael, when faced with an analogous departure from the field (moving
to a non-AI managerial career) resigned his Associate Editorship.
4. Two years is a graceful transition period in which to:
reward Bernard; find a new Editor; and allow the new person to find
out what Bernard knows about the process.
Does any of this modify any of your previously expressed thoughts
Ed
-------
Marvin and Ed:
Susie got married today to Dan Gunther, a prelaw student
and grade school classmate, and son of a Stanford law professor.
A fine wedding.
What follows is a statement on inviting Mrs. Scharansky that
Joe Weizenbaum thought I should ask you to adhere to. I think it's
a good thing, although more drastic measures will probably be appropriate.
∂09-Jul-78 1335 Robinson at SRI-KL (Ann Robinson) Shcharansky Press release
Date: 9 Jul 1978 1337-PDT
From: Robinson at SRI-KL (Ann Robinson)
Subject: Shcharansky Press release
To: jmc at SU-AI, joseph at MIT-ML
Attached is a copy of the final version of the Shcharansky press
release which is being released this afternoon.
July 9, 1978
FOR IMMEDIATE RELEASE
AVITAL SHCHARANSKY TO VISIT U.S. AT INVITATION OF U.S. SCIENTISTS
In a show of deep concern over the trial of Soviet dissident
Anatoly Shcharansky on Monday in the Soviet Union, a group of American
scientists has invited Shcharansky's wife, Avital, to visit the United
States during her husband's trial.
Mrs. Shcharansky has accepted the invitation.
In a telephoned statement yesterday from Israel, where she now resides,
Mrs. Shcharansky declared, "I welcome the opportunity to visit America. I
am extremely grateful for the invitation that has been extended to me by the
U.S. scientists. This will give me an opportunity to share with the American
people my grave concern over Anatoly's fate. I believe that coming to the
United States under the auspices of the American scientific community will
be extremely valuable in helping my husband."
Mrs. Shcharansky will visit the U.S. following a short stay in France.
She goes there at the invitation of Francois Mitterrand, First Secretary of
the French Socialist Party.
When she visits this country, Mrs. Shcharansky will be hosted by
members of a number of scientific research institutions and laboratories.
Among the American scientists who extended the invitation to Mrs. Shcharansky
were:
Owen Chamberlain, Physics, University of California at Berkeley
Malcolm Derrick, Physics, Argonne National Laboratory
Maurice Goldhaber, Physics, Brookhaven National Laboratory
Kurt Gottfried, Physics, Cornell University
Arthur Kornberg, Biochemistry, Stanford University
T.D. Lee, Physics, Columbia University
John McCarthy, Computer Science, Stanford University
Fred Mills, Physics, Fermi National Accelerator Laboratory
Burton Richter, Physics, Stanford Linear Accelerator Center
Joseph Weizenbaum, Computer Science, Massachusetts Institute of
Technology
American scientists have repeatedly expressed their concern over
Shcharansky's arrest and imprisonment. In letters, telegrams and
demonstrations of protest they have continually urged his release. In March
a group of Berkeley scientists issued a protest against the Soviet Union's
harassment and detention of Shcharansky. In that message, sent to
Soviet Secretary Leonid Brezhnev, they declared that as long as the
computer scientist was kept in prison, "we will find it dificult to
maintain open and generous sharing with our scientific colleagues in the
Soviet Union."
The Shcharansky trial follows the recent sentencing of another
Soviet scientist and dissident, Yuri Orlov. Orlov was sentenced to seven
years in prison followed by five years of enforced %xile for alleged slander
of the Soviet state. Both Orlov and Shcharansky were organizers of the
Moscow chapter of the Helsinki agreement monitoring committee. (The
agreement contains broad humanitarian declarations concerning individual
civil liberties.)
American scientists were chagrined by the Orlov sentence. They
now fear that its severity marks an escalation in the Soviet disregard for
the rights of dissident Soviet scientists and are alarmed by what may happen
to Shcharansky who is accused of a much greater crime than Orlov.
Anatoly Shcharansky goes on trial on Monday for "high treason in
the form of espionage." The charge carries with it a possible death
penalty. This is the first time since the Stalinist era that such a
severe sentence has been considered. The 30-year old computer scientist has
been accused by the Soviet press of being an agent of the U.S. Central
intelligence Agency, a charge President Jimmy Carter has repeatedly denied.
For more information call Dr. Morris Pripstein at (415) 642-0466
or (415) 843-2086.
-------
∂09-Jul-78 1939 SSO at SU-AI (Susan Owicki) Modal logic seminars
Date: 9 Jul 1978 1915-PDT
From: SSO at SU-AI (Susan Owicki)
Subject: Modal logic seminars
To: MODAL.DIS[1,SSO]:
July 13 -- continue dynamic logic
July 20 -- Pnuelli's temporal logic
July 27 -- Ashcroft's LUCID extended to modal logic (???)
The papers distributed last week are concerned with LUCID. I was going to
lead that discussion, but I will be leaving town July 21 for the rest of
the summer (more or less). So someone else will have to volunteer if that
session is to take place.
Let me know if you want to continue this discussion of modal logic in the fall.
-------
∂09-Jul-78 2200 Feigenbaum at SUMEX-AIM copy of my message to russell after he alerted me to trouble
Date: 9 Jul 1978 2201-PDT
From: Feigenbaum at SUMEX-AIM
Subject: copy of my message to russell after he alerted me to trouble
To: jmc at SAIL, zm at SAIL
Hi, Dave!
I too am away from the office, in Utah at a meeting.
I'll try to get hold of Bill Carlson in the next day or two. I have
a pretty good idea of how he is viewing things (though I have not
talked with him),i.e. if Cordell Green goes, so goes Cordell's money.
Fair enough; but John McCarthy has stong scientific views on why a
significant amount of funding should be put behind Zohar Manna and a
more mathematical approach to program synthesis. John should be allowed
to make his case (and Zohar as well) as completely and forcefully as possible.
I understand that John and Zohar want to go back to ARPA soon to argue the case
(i.e. as soon as you and Bill have a chance to read the proposal text).
In a sense, I am the wrong person to be discussing this, since at Stanford
each PI makes the best case he can for funds from the various agencies,and
the department chairman has little to do with the process. but I'm willing
to help if needed.
Will call you shortly after you get back to the office on Thursday.
Best regards,
Ed
-------
∂10-Jul-78 0030 MRC IAZ needs an intro to PUB
To: LES, JMC
IAZ is currently stalled in his programming efforts for Dialnet FTP because
I do not have the monitor code for Dialnet available for users yet. I am
hurrying on it as much as possible, but well, it's been "any day now" for
a week. So I was thinking maybe he might want to hack up the FTP protocol
document with PUB, etc. to have a suitable document for publication? I
should get on my ass too about getting the other stuff in pretty PUB format
too I guess.
Anyway, could one of you help him? My knowledge of PUB is rather minimal
since I hardly ever use it.
-- m
∂10-Jul-78 1305 PAT
the file is DIRECT.NAM[LET,JMC]. you may want to rename it - patte
∂10-Jul-78 1323 GLS Iterative top level for Interpreter
To: JMC
CC: GLS
Yes, indeed, this idea works out fine. See the papers
by Sussman and myself (MIT AI Memos 349, 353, 379, 443, and 452
-- the SCHEME papers); I have copies of these if you are
missing any.
Presently MacLISP makes use of the iteration idea only
spottily, not consistently.
The implementation must indeed be spotty.
(defun foo (n) (cond ((equal n 0) t) (t (foo (sub1 n)))))
followed by (foo 1000)
gives REG PDL OVERFLOW. An iterative interpreter would not
do that.
∂10-Jul-78 1553 GLS Spotty iteration
To: JMC
CC: GLS
That's right. Now if you change the word DEFUN to DEFINE
and give the same definition to SCHEME, you will not
get a pdl overflow.
∂10-Jul-78 2252 MINSKY at MIT-AI (Marvin Minsky)
Date: 11 JUL 1978 0154-EDT
From: MINSKY at MIT-AI (Marvin Minsky)
To: jmc at SU-AI
CC: MINSKY at MIT-AI
well, I got back too late to see Scharansky release in time.
Covey my best wishes to Susan
Re: Scharansky; something like this might have some practical sting:
From Minsky; McCarthy, Weizenbaum, Etc. etc.:
These scientists are considering a private ban on discussing recent work
in AI and advanced computer science with routine Soviet visitors.
They will, however, make exceptions when they feel it of personal value to
them or their friends.
The reduction of exchange of information is not considered a
significant scientific sacrifice by this group. They agree that
Soviet Computer Science is relatively primitive in
most areas -- with the exception of certain mathematical
areas in which Soviet Science has maintained most of its
traditional competence. But Soviet computer science in general
is far behind western work, and has shown little
sign of closing the gap.
Because Computer Science is felt by many to be likely the
most important area of science in the next few decades --
that is, because it will have enormous effects of industrial
productivity -- this particular action may have very serious
cosequences for the Soviets in the long run.
Unfortunately, there is an irony here, one of the american scientists observed ruefully.
"I'm afraid that soviet computer Science is already so far out of date that their bureaucrats
`ave no real idea of its importancep So itsunlikely to have
any short run effect, for example, on Scharansly7s
situation!".
I think something like your proposal is right. We need to think
more though. (1) I don't know whether general policies or individual
refusals are more likely to be effective. (2) I would like to see
bribery tried. Suppose the Russians had been told six months ago.
"We'll sell you the Cyber 186 if you release Scharansky". The people
who don't think much of imprisoning Scharansky anyway would be able
to argue against it on practical grounds without sticking their
necks out ideologically. My model is that the number of people
with a firm commitment to the ideology is small, so that what might
be taken as a gross insult to the ideology would in fact be effective.
It isn't clear to me how to apply this principle in the present
situation. I also think that the Russians will not be careful about
entering agreements that have unorthodox ideological consequences,
because the people who make the agreements won't care much about
whether other people would be embarassed ideologically. Thus the
Soviets agreed that the program committee of IJCAI 4 could invite
anyody, and the Lerner embarassment hit them probably to the secret
pleasure of some of them.
Replies
∂10-Jul-78 1549 JMC
To: GLS
The implementation must indeed be spotty.
(defun foo (n) (cond ((equal n 0) t) (t (foo (sub1 n)))))
followed by (foo 1000)
gives REG PDL OVERFLOW. An iterative interpreter would not
do that.
∂10-Jul-78 1249 JMC
To: PAT
Is there a file that lists the files deleted from let,jmc?
∂10-Jul-78 1202 JMC reference
To: DCO
"Representation of Recursive Programs in First Order Logic" to be
presented at International Conference on Mathematical Studies of
Information Processing, Aug 23-26, Kyoto, Japan.
∂10-Jul-78 1012 JMC
To: PHON.MSG[LET,JMC]
843 2740x6301 or 5771
∂09-Jul-78 1827 JMC
To: GLS
CC: RPG
Iterative top level for interpreter
I haven't worked this idea out fully, but it seems to me that the
LISP interpreter could be made to treat functions recursively on
the top level as follows:
When an EXPR is found, and say the top level is a COND or
a function application, the interpreter should JRST to the final
outer level function rather than call it recursively.
Inner functions still require recursive calls, but within the
inner functions, iteration will still be done where possible.
This will save running out of PDL and run faster. This corresponds
to a feature of some compilers.
∂09-Jul-78 1813 JMC
To: minsky at MIT-AI, ef at MIT-AI
Marvin and Ed:
Susie got married today to Dan Gunther, a prelaw student
and grade school classmate, and son of a Stanford law professor.
A fine wedding.
What follows is a statement on inviting Mrs. Scharansky that
Joe Weizenbaum thought I should ask you to adhere to. I think it's
a good thing, although more drastic measures will probably be appropriate.
∂09-Jul-78 1335 Robinson at SRI-KL (Ann Robinson) Shcharansky Press release
Date: 9 Jul 1978 1337-PDT
From: Robinson at SRI-KL (Ann Robinson)
Subject: Shcharansky Press release
To: jmc at SU-AI, joseph at MIT-ML
Attached is a copy of the final version of the Shcharansky press
release which is being released this afternoon.
July 9, 1978
FOR IMMEDIATE RELEASE
AVITAL SHCHARANSKY TO VISIT U.S. AT INVITATION OF U.S. SCIENTISTS
In a show of deep concern over the trial of Soviet dissident
Anatoly Shcharansky on Monday in the Soviet Union, a group of American
scientists has invited Shcharansky's wife, Avital, to visit the United
States during her husband's trial.
Mrs. Shcharansky has accepted the invitation.
In a telephoned statement yesterday from Israel, where she now resides,
Mrs. Shcharansky declared, "I welcome the opportunity to visit America. I
am extremely grateful for the invitation that has been extended to me by the
U.S. scientists. This will give me an opportunity to share with the American
people my grave concern over Anatoly's fate. I believe that coming to the
United States under the auspices of the American scientific community will
be extremely valuable in helping my husband."
Mrs. Shcharansky will visit the U.S. following a short stay in France.
She goes there at the invitation of Francois Mitterrand, First Secretary of
the French Socialist Party.
When she visits this country, Mrs. Shcharansky will be hosted by
members of a number of scientific research institutions and laboratories.
Among the American scientists who extended the invitation to Mrs. Shcharansky
were:
Owen Chamberlain, Physics, University of California at Berkeley
Malcolm Derrick, Physics, Argonne National Laboratory
Maurice Goldhaber, Physics, Brookhaven National Laboratory
Kurt Gottfried, Physics, Cornell University
Arthur Kornberg, Biochemistry, Stanford University
T.D. Lee, Physics, Columbia University
John McCarthy, Computer Science, Stanford University
Fred Mills, Physics, Fermi National Accelerator Laboratory
Burton Richter, Physics, Stanford Linear Accelerator Center
Joseph Weizenbaum, Computer Science, Massachusetts Institute of
Technology
American scientists have repeatedly expressed their concern over
Shcharansky's arrest and imprisonment. In letters, telegrams and
demonstrations of protest they have continually urged his release. In March
a group of Berkeley scientists issued a protest against the Soviet Union's
harassment and detention of Shcharansky. In that message, sent to
Soviet Secretary Leonid Brezhnev, they declared that as long as the
computer scientist was kept in prison, "we will find it dificult to
maintain open and generous sharing with our scientific colleagues in the
Soviet Union."
The Shcharansky trial follows the recent sentencing of another
Soviet scientist and dissident, Yuri Orlov. Orlov was sentenced to seven
years in prison followed by five years of enforced exile for alleged slander
of the Soviet state. Both Orlov and Shcharansky were organizers of the
Moscow chapter of the Helsinki agreement monitoring committee. (The
agreement contains broad humanitarian declarations concerning individual
civil liberties.)
American scientists were chagrined by the Orlov sentence. They
now fear that its severity marks an escalation in the Soviet disregard for
the rights of dissident Soviet scientists and are alarmed by what may happen
to Shcharansky who is accused of a much greater crime than Orlov.
Anatoly Shcharansky goes on trial on Monday for "high treason in
the form of espionage." The charge carries with it a possible death
penalty. This is the first time since the Stalinist era that such a
severe sentence has been considered. The 30-year old computer scientist has
been accused by the Soviet press of being an agent of the U.S. Central
intelligence Agency, a charge President Jimmy Carter has repeatedly denied.
For more information call Dr. Morris Pripstein at (415) 642-0466
or (415) 843-2086.
-------
∂09-Jul-78 1808 JMC Press release
To: robinson at SRI-KL
CC: joseph at MIT-ML
There are two phrases in the press release that can be taken in a
different sense from what is undoubtedly intended. The first sentence
says "in a show of concern". One common interpretation of that phrase is
that the concern isn't real - merely put on. Someone who reads the press
release will understand your sense, but if a newspaper reporter uses the
phrase in a story, it will read as though he considers our concern
feigned. The second phrase is the one containing the word "chagrined". A
person is chagrined if his plans have misfired. Soviet propaganda would
like to regard us as chagrined because our spy has been caught.
The first phrase could be replaced by "as an expression of
concern" and the second by "shocked" or "distressed" or even "horrified".
I agree with the intentions of the statement and will go along
with the wording, but I think it should be modified if there is still an
opportunity to do so.
I will ask Minsky and Fredkin if they are willing to adhere
to the statement.
∂08-Jul-78 1925 JMC bfloat
To: GLS
bfloat and fpprec fundamentally do the job. However, they should
be in LISP itself. However, a global variable for precision is
not in the spirit of LISP or any other recursive language. Rather
it seems that one should have a form fpprec(expression,precision)
which would evaluate <expression> using <precision> for its internal
computations except that inner calls to fpprec would control
the precision of the internal expression. <expression> itself need
not be a number at all; the form merely controls precision of anything
that occurs within the evaluation. The global value of fpprec would
still exist and would be the default value. Functional arguments would
presumably be used with the precision of the associated environment -
at least when other things are being give a lexical scoping.
∂08-Jul-78 1630 JMC
To: ellen at MIT-MC
How do I can version 9 of the Macsyma Reference Manual?
∂08-Jul-78 1455 JMC
To: joseph at MIT-ML
He must be the person who called me some weeks ago and who I agreed could
use my name on some statement. I'll call him.
∂08-Jul-78 1401 JMC
To: PLS
∀m n p.true Holds(0,K(SP,Not K(Name p,Holds(0,Not(Equal(M, Concept1 m)
And Equal(N,Concept1 n))))))
∂08-Jul-78 1336 JMC
To: joseph at MIT-ML
I replied to WEIZENBAUM. Can you retrieve it?
∂08-Jul-78 1100 JMC messages
To: WEIZENBAUM at MIT-ML
I received your message mentioning the alleged Medvedev letter.
I suspect I may have missed a previous communication, because I
don't know what help you are requesting - presumably names on
a statement. When you SEND me a message it only appears on the
screen if I am logged in. You must MAIL the message to be
confident of the result. I don't know Robinson but can try to
phone him Monday if that will help. As for Wiesner, I cannot
imagine that my influence with him is more than epsilon compared
to yours, Marvin's and Fredkin's.
Note that the Russians are perfectly capable of spreading
disinformation through unofficial channels to supplement their
official propaganda. The disinformation often seems more
reasonable than the direct propaganda. If you want to find out
whether Medvedev wrote such a letter, I suggest you phone him
in England. His American publishers would undoubtedly know
how to reach him, and your publisher would know how to reach
them.
I worry that Medvedev may be susceptible to disinformation,
because his story about the explosion in the nuclear waste dump
is scientifically implausible, and it reads to me like a cover
story. I.e. if an area must be evacuated after a mishap with
a bomb and nuclear weapons are absolutely secret, then the
people who aren't cleared to know about them but must help
with the evacuation have to be told some story. These people
are also sworn to secrecy and see themselves as having the inside
dope. It doesn't have to be fully rationally motivated in order
to happen that way.
∂08-Jul-78 0003 JMC dynamic precision
To: GLS
As far as I can tell from version 6 manual, bfloat offers only one extended
precision. Dynamic precision should allow any desired precision as
determined by the course of the program. Different variables may have
different precisions, and a program could decide to repeat a calculation
with a different precision. Also the facility should be in LISP - and
in the LISP manual.
∂07-Jul-78 1656 JMC
To: GLS
Let me borrow them.
∂07-Jul-78 1656 JMC
To: weizenbaum at MIT-ML
I know of no such letter unless you mean Lipavsky who didn't emigrate.
∂07-Jul-78 1639 JMC
To: GLS
Sorry, I'll have to read about that. Please give me a pointer.
∂07-Jul-78 1604 JMC Numerical analysis coup for LISP
To: GLS
If you have a bit of surplus energy, I have an idea for making LISP
a better language for some kinds numerical analysis than any that
the numerical analysts know and love. It is to make available
some form of dynamically controlled multiple precision floating
point, i.e. the program can decide with what precision it wants
to do subsequent operations on certain data. It requires some
thought to design it, but it probably isn't much work to execute
it. It would be a sort of coup.
∂07-Jul-78 1555 JMC
To: creary at SUMEX-AIM
I have the following Kowalski papers:
Kowalski, Robert\Predicate logic as programming language\UE - memo 70\Nov. 1973.
Kowalski, Robert and Patrick J. Hayes\Semantic trees in automatic theorem proving\
UE\Aug. 1968\theorem proving.
∂07-Jul-78 0023 JMC
To: MRC
1. I don't know how many of your points are worth mentioning in the renewal
proposal even in so far as I agree with them.
2. With Stanford's cut, we may not be far under budget.
3. I never expected to use operating systems' terminal service and
always supposed that Dialnet would require system modifications.
I also suppose that once Dialnet becomes attractive, manufacturers
would begin to implement it. It shouldn't require separate implementation
for each computer.
4. File transfer was always considered important.
5. While the immediate uses are point-to-point communication, so were
the immediate uses of the telephone. It will take a while after
Dialnet becomes popular, before it becomes common for people who
suddenly find themselves sharing an interest, will suddenly discover
that both of their installations have already implemented Dialnet.
∂06-Jul-78 2112 JMC
To: MRC, LES
See RENEW[S78,JMC] for a draft of Dialnet renewal proposal.
∂05-Jul-78 2024 JMC
To: LES
CC: DCL
General idea of DCL proposal has merit.
∂05-Jul-78 0025 JMC
To: LES
Take it away. FR[S78,JMC].
∂04-Jul-78 2321 JMC
To: LLW
It would be interesting to get together for dinner some time before
I go on my trip at the end of the month and discuss technological exotica.
∂04-Jul-78 2319 JMC
To: HPM
Dr. Rod Hyde, my Group's Advanced Design & Systems Studies Section Leader
(and the guy with whom I design fusion starships and other exotica from
time to time), was the one to whom you talked briefly re Hans Moravec's
rotating cable vertical lifter calculations on Saturday PM. He indicated
to me later that evening that he was looking forward to a detailed
discussion of Hans' thinking and calculations with him. Rod was ranked as
one of the most capable graduates ever of MIT's Aero-&-Astro Dept. when he
got his degree there 2 years ago, and I think Hans might find a discussion
with him to be quite productive.
∂11-Jul-78 0048 MRC dialnet
To: JMC, LES
well, i fixed whatever bug was causing my free storage problem, so a file
transfer no longer crashes the system; however a file transfer still puts
an extra record at the end and apparently any retransmission (caused by
missed packets) cause it all to get deadlocked. better than before though.
∂11-Jul-78 0151 JMC
To: CLT
CC: JMC
******* PAGE 3 MAIL$E.TXT[LET,JMC] ******* W U
Iterative computation sequence
Given a recursive program, we have considered a derived program that gives
a computation sequence. Consider a program that gives a trace, but allows
a program to call itself or another program at the outer level, i.e. go to
it. The trace shows the call at the same level of indentation as the
calling program and shows only one return. I suggest you try writing such
a derived function for a few examples and then a program that generates
the derived program.
***** Arrow at Line 2 of 9 ***** Page 3 of 3 ***** 1R -135C *****
∂11-Jul-78 1103 HVA Continuation of Your Appointment
CC: FWH, DCL, JMC, HVA
The matter is now resolved, I believe, and I will call Mary Clark at Bechtel
International Center re your visa.
∂11-Jul-78 1353 TOB
John
I received a call today from Ward McClure at TI. He said that they can
wrap up arrangements for the system with a condition. Heilmeier who would
be supporting it, wants a letter indicating that we intend to implement
LISP on the machine. The statement was about some language in which both
might do research, but he wants LISP, strongly. Initially, they thought
they could do it through TI foundation, but that is too much money for
them, so it must come out of someone's overhead, Heilmeier's.
Yes, it is George Heilmeier. An interesting quirk of fate I thought you
might enjoy.
The machine could be useful as a second arm controller or as a vision
machine which would provide fast response for demos. Do you think that it
is a good idea to make a statement that we will seek a student to
implement LISP? There is a PDP11 LISP. I am reluctant to do so because
we may have delays in finding someone to do it who would not be diverting
effort from research. I do favor making some contacts with TI because
they are moving strongly in robotics and will benefit the field. If we do
it, I want to renegotiate the machine configuration.
They are going into large scale production of their version of the
Scheinman arm.
I don't think LISP makes sense on the machine unless it has at least 128K
16 bit words. If they will pay for the student, I suppose one could be
found, but one couldn't promise them that it would be highest priority for
the best student.
∂11-Jul-78 1612 PLS Hayes' paper
The last page of the references was missing from the copy
you gave me - I had meant to tell you but forgot.
∂12-Jul-78 1003 Tajnai at SUMEX-AIM Principal Dissertation Advisor
Date: 12 Jul 1978 1001-PDT
From: Tajnai at SUMEX-AIM
Subject: Principal Dissertation Advisor
To: JMC at SAIL
cc: DPB at SAIL
I am holding the G-81 (Ph.D. Reading Committee) form for
Don Gennery. He has listed Tom Binford as Principal Dissertation
Advisory.
My question: What is the status of the request from you to Prof.
Feigenbaum requesting that Tom Binford be approved
to serve in this capacity?
Please advise.
Carolyn Tajnai
-------
I hereby make that request if I haven't formally made it already. JMC
∂12-Jul-78 1308 JB WISE MEN.
My first try on the proof of the wise men is disappointing:
the gain is about 0%.
The reason lies in the form of the axioms:
my goal structure is geared towards grinding goals into subgoals and it
does only a liminted combinatorial use of the facts needed towards the proof;
the several major goals (mile stones) in the proof of the wise men are very
simple (sometimes the goal is just "FALSE"), and the proof rests on making
heavy combinatorial use of the facts (axioms); the nature of this use, in this
particular proof, does not match the limited possibilities of combining facts
in my goal structure.
I did only minor changes to your axioms, namely those required to quantify
over "real persons" and thus express the axioms "initial" and "elwek" more
concisely. These changes weree not sufficient to modify the nature of the proof.
In order to obtain substantial gains with my goal structure, it will be necessary
to drastically modify the axioms, while still keeping the idea of having a "fool",
an accessibility relation and a "real world".
I'd like to talk with you before proceeding.
∂12-Jul-78 1417 RWW meta stuff from A. Bundy
I have put copies of the stuff he sent me in you mail box. I believe
the proofs are wrong but have not yet finished figuring out
what is there. The idea seems to be in the right direction.
rww
∂13-Jul-78 0226 KEN at MIT-AI (Kenneth Kahn) Your paper ''Ascribing Mental Qualities to Machines''
Date: 13 JUL 1978 0242-EDT
From: KEN at MIT-AI (Kenneth Kahn)
Subject: Your paper "Ascribing Mental Qualities to Machines"
To: jmc at SU-AI
CC: KEN at MIT-AI
I enjoyed your paper and have only one suggestion and one complaint. The
complaint concerns the part about how a robot could not predict its own
future and it might just be that I'm confused. You sketch a proof that if
it could predict its own future then it could disobey its prediction and
this contradiction leads one to think that it could not therefore predict
its own future. Also you argue that its simulation of itself must in
general be slower than itself. I assume you are talking about perfect (as
opposed to some approximate) prediction or simulation. Clearly if this
robot plans to simulate himself and then do otherwise he must take into
account this simulation and an infinite regress occurs. (If the
simulation was approximate then at each level of simulation might be
simplier and could converge I guess.) The robot could accurately predict
how he would act if he did not have the ability to simulate himself, but
then so what if he acts otherwise? Regarding the speed of the simulation,
I find it a little hard to think in terms of Life configurations but lets
suppose that the robot was a Lisp program instead. If the program is
running interpretively and wants to know what it will do then it should be
able to compile itself, run that for a while and see the results before he
himself could do it. Would it be cheating for a human to build an
accurate simulation of himself and run it some super-fast computer? I
personally would like to have seen a little more about why you might think
that certain mental qualities such as "humor and appreciation of beauty"
seem much harder to model. Do you think that something of greater
complexity is going on there? Or that such things are qualitatively very
different processes than having beliefs or solving problems? Or is it
just that we are more ignorant of what is going on there compared with
problem solving? Also I was a little surprised about your comments about
creativity. There seems to be more to creativity than the introduction of
elements not present in the problem statement. Somehow the process of
making choices is essential. I fail to see how ascribing creativity is
any easier than ascribing understanding to a machine. ---ken kahn
My argument is just a slight sharpening of the usual argument for the
unsolvability of the halting problem for Turing machines. That argument
says that if a Turing machine could solve the halting problem (or any
whether a given symbol will be printed, etc.), a related machine could
answer the question about what a machine does when it sees its own description
and express its answer by not halting when the object machine halts
and halting otherwise. The contradiction comes when we ask this machine
to look at its own description. My sharpening is merely to change the
question to that of whether the machine halts in n steps and remarking
that if the answer could be obtained in o(n) steps, the same contradiction
could be obtained by modifying it give the answer suitably. Most likely,
I should spell out the proof in more detail.
Incidentally, there is no infinite regress, because the machine
can use the same description twice - once as a description of the machine
and once as a description of the description. I just remembered that
my LISP class notes contain a LISP version of the unsolvability of
the halting problem that is trivial to follow for a LISP fan. If you
don't have access to a copy of the notes, I can send you one - or at
least the relevant chapter.
Having the program compile itself doesn't help, because the
behavior to be predicted has to include compiling itself, the execution of
the compiled program and getting its output.
My opinion (guess) about humor, etc. is that we merely don't
understand them. Anyway I don't have any theories.
I would be interested in any ideas about what more there is to
creativity than introducing new elements. However, if I seemed to claim
that creativity is easier to treat adequately than understanding, then I
withdraw the remark. I really have no opinion.
Many thanks for your comments, which are some of the most cogent I
have received. By the way, did you forget to mention your suggestion?
∂13-Jul-78 0120 LGC via AMES-TIP#57 Problem Solver
I have found it impossible to refrain from thinking about the problem
solver project, and the more I think about it, the more interested and
enthused I become. I would like to get your reactions to some of my ideas
about the problem solver (especially those that I have committed to writing),
and ask some questions about the etiquette (and/or politics) of requesting
information about other people's programs (including the code itself).
Accordingly, I'll appreciate it if you'll have a look at the file
PRBSLV.TXT[1,LGC] sometime soon, so that we can arrange to talk for a half-hour
or so this afternoon or tomorrow.
Well, I am pleased and flattered that you want to go in the direction of
the Advice Taker, and moreover, I agree. I agree with most of your
comments, but I have reservations about frames with slots at least if they
are intended to be used in the Minsky or KRL way. The problem is that
Minsky frames assume that what to do is determined by a main frame giving
pre-existing knowledge about what goals to pursue in a given type of
situation, and I am skeptical about this. Two days ago I received two
papers from Pat Hayes that you might find interesting. One of them is
"The logic of frames", and the other is entitled "The naive physics
manifesto".
∂13-Jul-78 1259 MRC Dialnet
To: JMC, LES
Dialnet has successfully transfered files with no errors. I left it up
for IAZ to work on FTP and he tells me he got pretty close to getting a
file transfered using the FTP protocol. Unfortunately, a bug in my
routines crashed the system, and I turned Dialnet off. The bug is fixed
in the source, so we'll try again tonite. -- m
∂13-Jul-78 1255 DCL Stanford -1 project
To: JMC
CC: LES
John, This project has been allowed to grow to monster proportions, over
30 people, and includes undergraduates, M.S. students and (hearsay) girlfriends.
The data disc channel problem is back again despite autologout of 15 mins.
I suggest we go ahead and implement the LOGIN EXCLUDE LIST feature I suggested
earlier. At any time yesterday this would have saved approx 4-5 channels
which was about the length of the queue.
∂14-Jul-78 0431 DCL
To: GROUP.DIS[VCG,DCL]:
***********************************************************************
JOINT SEMINAR
CONCURRENT SYSTEMS
and
VERIFICATION GROUP
TUESDAY 18th JULY
PLACE: ERL 237
TIME: 2:30 p.m.
TITLE: Distributed Termination of Parallel Processes
SPEAKER: Nissim Francez
U.S.C.
ABSTRACT
Several possible approaches to the termination of concurrent processes
will be pointed out. Distributed termination will be defined and the
pattern of such termination will be characterized. An algorithm to
achieve distributed termination will be shown and applied to an example.
∂14-Jul-78 0824 Feigenbaum at SUMEX-AIM AI Qual Exam for 1978-79 academic year
Date: 14 Jul 1978 0818-PDT
From: Feigenbaum at SUMEX-AIM
Subject: AI Qual Exam for 1978-79 academic year
To: winograd at PARC, jmc at SAIL, buchanan, lenat at CMUA
cc: feigenbaum
This note is just to remind everyone that Terry has volunteered to run
the AI Qual for the coming year and is hereby without further reminder
authorized to proceed at time and style of his choosing.
Re the four students who failed last time, I may want to redo one-hour
oral exams on one or more of them later this summer. More about that later.
Ed
-------
∂14-Jul-78 1250 RWW
John,
I have put my notes from PISA in your mailbox.
These notes contain the examples I refered to the other day.
A) Note 1 and 2 are about logic and the only thing that might be
necessary for an understanding of what follows are sections 8
and 9 on meta theory and reflection.
B) Notes 3 and 4 are a comparatively good description of the features
of FOL. Everything from section 4 on has not yet appeared in print.
C) Things of particular interest
1) example of syntatic simplification sect 6.1
2) solving linear equation example sect 10.1
3) axiomatizing heuristic search sect 10.2 (boyer and moore example)
D) minimization schema example is in note 6. It is worked out in
complete detail.
richard
∂14-Jul-78 1533 CARLSON at USC-ISI Re: Request for meeting
Date: 14 JUL 1978 1532-PDT
From: CARLSON at USC-ISI
Subject: Re: Request for meeting
To: JMC at SU-AI
cc: CARLSON
In response to your message sent 14 Jul 1978 1308-PDT
My plans have changed. I will now be at Standord Stanford on
Monday 20 July. I would like to come to the AI Lab on Monday
afternoon from 2-6 and discuss the proposal for next year.
I hope all the group leaders can be available for at least
part of that time.
My schedule is still not 100% firm,. I'll
get back to you immediately if the trip is cancelled again.
Bill
-------
20 July is not Monday. Is it 24 July you mean? If so, fine. I
hope we can see you sometime in July, because I will be gone
for a month starting early in August.
∂15-Jul-78 0617 DEK dinner tonite
Yes, thanks very much. What time? Should I bring some wine or something?
∂15-Jul-78 1502 JMC*
CSD affairs
LISP.PRO[S78,JMC] 22-Apr-78 Memo on changing catalog description of LISP
course.
CATALO[S78,JMC] 28-Apr-78 New catalog descriptions of my courses
VERIFI[S78,JMC] 14-May-78 Propaganda for verification in CS education
circulate verifi[s78,jmc]
∂15-Jul-78 1627 RAK Computer usage JULY 1-14--S1 takes off
To: DCL, JMC, LES, TOB
Computer usage for July 1-14 (note S1 is cleverly hidden under
three different projects; this is their sum)
ALL USERS 12.216 100%
MUSIC 2.358 19.3%
HAND-EYE 1.846 15.1%
S1 1.468 12.0%
PROG VERIF 1.371 11.2%
M222 1.194 9.8%
COMP SCI .848 6.9%
SPEECH
MODELLING .637 5.2%
FORMAL
REASONING .466 3.8%
DIALNET,
NAT LANG,
MTC .579 4.9%
ALL OTHER
NON-ARPA,
NON-NSF 1.449 11.8%
Restated:
Hand-eye, Prog verif, Formal reasoning, dialnet, nat lang, MTC:
4.262 doubloons or 35%
Others: 7.954 doubloons or 65%
I think this explains the tremendous waits and high afternoon load
averages.
Solution: Get HPM to make more movies, so that Hand-eye becomes bigger.
Dick
∂15-Jul-78 2028 MRC Dialnet proposal
To: JMC, LES
You can also say in it something like:
It turned out that implementing a suitable line transmission
protocol was the most difficult problem faced. Almost immediately
after the completion and debugging of the line transmission protocol
code in the system, file transfers were being done. An early
program, CRUFT, was in fact used in the debugging of the system
code! A few days later, the official Dialnet file transfer
implementation successfully transmitted a file with no errors. Work
now is being done on tuning this implementation so that it is suitable
for users, adding new features, etc.
At this time work has started on installing Dialnet at LOTS.
∂16-Jul-78 0152 MRC Dialnet protocol documentation
To: JMC, LES
I have PUBified some of the Dialnet protocol documentation. XSPOOL
PROTOC.XGP[DLN,MRC] for your own copy. PLEASE READ THIS!!!!!!!!!
∂16-Jul-78 1613 MRC runaway REMUX
To: HWC
CC: LES, JMC
I have killed a REMUX which was accumulating 2+ hours of runtime. It
was detached and had nobody on its PTY's, so it appeared to be a runaway.
I would have left it for you to debug, but RAK was breathing down everybody's
shoulder and I wanted to avoid making trouble for everybody concerned.
-- m
∂16-Jul-78 2257 ARK S1 Meeting
To: S1
CC: Pickering at SUMEX-AIM
The next S1 CS Group meeting will be on Thursday, July 20 in the small
conference room at SAIL at 1330.
Arthur
∂17-Jul-78 0201 HWC REMUX
To: RAK, DCL, PMF, LLW, JMC, LES
CC: PMF, LCW, MRC, BH, ME, JBR, HWC
To begin with, the metering data given by WHO is incorrect. Times from
TIME indicate a CPU load of less than 1/2 % when REMUX is actively doing
output. When it is detached, the worse it does is to wake up every two
seconds to check the output buffers, a matter of less than a millisec.
The two plus hours of CPU time indicated by WHO represented usage from
before 2200 on Jul 15 and was extremeley overstated. Before drastic
measures, the actual time as measured by TIME should be checked.
Under the current situation, it would be correct to kill a detached REMUX
since we are currently using a dial-up line as a development testbed.
Later we will require a detached REMUX around at all times as a phantom
process to service the dedicated line to Livermore. It will serve to
support multiple remote terminals and file-transport to the S1 project at
Lawrence Livermore Lab over a single high-speed line. This process is
meant to run on the primary SU-AI machine and not offloaded onto any
other. In any case, the load caused by REMUX is not the reason for the
general slowness of the system during prime afternoon time when the
complaints were registered.
Less flaming on these issues would occur if the metering data available to
the general community were as accurate as that available to those
knowledgeable about the system. There would be much less likelihood of
there being a randomly detached job if the line scanner did not randomly
decide that lines are disconnecting.
REMUX will continue to be improved and optimized so it can provide the
services required on the S1 project with as little impact as possible to
the rest of the system. Please be assured that it does not and will not
take anywhere the amount of resources indicated by WHO and that it
performs a useful and required service running on SAIL.
----------------
∂16-Jul-78 1428 RAK NREMUX
To: DCL, LES, LCW, HWC
I take it there is some very important reason why right now
NREMUX is detached in the RUNQ with 2:06:13 processor time and
is slowing down everyone else's line editor and E, while continually
using from 20-35% of the processor.
I suggest that if this important reason continues, we find some other
machine for NREMUX.
Dick
----------------
∂16-Jul-78 1613 MRC runaway REMUX
To: HWC
CC: LES, JMC
I have killed a REMUX which was accumulating 2+ hours of runtime. It
was detached and had nobody on its PTY's, so it appeared to be a runaway.
I would have left it for you to debug, but RAK was breathing down everybody's
shoulder and I wanted to avoid making trouble for everybody concerned.
-- m
∂17-Jul-78 0216 RAK REMUX--last comment from me, promise
To: DCL, LES, JMC
The system reported repeatedly that the % of the processor being used
by REMUX ranged as high as 40% and never got below 5%, averaging 25% for
almost all of the many times I looked. I am glad to hear that is a lie;
it is probably a lie for the verifier too, which never uses any real CPU
time at all. Let us adjust the who function so that knowledgable people
will see that.
I am very glad to hear that REMUX will use less overhead so that it can
support multiple remote users and file transfer from LLL. We really are
underutilized now.
We could solve our problems by running program verification, hand-eye,
FOL, etc., over at LOTS. Unfortunately, LOTS is reserved only for unsupported
research, but if we keep giving away our current machine, that probably won't
be a problem for long.
Dick
∂17-Jul-78 0230 MRC forwarding mail without sender's consent
To: HWC
CC: LES, JMC
I wish you had not forwarded my message to all those people. That
comment about RAK wasn't intended to go any further; I certainly
didn't want to embarass him to his face like that. The fact was
that RAK was complaining to anybody who would listen about this
and how this "random" was "stealing time" from his work (of course,
taking infinite time doing all this complaining, running WHO, etc.).
When he announced he was going to kill your job, I talked him out
of it and got him to try to call you; when that failed, I looked at
what it was up to myself, killing it only after I determined that
it was detached and nobody was on its PTY's, indicating a probable
runaway.
I certainly am not passing judgement on REMUX or its validity on
this system. In fact, I consider it to be important, since it
facilitates the S-1 group's usage of this system, and since it is
a data communications program (and that IS my area), I have an
interest in helping you out; also, since I believe that Dialnet is
exactly what you want for it, that is even more of a reason.
-- m
∂17-Jul-78 1433 JB ORALS.
The last date for Orals in the Summer Quarter is Thursday, August 10.
Scheduling must be three weeks in advance, that is this week.
I'd like to be able to do my Oral this Summer. I could produce one week
in advance to the chosen date a more comprehensive report than the one you saw,
and that could be a first draft of my thesis. Please let me know whether
I can schedule; if yes, please let me know of one or two convenient times
around the 9th or 10th of August.
∂17-Jul-78 1550 LGC via AMES-TIP#20 Problem Solver
I was very happy to hear that you approve of the basic direction I
want to take in the problem solver project. Though I didn't realize it
at first, I have gradually come to see that this project and I are
tailor-made for each other in very many ways. My enthusiasm for this
particular endeavor (well-founded, I believe) is now unbounded.
I have added sections to the first page of PRBSLV.TXT[1,LGC] on
"Motivation" and "Relation to FOL," and have also added a new major
design goal on the same page. Thus, once again I would like to have
your reactions, this time to the new things.
By the way, I am puzzled by something you say in column 1 of
page 5 of the ARPA proposal: "... mathematicians have proved the
completeness of the rules of inference of first-order logic; admitting
any general laws that generate conclusions not attainable by the
old laws makes the system inconsistent." This remark would be
appropriate if the predicate calculus were complete in the syntactic
sense that the addition of any non-theorem as an axiom leads to
inconsistency. However, the predicate calculus provably is NOT
complete in this syntactic sense (though the propositional calculus
provably is); rather the predicate calculus is complete in the semantic
sense that every valid wff is a theorem. Thus, your remark seems wrong.
Have I missed something here?
I'll stop by sometime soon to meet Richard and ask you those
questions I mentioned in my last message.
∂17-Jul-78 1648 LCW S1FORT Meeting
To: S1
Gio suggested that the S1FORT meeting not be held at 1PM. We will make it
the last meeting on Tuesdays and Fridays, ie. at 4PM, starting tomorrow.
Curt
∂18-Jul-78 0310 MRC PROTOC.XGP[DLN,MRC]
To: JMC, LES
Please print this file and read it. This is the thing I think should go
in the proposal renewal.
-- m
∂18-Jul-78 1402 LES
∂18-Jul-78 0705 GEORGE.ROBERTSON(A110GR02) at CMU-10A ARPA
Date: 18 Jul 1978 1003-EDT
From: GEORGE.ROBERTSON(A110GR02) at CMU-10A
Subject: ARPA
To: earnest at SU-AI
Les: Since we last met, I have decided not to accept the position
at ARPA. The main reasons had to do with the tradeoffs between
research and research management. I feel this is the wrong point
in my career to move to a management position. So, I will be staying
at CMU in a research position.
I don't know which program manager at IPTO will be handling your
renewal. You should talk with Col. Russell about that.
I just received an advance copy of your proposal which I will return
to you. I wish you luck with it.
George
∂18-Jul-78 1615 Hart at SRI-KL (Peter Hart) Invitation
Date: 18 Jul 1978 1617-PDT
From: Hart at SRI-KL (Peter Hart)
Subject: Invitation
To: Sacerdoti, Tenenbaum, Hendrix, Nilsson, Hart,
To: Feigenbaum at SUMEX-AIM, Buchanan at SUMEX-AIM,
To: Lenat at SUMEX-AIM, McCarthy at SU-AI, TW at SU-AI,
To: Goldstein at PARC, Bobrow at PARC, CCG at SU-AI,
To: Fikes at PARC, Kay at PARC
cc: Denicoff at USC-ISI
As you may know, Marvin Denicoff of the Office of Naval Research will
be visiting California later this summer. One of his purposes is to
hold an invited 2 day workshop to discuss future possibiltities for AI
research. The planned emphasis of the workshop is on brainstorming
with the aim of creating promising new directions, rather than on
reviewing and critiquing past work.
The workshop has been scheduled for August 28 and 29 at SRI. I
appreciate that the notice is rather short, but because of the
probable importance of the meeting (not to mention the intellectual
excitement), I hope you will find it possible to attend.
Please respond at your earliest opportunity to Nils Nilsson (Nilsson@
SRI-KL.) Additional suggestions for Northern California AI invitees
are welcome, but note that the intention is to limit the number of
attendees rather severely in order to encourage a free-wheeling
discussion.
Peter Hart
-------
I shall be in Japan on the 28th and 29th, although I would have
liked to attend the meeting. Should it have to be postponed for
some reason, please keep me in mind.
∂18-Jul-78 1646 LCW S1 CS Meeting Times
To: S1
The S1 CS Group meeting times have been changed again to
S1FORT 2PM
SOPA 3PM
MANUAL 4PM
Curt
∂18-Jul-78 2001 MRC mailed without comment
To: JMC, LES
a292 1944 18 Jul 78
AM-Mrs Shcharansky,750
By LAWRENCE L. KNUTSON
Associated Press Writer
WASHINGTON (AP) - Mrs. Anatoly Shcharansky and members of the
American scientific community disagreed publicly Tuesday over what
tactics to use to free her husband and other jailed Soviet dissident3.
The disagreement came as Mrs. Shcharansky continued to appeal
directly to government officials and influential Americans for
official action putting pressure on the Soviet Union so it will be
compelled to commute her husband's 13-year jail sentence.
She met in a House committee room with a several members of Congress
and a panel of scientists on Tuesday to discuss methods of exerting
that pressure on behalf of her husband, a computer programmer and
scientist who was convicted of treason.
But several representatives of American scientific associations said
any pressure from the scientific community should come through
individual efforts and not by what they called a destructive
curtailment of scientific exchange programs by the U.S. governnment.
They said they believe it would be more productive to exert pressure
by curtailing economic and commercial activitiies.
One scientist, Dr. John McCarthy of the Department of Computer
Science at Stanford University, suggested that the jailed dissidents
be ransomed in exchange for U.S. technology desired by the Soviet
Union.
''We can trade material concessions on our part for moral
concessions on their part,'' he said.
Speaking through an interpreter, Mrs. Shcharansky said the 1980
Olympic games should be shifted from Moscow to a city in the West.
She also said it would be wrong to sell advanced technology such as
computers to the Soviets because they might use such devices to
monitor dissidents. And she said selling oil drilling bits to the
Soviet regime ''would be a form of reward and ignoring their inhuman
acts.''
But she found little support from the panel of scientists when she
called for re-examination of scientific exchange programs with the
Soviet Union ''to evaluate whether they are positive programs for this
country in the current climate.''
''It seems absurd to give the Soviet Union the benefits of American
science when they mistreat their people so cruelly,'' she said.
Several scientists said that while they approved of the threat of
non-cooperation with the Soviet Union by individual American
scientists, they oppose government intervention.
Such a threat is contained in a so-called ''Declaration of
Conscience'' signed by about 400 scientists at more than 20 U. .
educational and scientific institutions and prepared for delivery to
Soviet President Leonid Brezhnev and the Soviet Academy of Sciences.
The signers pledged to ''withdraw all personal cooperation with the
Soviet Union until Anatoly Shcharansky and Yuri Orlov (another jailed
scientist) are released from prison.''
But Jeremy Stone, director of the Federation of American Scientists,
contended that the scientific exchange program must be preserved and
cast doubt on the effectiveness of any action to free the dissidents
or to affect Soviet policy.
''Whatever threats are made to the Soviet Union, they should not be
so naive as to think that the Russians will stop being Russians, that
the Russians will let civil liberties hold sway in Moscow or permit
all Jews who wish to leave to do so,'' he said.
''With all respect, I disagree with you,'' Mrs. Shcharansky said.
''To say that the Russians will always be Russians and cannot be
changed reminds me of the attitude of the West toward the Nazis in
Germany before World War II.
''You say that we want revenge,'' she said. ''... It is not
retaliation which we are planning. I don't think my husband is
concerned with revenge. He just wants to be free. ... Do you help him
in that? Or do you only want to talk to him about science?''
Stone replied that it would be self-defeating to cut Soviet
scientists off from all Western contact and influence but said he does
not mean ''that the situation cannot be improved in the Soviet Union
under pressure.''
''For example, 20 years ago Shcharansky would have been shot,'' he
said.
Meanwhile, Sen. Charles Percy, R-Ill., delivered a letter to the
White House on Tuesday calling on President Carter to request a
meeting of allied foreign ministers to discuss the status of the
Helsinki agreement.
Percy said the Shcharansky trial had ''ripped the fabric'' of the
human rights provisions of the accord, which was signed by the
Soviets, raising questions about whether the Russians would abide by
other provisions of the agreement.
Percy said he gave the letter to national security adviser Zbigniew
Brzezinski, who promised to pass it on to Carter. Percy would not
describe Brzezinski's reaction to letter.
ap-ny-07-18 2247EDT
***************
∂19-Jul-78 1201 Creary at SUMEX-AIM Appointment Request
Date: 19 Jul 1978 1202-PDT
From: Creary at SUMEX-AIM
Subject: Appointment Request
To: JMC at SU-AI
Will you have some time to talk if I stop by your office at 2:00 this
afternoon? A half-hour or so should do. Maybe I'll also be able to see RWW
if he's around.
-------
Just got here. Now or tomorrow ok?
∂19-Jul-78 1306 DCL
To: LES
CC: JMC
Can we get a quick decision on EXCULDE LIST implementation now?
We are going to do it.
∂19-Jul-78 1641 MRC merge of opcode and channel?
What do people think of merging the opcode and the channel number into
one byte (making each field 4 bits)? That would limit each to only 16,
but that is more than enough channels (FTP only uses 3) and opcodes
(only 9 are assigned now, and no more are anticipted).
This has an advantage: the packet header would one storage word on both
the pdp10 and any ibm implementation, hence the data area would start on
a word boundary. This would allow for much more efficient implementations.
∂19-Jul-78 2136 MRC ARPAnet directory
To: ARPDIR.DIS[S,SYS]:
People:
You are currently listed in the ARPAnet directory as belonging to SAIL.
You can see your entry by saying
FIND xxx IN ARPA.DIR[NET,MRC]
where xxx is your SAIL mailbox. If you wish your entry deleted or
modified, please send me a message. You need not send me a message πf
it's okay; I will default to leaving everybody in unmodified unless
told otherwise.
∂20-Jul-78 1204 DCL Arpa Proposal
To: LES
CC: JMC
A copy of our current ARPA proposal has found its way onto open reports shelf
in the CS library. I only hope it has not been mailed out to competing
institutions.
I object very strongly to my research plans being put on open library shelfs
and I am asking you to request the library to withdraw it and return it.
-David
∂20-Jul-78 1356 MRC
∂20-Jul-78 1221 JMC
Manna, Binford, Weyhrauch and Luckham should be in Arpanet directory.
mrc - they are and have been in.
∂21-Jul-78 1041 TOB music 222
To: JC, JMC
As I understand, music 222 is to be off the
machine at 9am. At 10:45, there are 15 jobs
of music 222 and a 15 minute wait for DD channels.
∂21-Jul-78 1226 WINOGRAD at PARC-MAXC2 Splenitude
Date: 21 JUL 1978 1111-PDT
From: WINOGRAD at PARC-MAXC2
Subject: Splenitude
To: mccarthy at SAIL
cc: winograd
I don't remember exactly, but my vague recollection is
that it was intentionally meaningless -- that you
started to set up a system of interactions in which
it played a role with nobody ever having a definition.
The discussion in general reminds me of "if a tree
falls in the forest and noone is there....". It
seems plausible to think of "owe" like "noise" in
having an inescapable component of perception
(or in this case thought). As you point out, the
possessor of this thought could be a computer
rather than a person, but the owing doesn't exist
until the thought.
-t
-------
∂21-Jul-78 1231 Blum at SUMEX-AIM New discussion group on medical computing applications
Date: 21 Jul 1978 1205-PDT
From: Blum at SUMEX-AIM
Subject: New discussion group on medical computing applications
To: siglunch:
A new monthly lecture series / discussion group on computers in
medicine will be starting on August 3, 1978 at 8:00 p.m. These meetings
are tentatively scheduled for the first Thursday of every month and will
be held in my home at 2781 Ross Road in Palo Alto.
At the first meeting Larry Fagan will present his research on rule-
guided management of mechanical ventilation done in conjunction with the
pulmonary research group at Pacific Medical Center.
The goal of these meetings is to provide an opportunity for
interchange among the many research groups developing medical computer
applicatiions. More detail on some of these groups exists on
<blum>club.announcement.
For further information please contact me via BLUM@SUMEX.
Bob Blum
-------
∂22-Jul-78 0310 DCL
To: GROUP.DIS[VCG,DCL]:
***********************************************************************
JOINT SEMINAR
CONCURRENT SYSTEMS
and
VERIFICATION GROUP
TUESDAY 25th JULY
PLACE: ERL 237
TIME: 2:30 p.m.
TITLE: Nondeterminism and Concurrency
SPEAKER: Larry Flon
S.R.I. and U.S.C.
ABSTRACT
In this talk we will explore properties of nondeterministic programs
that seem to be of use in treating the correctness of parallel
programs expressed in nondeterministic form. We define a number
of correctness criteria for nonterminating nondeterministic programs,
including Invariance, Potentiality, Inevitability, Blocking, Deadlock,
and Starvation. A system of proof rules for these properties exists.
We have shown the consistency and completeness of the system by
arguing about weakest pre-conditions and extremum fixed-points.
We will present several examples of parallel programs that can
be successfully proven in our system. Hopefully we can stimulate
some discussion of the advantages/disadvantages of the nondeterminism-based
approach.
TIME: Please will the following gentlemen be punctual:
DCL,RAK,SMG,LCP
∂22-Jul-78 0556 MRC Datacomputer availability
To: LES, JMC
Date: 22 JUL 1978 0852-EDT
From: DEE at CCA
Subject: Datacomputer
I am not quite sure of just what kind of statistics you want. Some
are in our Technical reports and I could send off a copy of the
report covering 1977 if given a US Post Office address.
Assuming that you are interested in long term reliability, availability,
and capacity I would say the following:
1) The Datacomputer reliability so far has been "perfect" in that, so
far as we know, we have never lost a user bit.
2) On availability, I recently sent the following off in a letter to
someone else on this subject:
"For the six month period from December 1, 1977, through May 31,
1978, there were 73 intervals of unavailability during prime time [9AM
to 8PM weekdays: the Datacompter is generally available outside these
hours but is not guaranteed to be attended]. Total prime hours unavailable
were 194.5 or an average of 2.7 hours each. However, this includes the
time during wich a state of emergency was declared in Massachusetts due
to the Great Blizzard of 1978 and its aftermath. Ignoring the blizzard,
which was the longest service interruption we have had, reduces the average
prime unavailability interval to 1.6 hours. During these same six months,
our percentage of prime time available was as follows: December 90%,
January 96%, February 86% (ignoring blizzard), March 95%, April 88%, and
May 91%."
3) On capacity, we have one tape, with over 40 billion bits of capacity,
available for general use and it's about half full. However, we do
not currently re-use areas of tape for a file other than the one
originally written there and we make back-up copies on the same tape
under some circumstances. Periodically we do a compacting copy to a
new tape and store the old one for back-up. If the general tape got
too full we would be spending all our time doing compacting copies. Of
course, the TBM hardware can be potentially expanded up to 64 tape drives
which would be a few trillion bits.
-------
∂22-Jul-78 1624 LGC via AMES-TIP#50 Oral presentation
Reflection has made it plain that I will indeed be able to give a decent
oral presentation for the ARPA site visit on Monday. Shall I go ahead
with the preparations? Later this evening there will be a new version
of my ARPA appendix for you to look at. It will contain some changes
from earlier versions you have seen, and also some new things.
Yes, please.
∂22-Jul-78 1649 LGC via AMES-TIP#50 Message Addendum
Forgot to mention in the earlier message that I have given hour-long oral
presentations on Meta-DENDRAL at both the RAND Corporation and the IBM
Watson Research Center.
∂22-Jul-78 1728 MRC APRAnet directory
To: OFFICE.DIS[1,MRC]:
You are currently listed in the ARPAnet directory with your office phone
as the phone number instead of the lab phone number. Please tell me if
you want the lab phone number instead.
Note that LES says that in the new building, there will be no paging
system and there will be secretarial coverage of all line, hence using
a lab number will be definitely sub-optimal there.
If you're happy with it the way it is, just ignore this message.
∂22-Jul-78 2352 LGC via AMES-TIP#54 ARPA Appendix
Please read over everything on PRBARP.TXT[1,LGC], excepting possibly the
Bibliography. There are several changes, and some possibly
controversial new additions. I would appreciate hearing, by Sunday
noon, about any changes or additions that you think are necessary or
desirable. Sunday at noon I plan to make the file available to LES for
PUBbing. If you'll give me the complete reference to Hayes's paper
"The Logic of Frames," I'll put it in the Bibliography.
Also, I'll appreciate some indication of the desired length of my oral
presentation in minutes. Are hand-drawn transparencies ok, or should
they be xeroxed from XGP copy? As to content, I am simply planning to
summarize the ARPA appendix, perhaps adding a few examples of some of
the abstract concepts. How many ARPA people will be in the audience for
the oral presentations?
Today's work has caused me to see that Nilsson's deduction system is not
the first thing that I should be planning to implement; that would be
something like putting the cart before the horse. What I now think I
should implement first is the top-level problem solving program itself,
using the example of getting to the airport provided in your paper as
the first test case for a paper simulation. It won't be clear 'till
later exactly what deductive capabilities are needed. Any comments?
Please plan for 20 minutes, and hand-drawn transparencies are fine.
So far as I know, Carlson will be the only person from ARPA, and
there is not guarantee that he will be enormously interested. It will
depend on whether he is taking over the supervision of basic AI or
is merely filling in. (George Robertson, from CMU, who was scheduled
to do it, decided he didn't want to work for ARPA after all).
∂23-Jul-78 0421 LLW Dinner
To: JMC
CC: LLW
I'm looking forward to dinner with you, Hans and Rod. Unfortunately, I can't
make it tonight, as I've already accepted another invitation. We're expecting
to be moving in to our new place this coming weekend, and will be having our
next party within a week thereafter (as I mentioned in my last note to you)--
does this stretch matters out too much? Lowell
∂23-Jul-78 1246 LGC via AMES-TIP#53 Problem Solver ARPA Appendix
To: LES
CC: JMC
I have finished with my ARPA Appendix; it is now ready for you to PUBify it.
The file is PRBARP.TXT[1,LGC]. Please do not delete my file, since I need to
keep it for future reference. I'll appreciate knowing what file the finished
version is on, so I can spool a copy for myself. Thanks. -- Lew
∂23-Jul-78 1938 TOB
John
The inquiry about machine usage brought the response from hand/eye
that things are much better. Suggested changes to service level or
data disc priority got no real response. Some fear expressed and
with reality that usage expands to overload available facilities.
Tom
∂23-Jul-78 2319 LES Appendix to ARPA proposal
To: LGC
CC: JMC
The source file is PRBARP.PUB[1,LGC] and the XGP file is PRBARP.XGP, also in
your area. The Pub run also left several garbage files in your area, apparently
because the default file protection prevented their deletion.
∂24-Jul-78 0000 JMC*
ticket
∂24-Jul-78 0310 LGC via AMES-TIP#52
I'm having trouble preparing a talk for a man I know too little about.
What is Carlson's main area of expertise, and what is his level of
training in it? How much does he know about AI? How about computer
science generally? How deeply is he likely to probe with questions,
and in what directions? I have no idea what kinds of things he will
need (or want) explained in detail, or what kinds of explanation
would bore him to death (or even insult him) with things he already
understands very well. Nor do I have any basis for guessing
what aspects of the problem solver project would be most likely
to interest and/or impress him. Can you help me with any of this?
Don't orient it too closely to any hypothesis about Carlson personally.
The programs he manages are all in computer science, so he surely
knows something about it. My gues is that his training is in some
other area of science; meteorology comes to mind.
He has not specialized in AI. The previous AI specialist went back
to flying airplanes, and his intended successor decided not to play
ARPA after all.
∂24-Jul-78 1000 JMC*
stuff for pripstein
∂24-Jul-78 1854 Creary at SUMEX-AIM Phone Numbers
Date: 24 Jul 1978 1855-PDT
From: Creary at SUMEX-AIM
Subject: Phone Numbers
To: JMC at SU-AI
If you should get definitive information concerning ARPA funding of my salary
next year, I would very much appreciate a short letter or phone call on the
subject. I can be reached at P.O. Box 593,
Lewiston, Michigan 49756
or at 517-786-4352 after 8/1 (Lewiston),
or at 313-521-1117 after 7/28 but before 8/1 (Detroit). Thanks very much.
-------
∂25-Jul-78 1008 PEG Rides to/from SAIL
To: S1.DIS[S1D,PMF]:
The so-called S1 Projectmobile is now in operation. To see when it
will be journeying between campus and SAIL, look at RIDES[1,PEG]. Those
interested in riding please contact me.
Phil Gerring
∂25-Jul-78 1305 DCL via SU-TIP#10 Delete list
To: LES
CC: JMC
Les, again our resources are being used at peak times by Guests and non
project related users:
1300 today: NO PHONE LINES
Among the five phoneline users at this time were:
SL (a guest) SCD(a guest) THL (Ph D student not on any project).
One phoneline must have been left hung up after logout.
Can we implement your proposed restrictions NOW?
∂25-Jul-78 1532 ME COPY/DIRED within E
∂25-Jul-78 1253 JMC COPY within E
It would be very convenient to be able to do all COPY funciions within E
including DELETE and DIRECTORY. Would this be difficult?
Preferably, the command notation mapping would be systematic.
ME - The only way to permit doing all COPY functions in E would be to
actually have E run COPY through a PTY, but I don't think that that is a
very good idea. There is also the question of where such COPY output
would appear -- in the 3 lines at the bottom of the screen? -- That's a
pretty small area for, say, a DIRECTORY listing. E already has the
capability of copying E-format files (e.g., ET NEW←OLD<cr>). It wouldn't
be too difficult to implement DELETE, perhaps only to work on the file you
are editing. The general solution would be to implement DIRED in E; I've
sort of wanted to do that for some time, but it is certainly a non-trivial
task, one that I wouldn't want to take on right now.
∂26-Jul-78 0646 FB via SU-TIP#1 big micros
There is a short blurb on the Motorola M68000 microporcessor in the
latest issue of Electronic Design. There will be a longer article on
the same subject in the same magazine in a few weeks. Besides the 16
megabyte address space, the most prominent feature of the M68000 is
16 32-bit registers.
∂26-Jul-78 1035 CLT
individual chapters are now pubbable. see lspcmt/4p
∂26-Jul-78 1410 DAN
The spacewar boxes are in bad shape. I would like to fix them. To do it
I will need to buy some new switches. It seems reasonable for the lab to
pay for anything that I need to buy. Does the lab have an account at EE stores?
And If so, could I use it to buy the stuff?
(I would have asked LES, but he is out of town.)
If TED agrees, I will go along, assuming it isn't much.
∂26-Jul-78 1754 MRC autologout
To: TVR, DON
CC: RWG, JMC, JBR
Welcome to the club. As for the autodetach business, EN LIV does work
and I do use it. In fact, it is a function of my logging in. I have
been saved many times by this when a user has called me away briefly to
ask me for assistance.
I had changed the SETPRV UUO so that any user could turn on LIVPRV, but
BH removed my code, on the grounds that it was somehow "immoral" for a
user to be able to turn on this half-useless privilege by default
(ignoring of course that LOGRUN can do it just as well, although more
sloppily).
I fully intend to bring this up at the system meeting. As LIV cannot
be used in an antisocial manner any more, I have no qualms about having
it turned on by default and will continue to do so in spite of what BH
says. My work comes above BH's esthetic considerations.
∂26-Jul-78 2231 MRC file transfers
To: Dialnet, JBR
For the first time today, large (>16 disk tracks) files have been
transferred using Dialnet from SAIL to SAIL, utilizing the CRUFT
protocol (CRUFT = CRetinous User File Transfer). Small files (<2
disk tracks) had previously been transmitted with no problem, but
a system bug caused a deadlock when one attempted to transmit a
file larger than that.
I have not tried it with the official Dialnet file transfer program
as its author is currently out of the country, but there is every
reason to assume it would work just as well.
∂26-Jul-78 2358 MRC Is LOTS now urgent
Yes. I believe now that the code here has been debugged to the point
where it is safe to proceed to LOTS.
∂27-Jul-78 0843 REG via AMES-TIP#54 If you see this in time, XS
GORIN3.XGP[LOT,JMC] and bring it with you to Lieberman
∂27-Jul-78 1013 REG CANCEL MY PREVIOUS
I'M ALL SET WITH MEMO FOR THE MEETING
∂27-Jul-78 2324 Davis at SUMEX-AIM car
Date: 27 Jul 1978 1658-PDT
From: Davis at SUMEX-AIM
Subject: car
To: siglunch:, STANFORD USERS DISTRIBUTION:
[As in automobile.]
FOR SALE:
71 Ford Pinto, 2-door, auto transmission, only 69,000 miles, orignal owner.
Excellent condition, good mph, fair acceleration [after all, it's only a
subcompact].
For more info, call Randy at 7-6970, or 329-0389.
-------
∂28-Jul-78 0040 PEG Dull lunches at SAIL
To: S1.DIS[S1D,PMF]:
Wouldn't you rather go back to the relaxing atmosphere of the
beautiful S.U. campus to eat lunch, instead of spending your lunch hour
hunched over a terminal while your program runs (maybe), contending with
high load averages while eating that egg-salad sandwich that the Pony
overcharged you for, and spilling your coke (that the machine charged you
sixty cents for) all over the terminal while waiting for the system to
continue or be reloaded (with 5 seconds warning to save your edits...)???
RIDERS WANTED! The Gerringmobile will take you to campus and bring you
back. For details, see RIDES[1,PEG].
∂28-Jul-78 1311 DCL via SU-TIP#10 Phone lines and delete list
To: LES
CC: JMC
Again, all lines are busy. The following "no project" students are
using lines at this time: CVW, BIA.
∂28-Jul-78 1946 Wiederhold at SUMEX-AIM datamedias
To: S1
Date: 28 Jul 1978 1947-PDT
From: Wiederhold at SUMEX-AIM
Subject: datamedias
To: s1 at SAIL
People report datamedia problems from the Sera trailer to Sail. Use
of 9------- lines rateher than 7---- centrex lines works better, but
is more expensive. Suggest try centrex first, if trouble go outside - directly
or via one of the TIPs. Pleas report behaviour to me if problematic. Gio
-------
∂28-Jul-78 2324 PMF FSIM
To: S1.DIS[S1D,PMF]:
I am not responsible for fsim. JBR and MLB are. Please do not send
complaints to me.
∂29-Jul-78 0043 RWW
∂28-Jul-78 1735 JMC Woody Bledsoe wants to see you.
He will be at 714 440-5898 San Diego this weekend thru Monday.
He would like to come to AI Lab at 10:30, and we can all go tt
lunch at noon.
He wants lunch on what day.
Sorry. He will come Tuesday AM, and lunch is proposed for Tuesday.
∂29-Jul-78 0052 RWW
That's fine with me I can set aside the entire day, do you have
any idea why he wants to see me? Also should I call him at the
number you gave me?
Yes, call. No real idea.
∂29-Jul-78 0203 MRC Dialnet modem at LOTS
To: LES, JMC
We have reason to suspect that the dialer at LOTS is losing. We are
investigating this further, but if our fears are proven correct it will
be quite a serious problem!!!
∂29-Jul-78 1407 JMC NSF proposal for program-prover cum proof-checker for LISP course
CC: JMC, RWW, CLT, boyer at SRI-KL, JRA
NSF awards grants of up to $25K for "local course improvement" in
science. PROOF.PRO[E78,JMC] is a start on an application for a grant
to make an interactive program-prover-proof-checker to be integrated wiih
MACLISP and used in CS206 - which will be renamed CS126 next year.
If you are interested in participating in some form or if you
have suggestions, please communicate.
∂30-Jul-78 1441 ARK S1 Meeting
To: S1
CC: Pickering at SUMEX-AIM
The next S1 CS Group meeting will be on Thursday, August 3 in the small
conference room at SAIL at 1330.
Arthur
∂30-Jul-78 1504 MRC dialer at LOTS
To: LES, JMC, TED
I'm now convinced that it is biting the bag. Can we arrange for somebody
competant with these modems to go and look at the thing and find out what
is wrong?
∂30-Jul-78 1511 MRC modem
I'd rather have somebody (hardwarily) competant do that sort of experimentation.
Most of what I was doing here was fooling around with the option switches
seeing that they were reasonable. Can we get Ted or Ted's replacement (if
such a creature gets generated) to do this?
∂30-Jul-78 1625 MRC LOSING VADIC
To: JMC, LES, TED
The losing Vadic is now the one at SAIL. I'm taking the winner down to
LOTS.
-- m
∂30-Jul-78 1643 DCL
To: GROUP.DIS[VCG,DCL]:
***********************************************************************
JOINT SEMINAR
CONCURRENT SYSTEMS
and
VERIFICATION GROUP
TUESDAY 1st AUGUST
PLACE: ERL 237
TIME: 2:30 p.m.
TITLE: Verifying Programs which Manipulate Pointers.
SPEAKER: Greg Nelson
ABSTRACT
The current methods for reducing the correctness of pointer
manipulation programs to the validity of first-order logical formulas
produce monstrous expressions which are almost impossible to relate to the
original program. As an alternative to these approachs, I will describe a
method for reducing the correctness of such programs to the validity of
formulas in a very powerful logical system--one in which all general
recursive functions are expressible. The added expressiveness of the
logical language makes it possible to mechanically reduce the correctness
of pointer programs to logical assertions which are short, succinct, and
readable. They are more difficult to prove than first-order formulas,
however. I will describe a method for proving them, and an experimental
theorem prover which uses the method.
∂30-Jul-78 2214 Boyer at SRI-KL (Bob Boyer) Proposal for Course Improvement
Date: 30 Jul 1978 2219-PDT
From: Boyer at SRI-KL (Bob Boyer)
Subject: Proposal for Course Improvement
To: JMC at SU-AI
We are definitely interested in your plan to propose the development
of a system for use in a LISP course to prove programs correct.
You are welcome to use the system we have been working on for several
years either piecemeal or in toto as part of the system you propose.
At the least, our participation in such a project could include
delivering the system we have to the LOTS computer and describing
its use and structure. In fact, it is currently well documented in
the manuscript we sent.
Incidentally, at SU-AI you have two students, Greg Nelson and Richard
Pattis, who are well versed in the structure of our system. They might
be interested in your plan.
Besides our delivering or describing our system what other participation
by us would interest you?
Bob & J
-------
Bob and J:
I am glad to hear you are interested. I will be abroad until
early September, but after that, I think all interested should get
together.
John
∂31-Jul-78 1018 Engelmore at SUMEX-AIM Siglunch for Thursday, August 3
Date: 31 Jul 1978 1019-PDT
From: Engelmore at SUMEX-AIM
Subject: Siglunch for Thursday, August 3
To: siglunch:
This week Jan Aikins will the work she is doing for her dissertation.
Prototype-Directed Problem Solving
A system has been developed which uses a set of
domain-specific prototypes of typical situations to direct a
consultation and guide the invocation of production rules. These
prototypes serve as hypotheses about how to classify the data in a
particular case. This system has been implemented in the domain of
pulmonary physiology in which the prototypes represent various
pulmonary diseases. This talk will focus on the capabilities of the
current system and those extensions which would allow the system to
perform multiple problem-solving tasks using the same knowledge base.
-------
∂31-Jul-78 1024 RWW woody
He will be here tuesday at 10:30. How about lunch at 12-12:15 as you
suggested. He needs to be at SRI by 1:30.
rww
∂31-Jul-78 1035 Tajnai at SUMEX-AIM cs206
Date: 31 Jul 1978 1036-PDT
From: Tajnai at SUMEX-AIM
Subject: cs206
To: jmc at SAIL
cc: pat at SAIL
CS206 aut TTH 11-12:15 ESMB 113
Please check suitability of classroom in advance.
Thanks, Carolyn
-------
∂31-Jul-78 1045 Boyer at SRI-KL (Bob Boyer)
Date: 31 Jul 1978 1047-PDT
From: Boyer at SRI-KL (Bob Boyer)
To: JMC at SU-AI
Since J will be gone all August, the beginning of September
is the earliest we could both meet with you anyway. See
you then.
-------
∂31-Jul-78 1153 CLT
To: JMC, CLT
I am interested in participating in the "program-prover-proof-checker" project.
I have a comment and a question regarding PROOF.PRO. The comment is that I
think the statement
"FOL has many grammatical crochets, so that mistakes in using it are frequent."
is unfair. I'm not sure what you mean by "grammatical crochets", but I find
the FOL syntax and command language very easy to use and the parser is
one of the most robust and helpful I know of.
In a proposal of this sort would it be useful to have examples of the
kinds of statments we hope to have students proving, and perhaps a
sample proof, or is that overkill?
There are really a lot of features of FOL that can only be described as
crochets. For example axioms must end in double semi-colon, a function
is declared by writing f(INTEGER) = INTEGER while an integer variable
is declared by n ε INTEGER. 3:#2#3 could be improved also. One can
never guess FOL syntax, but it has to be looked up in the manual.
I believe enormous improvements are possible. Incidentally, Boyer and
Moore have expressed interest, and I think we should all meet in September.
We can include some of the work we have already done, e.g. the theory
chapter of our book and a FOL proof, as appendices. That should suffice,
since I think what we propose to do will be more innovative than almost
everything they get.
∂31-Jul-78 1246 Rsmith at SUMEX-AIM RUTGERS
Date: 31 Jul 1978 1247-PDT
From: Rsmith at SUMEX-AIM
Subject: RUTGERS
To: JMC at SAIL
We are installing our 2050T system. The
installation is proceeding on schedule. We hope to have
it back up by next week or so for users; the system
has actually been running for two weeks but DEC is testing
it out. I will send you a message when it is up.
I think the machine will be called RUTGERS instead
of RUTGERS-10 since, strictly speaking, it is no longer
a 10!
I have been thinking about the frame problem and trying to
apply modal logic to it recently. Do you have any
thoughts/suggestions/papers about this?
Regards,
Bob Smith
-------
My present idea is to do the frame problem as an application
of circumscription. See EXAMPL[W78,JMC].
∂31-Jul-78 1327 MRC losing Dialnet modem
To: JMC
CC: LES
As I told JMC yesterday, I have verified that one of the dialers is losing.
The modem with the losing dialer is now at SAIL, while the winner is at
LOTS. The loser can receive calls alright, but can't initiate. I have
verified that the software both at SAIL and LOTS is winning, as LOTS worked
after the modems were swapped, and previously SAIL had worked and now is
getting the same problems as were at LOTS.
JMC told me to call Vadic, but I have no idea whatsoever who or where to
call! I spoke to Hersche, and the best she could do was tell me to wait
for Ted to come in tommorrow. I'll try to be in early tommorrow, but I have
a dentist appointment in the afternoon, so anything that will be lengthy
will have to be deferred until later; also I am going to have to take my
car in for a checkup sometime soon (although it can be put off a day or two).
Anyway, that's what's up (or down) right now, and I can't do anything about
it until I get more information.
∂01-Aug-78 1148 TED VADIC MODEMS
To: MRC
CC: LES, JMC
What is the true story on the VADIC modems? What makes you think there is
something wrong with the modem and not somewhere else. If we can pin it
down to the modem, please tell me the exact problem and I will communicate
with VADIC and try to get is straightened out. TED
∂01-Aug-78 1253 LCW PASCAL Header Page
To: S1
A header page for PASCAL programs now exists as HEADER.PAS[S1D,PMF]. It
contains a description of options for the PDP-10, SLAC, and S1. It will
be updated as necessary. People responsible for PCPASC, SOPA, S1FORT, and
all other S1 PASCAL programs, please prefix your code with HEADER.PAS in
order to provide in-code options documentation.
Curt
∂01-Aug-78 1300 MRC modems
To: TED
CC: LES, JMC
I took the loser in to VADIC, who agreed that there was a problem. The
symptom of the lossage was that it responded "D" to virtually everything.
The VADIC guys think it is that funny protocol adapter frob but are
looking at it.
∂01-Aug-78 1905 REM at MIT-AI (Robert Elton Maas) Interval-Refinement Stochastic-Mapping
Date: 1 AUG 1978 2205-EDT
From: REM at MIT-AI (Robert Elton Maas)
Subject: Interval-Refinement Stochastic-Mapping
To: RWG at SU-AI, JMC at SU-AI, DEK at SU-AI, WILBER at SRI-KL
To: LAUREN at UCLA-SECURITY
CC: MRC at MIT-AI, RMS at MIT-AI, KLH at MIT-AI, REM at SU-AI
CC: FFM at MIT-MC, ELLEN at MIT-MC
One of my moments of inspiration.... I have just spent the past 8
hours writing an article suitable (??) for publication describing
nicely (??) the pre-intcom algorithm of mine for radix-conversion etc.
(Using EMACS through CRTSTY REMB here on my Beehive+6502 with
simulation of Datamedia 2500 random-cursor-positioning.)
I will deposit a copy of the file with you unless you tell me you
don't want to see it before I time out and send it. The file
is nearly 3000 PDP-10 words (15000 characters) long.
(CC: recipients, default is NOT to get the file, MRC may sigh relief)
∂01-Aug-78 2046 REM via AMES-TIP#66
To: RWG, JMC, DEK
See IRSM.WRU[1,REM]
∂01-Aug-78 2146 PMF PDP-10 pascal misfeature
To: S1
Files in pdp-10 pascal with extension TMP are put in tmpcor. This is usually
not what was intended and so should be avoided.
∂02-Aug-78 1611 YAW possible visit
John,
I talked tentatively to LES about the following and he
seemed to think it might be possible: Id like to be
in the area next Spring/Summer for up to 3 months
and would be grateful for a desk if you had one free.
Im reasonably confident that a NATO fellowship
would cover fares and cash, but they would need
a host this end as a base from which I could talk to
people here/SRI/XEROX-PARC etc. If you agreed
to let me mention you/the Lab. in this role
it would cost you not a penny. Do you feel positive
about the idea?
Yorick.
I think it's possible, and I am willing to bet that we will
be able still to find the desk space when we move to campus,
so you can mention us. The degree of sharing that might
be involved cannot now be guaranteed.
∂02-Aug-78 1854 DCL Meeting
To: JMC
CC: LES
John, Les and I would like to have a short meeting with you before you
leave on your trip. I guess it has to be Thursday (tomorrow).
This is to discuss two items of business:
1. My wish to move one of my NSF grants to DSL (the new one).
2. A priority scheme for login on Data media phones and data disc channels
during busy hours.
Please let us know what time.
-David
I'll be in all day, I think.
∂02-Aug-78 1859 DCL
Good, shall we say 2.00pm?
∂02-Aug-78 2305 CLT
Proposed (revised) organization of Book
The idea is to divide the book in two parts (conceptually if not actually).
Part 1 covers the basic topics which would be the core of the course material.
Part 2 is a collection of special topics, more advanced material, examples
of programs and additional theoretical techniques....
For text book purposes material from part 2 could be selected according
to taste and interest of teacher and students.
Part 1:
I Reading LISP programs (function definitions)
II Writing LISP programs
III Proving statements about LISP programs
IV Dynamic LISP --- prog, rplacs, etc.
V Compiling
VI LISP implementation (how it really works)
Remarks: Chapters I-V exist. II has been revised, but requires polish.
III needs revision (I have an outline). IV needs to be extended.
VI is yet to come.
Part 2:
SEARCH Searching programs
EXTEND Transforming LISP programs
ABSNTX Abstract Syntax
COMPUT Computatibility/Non-computability
Control Structures
Pattern matching, syntax directed computation, data driven computation
Dialects of LISP, systems written in LISP
Collection of substantial programming and/or proving projects
Remarks: Chapters with names exist is some form those with out names do not.
For more details on contents of chapters see LSPCMT[LSP,CLT]/11p and /12p.
A good idea, but compiling should be in part II, since knowing how to write
a compiler is really an optional topic.
∂02-Aug-78 2327 CLT
The reason I put compiling in part one was that I wanted to discuss the
interaction of compiled and interpreted code in LISP implementations.
It seems like useful information (also a nice feature) which is not
easily extracted from any manual I have read.
∂03-Aug-78 0435 REM at SU-AI at MIT-MC (Robert Elton Maas) More good data-compression with INTCOM/IRSM
Date: 3 AUG 1978 0736-EDT
From: REM at SU-AI at MIT-MC (Robert Elton Maas)
Sent-by: FFM at MIT-MC
Subject: More good data-compression with INTCOM/IRSM
To: JMC at SU-AI, RWG at SU-AI, WILBER at SRI-KL
To: LAUREN at UCLA-SECURITY
CC: WD at SU-AI, REM at SU-AI
While devising an improvement on SDLC et al synchronous data
communication protocols that used my INTCOM/IRSM, after getting
somewhat-improved average-case performance and MUCH-better worst-case
performance, I decided to return to the work that led to INTCOM
in the first place on Feb.12, namely slight improvements on Huffman
code that allow code-spaces of size 3/2↑n, 5/2↑n, etc. to be filled
as well as the usual 1/2↑n. The idea is to use my INTCOM methods
(which have lots of multiplies and divides) to construct a code,
but rather than have it adaptive (one of the options in INTCOM),
to decide at the start that it will be fixed and construct a state
table that can be held in a small memory to drive the decision tree
that operates the shift-registers and the I/O in the INTCOM algorithm.
Example. Instead of eating two out of every 10 bits for start/stop,
thus getting MAXIMUM thruput of .8 if all eight data bits are fully
entropic, and requiring clock tolerances of 5% to avoid a half-bit slop
at the stop bit, synchronous communication methods require that there
be at least one transition every 6 bits, thus allowing slightly looser
tolerances and not eating up 1/5 of the bits. But this eats slightly
into the code space, and the obvious (and proposed standard) mapping
between data and line-bits is sub-optimal (worst case is 5 bits of
data in each 6 bits on the line for thruput of .833333, only slightly
better than asynchronous, and average only .984127, whereas optimum
thruput is a uniform .988 appx.). INTCOM provides essentially optimum
thruput at the expense of some 16-bit multiplies and divides for each
bit processed (8-bit mul and div would still be near-optimum, but faster,
but not fast enough). I have manually computed what INTCOM would do
with 3 bits and with 4 bits:
number of bits: 3 4
worst-case thruput: .90000 .94100
average thruput not computed
number of states: 10 22
number of bits of PROM needed
to hold control table: 12*13=156 24*17=408
It looks like with a moderately small PROM we could drive the INTCOM
stuff so close to optimum as to be difficult to exceed by other means.
Note, the arbitrary buffering-delay which occurs when using the additive
shift-register (in the crunch pipeline) occurs when TRANSMITTING a packet,
whereas when RECEIVING a packet and translating back to pure-binary we
are using the subtractive shift-register (in the uncrunch pipeline)
which has guaranteed finite delay. Thus when receiving data we never
fall behind and lose a packet due to delay in the SDLC-->BINARY mapper,
and when transmitting we can probably compute the packet early or else
truncate the packet early if encoding falls too far behind. Thus I
think my algorithm and implementation is appropriate for improving the
state of the art in synchronous communication. (Details remain semi-secret
until I'm well along the way toward getting a patent, but thought you all
would want to hear the usefulness I found.)
I even have a way to make this work in my 6502 without requiring either
a permanently-fixed code or great compute power -- when remote computer
is sending data to my 6502, it uses the currently-established code,
otherwise it uses the idle time to transmit updates to the code table.
I may implement this in the not-too-distant future .... but I think I
have a solid case for claiming my patent is the reducing-to-practice
of an otherwise infeasible mathematical technique (circa 1972), in
a way that is not obvious from the mathematical algorithm itself,
and thus patentable. I may be writing up my arguments for patentability
soon, in preparation for contacting the expert at Stanford who helps
Stanford-affiliated scientists get their inventions patented (and more
important, marketed) in exchange for splitting the proceeds.....
∂03-Aug-78 1054 PAT call from Laske
John Laske called and left the following message:
"I do not think, following first reaction of Carolyn Talcott,
that the induction principle you proposed is seriously unsound.
Your (JMC's) counter-example is not a true one, but there is
some trouble."
If he does not talk to you before you leave, Audrey and John give you
their best.
The above message is to go with a photocopy of something of Carolyn's but
she has not come in yet so I don't have it.
∂03-Aug-78 1119 CLT
did you find your copy of the Boyer/Moore manuscript?
Not yet. I'll look at home.
∂03-Aug-78 1209 REM via AMES-TIP#20 First draft of intended paper (public, CACM or IEEEx etc.)
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
The file IRSM.WRU[1,REM] (also stored as MIT-MC:REM;ISRM >) contains
a 15k-character file which I hope to publish somewhere soon. It does
not reveal the method of reducing the algorithm to practice, which I
invented 1978.Feb.12 (with a few later enhancements) and which I intend
to patent, however since it does reveal my earlier invention of general
radix-conversion (i.e. stochastic-mapping) it should be considered as
a limited-publication item for now, i.e. don't make more copies than
the one you need for your own reading, and don't advertise it to others.
TITLE: Interval-Refinement Stochastic-Mapping
POSSIBLE CHANGE IN TITLE: Stochastic Mapping by Interval-Refinement
Abstract not yet written.
You should FTP in text (ascii = TYPE A, BYTE 8) mode if you are not
local to SU-AI or MIT-MC and you want to read it. If you have trouble
FTPing it, contact me and I'll net-MAIL it to you. (No password is
needed for FTP from here.)
∂03-Aug-78 1252 MRC DLNSER documentation
The DLNSER documentation now contains information for the benefit of
monitor programmers who want to convert the SAIL Dialnet routines to
run on a Tops-10 system. I'm not as familiar with the internals of
Bottoms-10 as I am with WAITS, ITS, or Tenex, but I tried to give a
reasonable amount of information. The code should be straightforward
enough in most places for a wizard to have a fairly simple conversion
job; just a lot of TECOing.
∂03-Aug-78 1624 TW via PARC-MAXC2
I would like to refer to your minimal entailment paper in
something I am writing. Is there a latest version, and what is
the appropriate citation?
--terry
Here is the citation. I suggest you cite the computer file
analogously to citing an internal report and also as to be published.
MINIMA[S77,JMC] 11-Jul-78 CIRCUMSCRIPTION INDUCTION
- A WAY OF JUMPING TO CONCLUSIONS
∂03-Aug-78 2151 MRC SPOOL
Thanks. I wish you had used gripe though. I don't know how much of
what you said to me went out over the comm line, but I find that sort
of thing really humiliating, especially when I had nothing to do with
the lossage. Others hearing such things can come to the conclusion
(reinforced by their prejudices about hackers in general) that my main
function in life is to make "changes" whose utility is debatable,
except to keep system software unreliable and confuse users. Such
attitudes on the part of the user community, especially among those
who should know better or are in one way or another influential, are
not compatable with keeping a healthy atmosphere of system software
development!
∂03-Aug-78 2309 MLB New FSIM
To: S1.DIS[S1D,PMF]:
New FSIM up on system. New eXtended commands MUTTER,SAVE,HELP, miscellany.
Reformatted FSIM.0[S1D,PMF] tells all. R FSIM gets what you asked for.
∂04-Aug-78 1037 Tajnai at SUMEX-AIM Syllabus for Comprehensive
Date: 4 Aug 1978 1038-PDT
From: Tajnai at SUMEX-AIM
Subject: Syllabus for Comprehensive
To: JMC at SAIL
cc: DPB at SAIL
John, at the 5/30 Faculty Meeting you commented that some of the
students didn't know what an invariant was and that reflected on
the curriculum and/or syllabus.
I have asked Frances Yao if there are changes to be made in the
syllabus. She has none.
Because of your comment at the Faculty Meeting, I thought you might
have an addition or two in mind for the syllabus.
Please advise.
Carolyn
-------
This is in response to your message of 4 August - the day I left.
The readings should include the chapter in Manna's book that cover
invariants if they aren't already on the syllabus, and it should
also include McCarthy and Talcott - LISP, Programming and proving.
The following sentence should be included: "The examination may
require proving statements about properties of programs by constructing
a suitable invariant for sequential programs (see Manna) or by
representing the program in first order logic for recursive
programs (see McCarthy and Talcott)." If the syllabus is already out,
this amendment should be included as a supplement unless Frances
objects.
∂04-Aug-78 1906 Wiederhold at SUMEX-AIM Minutes from three weeks ago.
To: S1
Date: 4 Aug 1978 1903-PDT
From: Wiederhold at SUMEX-AIM
Subject: Minutes from three weeks ago.
To: s1 at SAIL
Minutes S-1 meeting 20Jul78 Small Conference Room at AI-Lab
Present mlb, ark, hwc, hal, peg, fac, blh, alp, gls, pn, fc, pbk,ejg
ptz, lcw, gio, ko
absent bth
OPERATIONS (peg,ejg)
Backup of UDP will be done Fridays or Thursday evenings. When a new
disk drive comes to the AI-Lab, allocations can be increased.
There are problems in high segment sharing. Two job slots are needed
to run PDP-10 PASCAL jobs. A 9600 baud line between SAIL and LLL
should work soon (hwc) and can be used to transfer data. Program testing
will remain mainly at the AI-Lab using the simulator.
PASCAL (ko etal)
Standard PASCAL from San Diego seems to be established as a minimum
level.
ko will determine how MODEL gets to have OWN variables, and present
their solution for our consideration.
Expect soon (underscore) instead of (left arrow). New cross reference
does not handle #, else works. Runtime still needs a change. Option
U and bug has been fixed. EOF not recognized.
LINKER (ark)
Parts of code are beginning work. 30/36 bit address is open.
FORTRAN (pn etal)
Complex and conversions (pn), subroutine calls (fac), format conversions
(fc) are in progress. The question of base register allocation by
procedure versus global allocation was raised.
SOPA OPTIMIZATION (pbk)
Analysis of good tasks is in progress. Register allocation is
promising. There is quite a bit of dependency on Sites work on the
PCode-to PCode optimizer in San Diego.
HARDWARE (lcw,pmf)
The machine is doing better, but routine use of the Mark I from here
is quite aways off.
-------
∂04-Aug-78 2253 LLW Forest Baskett-SAIL Affiliation
To: JMC, LES
CC: LCW, LLW
Forest Baskett spoke with me today regarding the details of his return to
Stanford as they impacted his long-anticipated affiliation with the S-1
Project. We concluded that the best interim arrangement would involve his
joining the SAIL portion of the effort. This offers the advantages of
minimizing administrative overhead at my end of things, of placing him in
that portion of the Project where his primary interests lie (systems,
particularly OSs), and of immediate availability. I hope that such an
arrangement might be seen by you as strengthening your OS-related research
effort, while having essentially no drawbacks, and thus be quite
agreeable.
Assuming that this would be the case, I asked Forest to call you, Les, as
to the details of the arrangements. He would like to start on Sept. 11,
and any fraction of the time that he would care to work, both before and
during the academic year, is quite agreeable to me. I believe that you
are underrunning your contrαct sufficiently at present that his joining
your effort should pose no problem from the financial plan standpoint.
I hope that you are as pleased by the prospect of Forest's participation
as am I.
Lowell
I just returned, and this is in response to yours of 4 August about Baskett.
His being affiliated with the SAIL part seems fine to me, and now we can
discuss the next proposal.
∂06-Aug-78 2213 REM via SU-TIP#1 INTCOM WORKING WITH LEFT-CONTEXT ADAPTIVE ALGORITHM
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
I finally found time to work on my new data-compression algorithm,
adding left-context so it can really be tested. The first raw test
was crunching a file consisting entirely of "QU" over and over, whereas
without context it learned that each character had 50% chance thus needed
about 1 bit, thus 4k file crunched to 1/2 k, with context it crunched
the entire file down to about 12 words!! The first real test was on
its own source which was 8230 words. It crunched it down to 3824 words.
Thus it does slightly better than previous program using Huffman code
with context (non-adaptive).
With NFORK (number of instances of a given context before that context
gets its own code) = 5, the program crunched the test file ok, but used
up too much memory (510 pages = 255 k) and crashed when it couldn't get
any more core. With NFORK = 10 it did the 8230-->3824 reported above.
Now to play around with parameters and other files....
ANYBODY WANT TO PLAY WITH IT? (SU-AI USERS ONLY)
∂06-Aug-78 2307 REM via SU-TIP#1 More on working version of INTCOM/CONTEXT
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
My nutrition database is 21456 words. By old program it crunched to
6489 words. With INTCOM/CONTEXT it crunches to 4644 words.
My nutrition database is a tabular file using spaces and tabs between
columns. INTCOM probably reduces a character to less than one bit in
many cases, thus saving over best Huffman or other discrete-bit code.
The deeper context, 5 instead of 3, also probably helps, allowing full
word-recognition on many of the vitamin/mineral names and many of the
common numeric values like 0.0 and x.33333 etc.
∂07-Aug-78 0752 MARG at PARC-MAXC hi
Date: 7 AUG 1978 0755-PDT
From: MARG at PARC-MAXC
Subject: hi
To: jmc at SAIL
Hi. I heard that you're involved in L5. I am looking for the whereabouts (mailing
address) of one of them named Eric Drexler. Maybe I can find some contacyts
from you to locate him. E.g. Carolyn Hensen might know where he is.
Otherwise, just hi since I haven't seen you yet this summer.
-------
I just got back from my round the world trip and saw various messages from
you. If there is still time before you leave, perhaps you would like to
eat at Louie's with Sarah and me.
∂07-Aug-78 0937 Nilsson at SRI-KL (Nils Nilsson) A Word About the Future
Date: 7 Aug 1978 0938-PDT
From: Nilsson at SRI-KL (Nils Nilsson)
Subject: A Word About the Future
To: Feigenbaum at SUMEX-AIM, JMC at SU-AI, TW at SU-AI
cc: Nilsson
We were sorry to hear that you could not attend the workshop at SRI
being organized at the suggestion of Marvin Denicoff to talk about
future directions for AI technology and applications. If your plans
change so that you will be able to come, please do! Otherwise, if you
have ideas that you would like to put down on a couple of pages, I'll
read them at the workshop.
-Nils
-------
∂07-Aug-78 1028 REM via AMES-TIP#20 More on new version of INTCOM
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
I collected some transcript of actual online tty output, into a file.
I used the new program, and got a c.r. of about 55%, thus not as good
as normal text files (more entropic than normal homogeneous files).
Then I tried the old program and it wouldn't work at all because of
strange sequences of characters like <lf> not preceded by <cr>.
Note, yesterday I tried the new program on a core-image (not text) file
and it worked ok, although due to mismatched byte size and frequent
line-number bits on the compression wasn't good at all, but the old program
would have crashed totally! Thus, except for compute time, the new program
is as good as or better than the old program, as expected, except I
had hoped it would be significantly better in more cases, will investigate
after I finish some changes....
∂07-Aug-78 1101 REM via AMES-TIP#20 More good results on INTCOM/CONTEXT
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
The private version of the PCNet mailing list, original 9193 words,
crunched by old CRU2 program 5551 words, now by INTCOM it becomes 4891 words.
∂07-Aug-78 1557 DCL
To: GROUP.DIS[VCG,DCL]:
***********************************************************************
JOINT SEMINAR
CONCURRENT SYSTEMS
and
VERIFICATION GROUP
TUESDAY 15th AUGUST
PLACE: ERL 237
TIME: 2:30 p.m.
TITLE: Formal Derivation of Compilation Algorithms.
SPEAKER: Prof. Dienes Bjorner
*****************************************************************
NOTE: THERE IS NO SEMINAR TUESDAY 8TH. AUGUST
*****************************************************************
∂09-Aug-78 1040 Robinson at SRI-KL (Ann Robinson) Your trip to Washington for Scientists for Shcharansky
Date: 9 Aug 1978 1041-PDT
From: Robinson at SRI-KL (Ann Robinson)
Subject: Your trip to Washington for Scientists for Shcharansky
To: jmc at SU-AI
'Scientists for Shcharansky' is ready, able and willing to
pay the expenses for the trip you made to Washington. However,
they need a bill for the expenses. You can either SNDMSG me the
bill or mail it to
Scientists for Shcharansky
Public Information Department
Lawrence Berkeley Laboratory
Berkeley, California 94720
In addition to a total, we would like an itemization of the mmajor items,
especially plane fare and lodging.
Ann Robinson
-------
Just returned from my trip and received your message about expenses of
Washington trip. I thought I had sent them to Morris, but it turns out
I hadn't and will do so today.
∂11-Aug-78 2302 RWG re dcl:
To: LES, RWG, JMC
"In the longer run, we hope to augment the terminal capacity of the SAIL system so
as to circumvent this problem. In the meantime, we regret the inconvenience to
some users."
When the terminal capacity of SAIL is augmented, the "circumvention" will be to
convert the (system nigger) "problem" into a higher load average. Our friend will
then insist that you adapt the "solution" to read, eg, niggers will relinquish
their jobs as quickly as possible whenever the load average exceeds, say, 10.0. If
you do, it will soon be 5.0. For a while. And someday, perhaps, he will covet my
office for some reason... .
So my question is: just where do you plan to make your stand? Or will you just
become sorrier and sorrier for the inconvenience?
∂14-Aug-78 1835 Feigenbaum at SUMEX-AIM Re: AIL Usage
Date: 14 Aug 1978 1830-PDT
From: Feigenbaum at SUMEX-AIM
Subject: Re: AIL Usage
To: Levinthal
cc: tob at SU-AI, les at SU-AI, liebes, jmc at SAIL
In response to your message sent 8 Aug 1978 1419-PDT
I heartily endorse the concept of the collaboration described in your memo.
As department chairman, I conrol only one of the resources discussed in
your memo, namely the working space needed. Though space is tight in the
new building, we'll work you in as requested...somehow. However, for the
time being, hold on to your office space in SUMC.
I look forwaard to an affirmative response from John and Les, cementing the
whole deal. I sounds like an important addition to the department's
research profile and resources.
Best of luck on the new endeavor,
Ed
-------
∂15-Aug-78 0819 MLB
To: S1.DIS[S1D,PMF]:
The state of the simulator after using the new extended command SAVE is currently
slightly screwy. The saved dump file should run OK, though. Hope to have it fixed
in the near future. - Marc
∂15-Aug-78 1120 Levinthal at SUMEX-AIM 8/8/1978 Memo
Date: 15 Aug 1978 1103-PDT
From: Levinthal at SUMEX-AIM
Subject: 8/8/1978 Memo
To: jmc at SAIL
cc: FEIGENBAUM, tob at SU-AI, les at SU-AI, liebes
Dear John,
I would appreciate receiving your concurrence with the arrangements
outlined in the above memo. Thanks.
Elliott
-------
∂16-Aug-78 1351 JED AI Qual Reading List
To: SJW, ROD, JMC, buchanan at SUMEX-AIM
To: feigenbaum at SUMEX-AIM, bennett at SUMEX-AIM
To: winograd at PARC-MAXC
One suggestion which came out of this spring's qual was that an update be
created for the syllabus prepared previously by Winograd and Lenat.
A number of people indicated that this could be done, by
by either the students or faculty, to provide a basis for the next
oral exam.
I have generated a list of readings which I feel could be part of such an update;
it resides in UPDATE[QAL,JED] at SAIL. The format is the same as the original,
with the same abbreviations and symbols.
This is merely a list of suggestions, and not complete; the appropriate
faculty members can use this, to whatever degree desired, in preparing
the actual syllabus.
Jim
∂16-Aug-78 1715 Wiederhold at SUMEX-AIM S1 meeting
To: S1
Date: 16 Aug 1978 1706-PDT
From: Wiederhold at SUMEX-AIM
Subject: S1 meeting
To: s1 at SAIL
cc: pickering
Reminder : The bi-weekly meeting time is this thursday 8AUG78 1:30
in the AI=LAB rear conference room. Gio
-------
∂16-Aug-78 1738 WINOGRAD at PARC-MAXC2 Re: AI Qual Reading List
Date: 16 AUG 1978 1735-PDT
From: WINOGRAD at PARC-MAXC2
Subject: Re: AI Qual Reading List
To: JED at SU-AI, SJW at SU-AI, ROD at SU-AI, JMC at SU-AI,
To: buchanan at SUMEX-AIM, feigenbaum at SUMEX-AIM,
To: bennett at SUMEX-AIM, winograd at PARC-MAXC
In response to the message sent 16 Aug 1978 1351-PDT from JED@SU-AI
As you may remember, I volunteered last spring to work on the
syllabus and arrangements for next year's AI qual. Since
Jim's note refers to the "appropriate faculty members" I am somewhat
relevant. However, I am going to be away for the rest of the summer
and won't be able to do anything on it until fall quarter. I will be glad
to work on it then, but if people need a list sooner, some other
faculty member will have to take charge for now.
I think Jim should be commended for taking things in hand
and getting things going.
thanks --terry
-------
∂16-Aug-78 1734 LLW Call for Proposals
To: JMC, LES, wiederhold at SUMEX-AIM
CC: LLW, LCW, BS
This is a reminder that proposals for S-1-related software R&D for the
post-October 1, l978 period need to appear at LLL very shortly, if they
are to credibly propose work for the period commencing October 1. While I
presently expect to be able to extend your present contracts on a
no-added-cost basis well into the Fall (since both contracts have
substantial uxexpended funds, which could probably be stretched to cover
costs until Christmas), it would be reckless to throw away the 'pad' in
time which has been developed through these underruns, and go back to the
'Perils of Pauline' mode of meeting payrolls which Gio and I were in a
year ago. I therefore hope to get your proposals into LLL Procurement, in
final form, by the end of this month.
As before, I'll appreciate it if first Curt, and then I, can have an
opportunity for comment on them before they're formally submitted.
Thanks,
Lowell
∂17-Aug-78 0737 RSMITH at RUTGERS RUTGERS MACHINE
Date: 17 Aug 1978 1039-EDT
From: RSMITH at RUTGERS
Subject: RUTGERS MACHINE
To: JMC at SAIL
OUR TOPS20 SYSTEM IS NOW UP ON THE NET. IT IS STILL
KNOWN AS RUTGERS-10.
IF YOU WOULD LIKE AN ACCOUNT, PLEASE SNDMSG TO HEDRICK AND
GIVE HIM YOUR PREFERRED PASSWORD ETC.
REGARDS,
BOB SMITH
-------
Thanks for RSMITH message about possibility of account on RUTGERS-10.
I would like one if offer still open. The name should be MCCARTHY or
whatever is close to it, and the password should be 321-7580.
∂17-Aug-78 0851 Wiederhold at SUMEX-AIM Re: Call for Proposals
Date: 17 Aug 1978 0849-PDT
From: Wiederhold at SUMEX-AIM
Subject: Re: Call for Proposals
To: LLW at SU-AI, JMC at SU-AI, LES at SU-AI
cc: LCW at SU-AI, BS at SU-AI
In response to the message sent 16 Aug 1978 1737-PDT from LLW at SU-AI (Lowell Wood)
It was my concern for the proposals that prompted me to send earlier
messages via MAIL, Curt, and Tom McWilliAms to arrange a meeting .
A reasonable definition of tasks and goals is important to me before
I can generate proposal text. When? Where? Gio
-------
∂17-Aug-78 0902 Wiederhold at SUMEX-AIM (Response to message)
To: S1
Date: 17 Aug 1978 0854-PDT
From: Wiederhold at SUMEX-AIM
Subject: (Response to message)
To: BTH at SU-AI
cc: s1 at SAIL
In response to your message sent 16 Aug 1978 1949-PDT
oops , looked st the wrong digit on my watch, the data is of course
17AUG78
for the S1 mmeting. Gio
-------
∂17-Aug-78 0942 Wiederhold at SUMEX-AIM (Response to message)
To: S1
Date: 17 Aug 1978 0854-PDT
From: Wiederhold at SUMEX-AIM
Subject: (Response to message)
To: BTH at SU-AI
cc: s1 at SAIL
In response to your message sent 16 Aug 1978 1949-PDT
oops , looked st the wrong digit on my watch, the data is of course
17AUG78
for the S1 mmeting. Gio
-------
∂17-Aug-78 1116 ARK Reminder: S1 meeting
To: S1
The bi-weekly meeting time is this Thursday 17AUG78 1:30 pm in the AI lab
small conference room.
Arthur
∂17-Aug-78 2246 MRC Dialnet
To: Maxion at SRI-KL
CC: LES, JMC
Roy,
Well, you've come to the right person. Dialnet is a set of communications
protocols between computers using ordinary TelCo phone lines. We use
VADIC VA3400 1200/1200 full duplex modems with VA801 pulse dialer. In its
full form, Dialnet will have file transfer (in all its hairy forms), mail,
linking, telnetting, and any number of private protocols limited only by
the imagination of the participants. Essentially, it is a poor man's
ARPAnet; however, while it is expected that it will have all the USER
capabilities of the ARPAnet, it is in design infinitely simpler.
Right now, the line transmission protocol (equivalent to the combination
of ARPAnet IMP-IMP, IMP-Host, and Host-Host protocols) has been specified
and committed to stone tablets. An implementation of this protocol has
been written and tested for Stanford's WAITS operating system and has been
up for over a month. An implementation has also been written for DEC's
Tops-10 operating system, but at the current writing it hasn't been
tested. Both the Stanford and the Tops-10 implementations are implemented
as monitor routines. I am presently working on a user-mode implementation
for the Tops-20 monitor, which is being developed and tested at LOTS.
The file transfer protocol has been specified, but has not yet been
committed to stone; it is likely however that the final version will be
essentially unchanged from the current version. An implementation of the
file transfer protocol has been written for WAITS.
The other protocols are in one form or another in the process of being
designed.
It would be much better if we sent you hardcopy documentation instead of
online files, as our online files are PUB-language documents for output
on Stanford's XGP. If you send me a mailing address, I will have an
abstract of our original NSF proposal and the current line transmission
and file transfer protocol document sent to you.
I am quite happy to hear of interest within SRI for Dialnet. SRI does
have compatible modems with ours, although I don't know if your modems
have dialers. You should be able to run the LOTS implementation of
Dialnet on SRI-KL with little or no modification.
Cheers,
-- mark
∂18-Aug-78 0914 Nilsson at SRI-KL (Nils Nilsson) AI Workshop
Date: 18 Aug 1978 0917-PDT
From: Nilsson at SRI-KL (Nils Nilsson)
Subject: AI Workshop
To: Hart, Duda, Hendrix, Moore, Sacerdoti, Nilsson, Tenenbaum,
To: Taylor at XEROX-PARC, Bobrow at XEROX-PARC,
To: Fikes at XEROX-PARC, Goldstein at XEROX-PARC,
To: Kay at XEROX-PARC, Lenat at SUMEX-AIM, Buchanan at SUMEX-AIM,
To: CCG at SU-AI, Balzer at USC-ISI, Denicoff at USC-ISI,
To: Feigenbaum at SUMEX-AIM, JMC at SU-AI, TW at SU-AI
Here is the latest news on the brainstorming workshop on future directions
in AI. I'm sending an agenda along with this message. The
workshop will be held in the Keidanren Room (S109) of the SRI
"International Building" (the modern white building adjacent to the main
SRI administration building on Ravenswood Ave.) We'll start promptly
at 9:00 a.m., but plan to come a bit early for coffee and donuts.
Please note the opportunity to make a short personal statement early in
the meeting. (To those who said they could not attend this workshop:
If you would like me to read a statement expressing your views about
future AI directions, please send me it.)
Also note that we are planning to get together for a no-host dinner
on Monday evening. (If any of you have any suggestions for a nice,
quiet spot for good food, please let me have them. Can we do something
different than a Chinese restaurant this time?)
WORKSHOP ON ARTIFICIAL INTELLIGENCE IN THE EIGHTIES
Technology, Applications and Strategy
SRI International
August 28 and 29, 1978
Keidanran Room (S109) International Building at SRI
Ravenswood Avenue, Menlo Park
AGENDA
Monday, Aug 28
8:30-9:00 Coffee and Donuts
9:00-9:45 Introductory Remarks: Marvin Denicoff, ONR
9:45-10:00 Coffee
l0:00-12:00 Short (Seven Minute) Statements from Attendees
12:00-1:30 Lunch
1:30-5:00 Brainstorming Session on AI Technology
Present Status, New Directions, Important Problems
Moderator: Nils Nilsson
6:00 No-host dinner (place to be announced)
Tuesday, Aug 29
8:30-9:00 Coffee and Donuts
9:00-12:00 Brainstorming Session on New Application Opportunities
Moderator: P. Hart
12:00-1:30 Lunch
1:30:4:30 Brainstorming Session on Pragmatic, Organizational,
and Financing Strategies
4:30 Adjourn
--Nils Nilsson (NILSSON@SRI-KL)
-------
∂18-Aug-78 1328 BTH NEW ARCHITECTURE MANUAL
To: S1
As sections of SMA3 are completed, they will be placed in SMA3.XGP[s1d,pmf].
This is formatted for the xgp and can be listed by saying:
XS SMA3.XGP[S1D,PMF]
This manual is still in development.
We would appreciate being notified of any inaccuracies found.
BUGS→BTH.
∂21-Aug-78 1257 David Levy at MIT-AI (David A. Steele)
Date: 21 AUG 1978 1528-EDT
From: David Levy at MIT-AI (David A. Steele)
Sent-by: RG at MIT-AI
To: JMC at SU-AI
Dear John,
My present address is
104 Hamilton Terrace
London NW8 9UP
England
My final match will be played in Toronto beginning August 26 and ending
around Sept 2nd or 3rd. If I lose, I shall mail a cheque to you at your
Department address; otherwise I look forward to hearing from you on your
return.
Best regards,
David
∂22-Aug-78 1352 Tajnai at SUMEX-AIM New PHD Students - support
Date: 22 Aug 1978 1353-PDT
From: Tajnai at SUMEX-AIM
Subject: New PHD Students - support
To: faculty.list:
cc: Hart at SRI, Engelmore
The file NEWPHD[PUB,DPB] at SAIL contains a memo from me concerning
support for new students. Hard copies will go out in ID mail tomorrow
morning.
Denny Brown
-------
∂23-Aug-78 1500 LCW Sites' FY79 Proposal
To: S1
The technical portion of Sites' FY79 optimizer proposal is available in
the SAIL S1 Library.
Curt
∂23-Aug-78 2139 LCW via SU-TIP#11 Disk Areas
To: S1
The disk drive which the S1 Project is loaning to SAIL arrived today. LES
is setting up a pseudo-person, S1, and the plan is to keep S1-related
files on [*,S1] instead of on personal areas. I will move some files
myself, like *.*[S1D,PMF], but mainly people in charge of the files will
be responsible for moving them (and making changes, if necessary).
Working files of no general interest and strictly personal garbage should
still be kept on personal areas. The purger will send me messages when S1
is over allocation. We should have enough room to keep everything
important on-line; if not, then I will ask LES for more room. LES will
also be adjusting personal allocations as a result of this reorganization.
Allocations should become effective tonight when FONDLE runs.
Jeff and I came up with the following list of initial areas, the intended
functions of which are listed here and in AREAS[DOC,S1]. I have created
some of these, otherwise create them as necessary with the UFD command,
protection = 005. If you create a different area or change one of these
names on the basis on the basis of violent disagreement, then announce the
change to S1 and update AREAS[DOC,S1].
DOC All documentation (eg. FASM.0, SMA3, AMOD.SPC, slides,
distribution lists, mail records)
SOP SOPA
PCP PCPASC
FOR S1 FORTRAN
NIL NIL compiler
MIC Microcode assembler, microcode
MK1 Mark I drawings and related files
MK2 Mark II drawings and related files
SW Switch drawings and related files
MEM Memory drawings and related files
BSM BSM-controller drawings and related files
DRW DRW libraries
SCA SCALD programs, logic simulator
PRG Miscellaneous programs source and backup, eg. FASM, FSIM, UYK, DDT
LIB Program libraries (eg. scientific functions, PCPASC runtime)
DMP Current, public DMP files
Curt
∂23-Aug-78 2224 LCW via SU-TIP#11 AREAS
To: S1
Read "INDEX[DOC,S1]" instead of "AREAS[DOC,S1]" in the last message.
Curt
∂24-Aug-78 0017 S1 *,S1 Files
To: S1
In case I didn't make it clear:
1. Keep working files on your personal areas until they are usable by the
general community (ie. debugged).
Your areas will be made big enough for this.
2. Protect files on *,S1 with protection 005.
3. Do not put passwords on new areas. Login only to 1,S1 and only for
creating new UFD's.
Curt
∂24-Aug-78 1307 MSS updating distribution files
To: S1
Whoever is in charge of whatever log and distribution files should change
the latter to insure that messages are accumulated in the relocated
log files.
-Mark
∂24-Aug-78 1554 PAT phone call from Bob Albrecht
Bob Albrecht called and wants to get the WHOLE chess bet story from you
to be published in what was People's Computer Company Newspaper, which
became People's Computer Magazine, which will become Recreational Computing
Magazine. His number is 323-6117.
∂26-Aug-78 2247 LCW Distribution Lists
To: S1[DIS,S1]:
S1 distribution lists are now *[DIS,S1], eg. S1ARCH[DIS,S1]. Curt
∂27-Aug-78 1616 Feigenbaum at SUMEX-AIM retransmission of my July 9 message re Meltzer
Date: 27 Aug 1978 1609-PDT
From: Feigenbaum at SUMEX-AIM
Subject: retransmission of my July 9 message re Meltzer
To: newell at CMUA, jmc at SAIL
Dear John and Al,
Please forgive the delay in my response on the issue of Meltzer's
continuing as Journal Editor. I forgot that I had forgotten to send
it to you!
I am inclined to agree with John's suggestion that Bernard continue
for two more years, but I feel that the period should be LIMITED to
two years. Why
1. Bernard has done a fine job. Since he has implied that he wants to
continue in part at least because of the editorial stipend, the Board
can reward Bernard for his service by asking him to continue for a
while.
2. I do not, as a matter of principle, believe in an individual
being in any one position of power and responsibility too long
(e.g. deans, department chairmen, ARPA office directors,etc.)
Bernard has been Journal Editor a long time! Is it not time for him to
let some younger person have a chance (I don't however have a candidate
in mind).
3. Bernard's active knowledge of AI will decay rapidly after he retires
because he will not be in the stream of research, students, etc.
Raphael, when faced with an analogous departure from the field (moving
to a non-AI managerial career) resigned his Associate Editorship.
4. Two years is a graceful transition period in which to:
reward Bernard; find a new Editor; and allow the new person to find
out what Bernard knows about the process.
Does any of this modify any of your previously expressed thoughts
Ed
-------
I would prefer to continue to have the option of asking him to continue
after 2 years if he is still sufficiently in contact with the field. I
don't consider it right to declare someone useless for purely
chronological reasons. Anyway we need to reply promptly, don't we.
Perhaps a conference phone call is needed.
∂27-Aug-78 1616 Feigenbaum at SUMEX-AIM what to do about Meltzer's letter of June
Date: 27 Aug 1978 1606-PDT
From: Feigenbaum at SUMEX-AIM
Subject: what to do about Meltzer's letter of June
To: jmc at SAIL
Mail from CMU-10A rcvd at 27-Aug-78 1313-PDT
Date: 27 Aug 1978 1613-EDT
From: Allen Newell at CMU-10A (A310AN02)
Subject: Re: various
To: Feigenbaum at SUMEX-AIM
Message-ID: <27Aug78 161337 AN02@CMU-10A>
In-Reply-To: Feigenbaum's message of 27 Aug 78 14:59
Ed: My recollection was that you had some more extended thoughts on
the subject and we were waiting for your input. JMC said ok, I said,
I don't know much, but ok; someone said, short time fuse (couple of
years) is better so we don't get caught in a mess if it doesn't work
out. it was your move.
It is not clear to me that letter to Meltzer is right response. He was
after us to do something on his behalf with the editorial board
directly, ie (or eg?) propose some such arrangement. I gather he felt
he could not propose it himself to the whole board, though he could
to us.
I have done nothing. I am sufficiently far from AIJ policies and action
that I'm not sure I know how to proceed. For instance, apparently there
was some issue about Bert Raphael or something -- anyway, I knew nothing
about it, except Danny showed up as the US editor.
Maybe we should collectively write a note to the editorial board. I would
think JMC or you (?) should draft it, but I would be glad to sign it,
or to endorse it, if it goes via netmail. I hate to be the drafter of it,
since I feel I'm missing some chunks of context on the life of the journal.
AN
PS Forward this to JMC is you want to evoke a coordinated thing.
-------
∂27-Aug-78 1617 Feigenbaum at SUMEX-AIM Re: various
Date: 27 Aug 1978 1605-PDT
From: Feigenbaum at SUMEX-AIM
Subject: Re: various
To: Allen Newell at CMU-10A
cc: jmc at SAIL
In response to your message sent 27 Aug 1978 1613-EDT
Thanks, Al. That's OK with me. I'll draft something, coordinate with
John, and pass it by you.
Ed
-------
∂28-Aug-78 0733 RSMITH at RUTGERS our TOPS20 system...
Date: 28 Aug 1978 1032-EDT
From: RSMITH at RUTGERS
Subject: our TOPS20 system...
To: JMC at SAIL
is now up for users. It seems to be working quite well except for
some memory glitch that DEC is working on. All the software worked
like a charm.
Regards,
Bob Smith
-------
∂28-Aug-78 2242 DCL
To: GROUP.DIS[VCG,DCL]:
**** NO SEMINAR TUESDAY 29th AUGUST *************
∂28-Aug-78 2250 Kiessig at Rand-Unix Dialnet project
From: Kiessig at Rand-Unix
Date: 24 Aug 1978 at 1849-PDT
Message-Id: <[Rand-Unix]24-Aug-78 18:49:21.kiessig>
To: earnest@su-ai,mccarthy@su-ai
cc:
Subject: Dialnet project
I got your names from someone here at Rand who mentioned that you
either have been or are currently working on dialnet.
I was just wondering what the current state of the project is, and
if you have produced any readable documents on dialnet. If you
have, and they are on-line, perhaps you could point me to them. If
some exist only in hardcopy form, then perhaps there's a way to
order them.
Dialnet sounds like a very interesting project, but it seems
news about it has been scarce of late.
Thanks,
Rick Kiessig (kiessig@rand-unix)
I presume you answered Kiessig's 28 August message.
∂29-Aug-78 1448 HEDRICK at RUTGERS account at Rutgers
Date: 29 Aug 1978 1745-EDT
From: HEDRICK at RUTGERS
To: mccarthy at SAIL
Cc: rsmith at RUTGERS
Subject: account at Rutgers
I have given you an account at rutgers, password SHAKEY. Bob Smith
has promised in your behalf that you do not intend to use it to run
your A.I. course in Interlisp.
Regards,
Chuck Hedrick
Computing Director
Rutgers LCSR
-------
Replied to earlier message before seeing yours of 29 Aug. Thanks much
for the account. I wouldn't have my dog run Interlisp.
∂30-Aug-78 1127 Tajnai at SUMEX-AIM Max Shiffman
Date: 30 Aug 1978 1126-PDT
From: Tajnai at SUMEX-AIM
Subject: Max Shiffman
To: faculty.list:
Prof. Max Shiffman, Prof. of Math at Hayward, has left a couple of messages
with me for the faculty, and I have neglected to pass them on.
He recently left a letter with me, and if you are interested in seeing it,
please let me know.
The following excerpt covers it rather well.
".you are not to use or administer any of my mathematical articles and
material without my permission, including financial permission."
Carolyn Tajnai
-------
∂31-Aug-78 1408 PBK
To: S1[DIS,S1]:
linkage protocol manuals are available under curt's desk
∂31-Aug-78 1606 LCW Linkage Conventions
To: S1[DIS,S1]:
Comment soon on the linkage conventions (copies on my desk), or forever
hold your peace.
Curt
∂01-Sep-78 0001 JMC*
renew Science (see right drawer)
∂01-Sep-78 1619 BALZER at USC-ISIB STRAWMAN PROPOSAL TO HELP FOCUS DISCUSSION AT L.A. MEETING
Date: 1 SEP 1978 1620-PDT
From: BALZER at USC-ISIB
Subject: STRAWMAN PROPOSAL TO HELP FOCUS DISCUSSION AT L.A. MEETING
To: NILSSON at SRI-KL, HART at SRI-KL, HENDRIX at SRI-KL,
To: MOORE at SRI-KL, SACERDOTI at SRI-KL, TENENBAUM at SRI-KL,
To: BOBROW at XEROX-PARC, FIKES at XEROX-PARC,
To: GOLDSTEIN at XEROX-PARC, LENAT at SUMEX-AIM,
To: BUCHANAN at SUMEX-AIM, CCG at SU-AI, DENICOFF at USC-ISI,
To: FEIGENBAUM at SUMEX-AIM, JMC at SU-AI, TW at SU-AI
cc: BALZER
STRAWMAN AI PROGRAM SCHEMA
I. 5 Year Basic Research Program--[proposed: KNOWLEDGE REPRESENTATION]
A. Pursued via separate investigations of core research issues
[see STRAWMAN list below]
B. Progress measured via independent criteria established for each core
research issue
II. 5 year development plan to follow research plan
A. Demonstrate ability to integrate basic capabilities developed separately
in Basic Program
B. Initiate several separate mission oriented projects
******************************************************************************
Strawman List of Core A.I. Issues
1. Representation Theory -
Formal investigation of the Space of Representations, the formal
mappings between them, and strategies of selection.
2. Fusion: Multi source knowledge integration -
Coordinating multiple streams of data, recognizing common
objects or events in separate streams, integrating knowledge of
common objects and events from different types of sensors, and
sensor control.
3. Multiple uses of knowledge -
Using the same knowledge for multiple purposes such as
recognition, generation, planning and execution.
4. Theory Structuring: Knowledge acquisition and assimilation,
knowledge engineering -
Technology for forming individual "rules" into theories, testing
and debugging such theories, analyzing them for deficiencies,
and identifying causes of failure.
5. Process understanding and modification -
Process analysis, documentation, and explanation; explicit
process modifications, goal directed process modification;
process simulation and description with actual and symbolic
data.
6. System Self Awareness -
Multi-level explanation capabilities for previous behavior,
causes of non-occurring behavior, conditions required for
specified behavior, effects of altered environment or system on
subsequent behavior. Advice taking capability to alter system
behavior.
7. Knowledge Compilation -
Automatic and/or programmer directed optimization of
A.I. systems to substitute compile time analysis for run-time
decision making (e.g., Harpy speech system).
8. Robustness -
Develop technology for graceful degradation of system
performance under partial and/or errorfull information.
9. Partial Pattern Matching -
Develop technology for dealing with inexact or "fuzzy" data
and/or situations. Trying harder when no response is elicited
(threshold reduction). Analogical reasoning to produce
alternative viewpoints.
10. User Modeling -
Develop capability to model users knowledge of system state and
his ability to alter his knowledge base. Gear system
interactions to achieving desired transfer of information to
user (computer coaching). Interpret user input on basis of user
model and goal structure.
11. Search Control -
Investigate alternative architectures and strategies for
controlling search.
-------
∂03-Sep-78 1053 JLH Pascal improvement project
To: JMC, GIO, SSO, DPB
GOOD NEWS!
Armando has completed the load and go version of the compiler and it has been
tested fairly throughly.
Preliminary benchmark comparasion:s 815 lines (450-500 with code).
CPU time - compile only
Pascal 3.362 s
PASSGO 2.938 s 12% increase
COMPILE and LINK - ELAPSED TIME
Pascal 15s
PASSGO 7S 65% increase
The latter figure is more mprotant as it measures overall
usage of the system by the compiler. Additionally, the above represents a lightly
loaded system - we expect the improvement to be more dramatic as the load increases.
Naturally the code is identical so execution speeds are the same.
All Pascal programs except those with external procedures (obviously)
will run using the load and go version.
∂04-Sep-78 0017 MRC
hippie birthday
∂04-Sep-78 0028 LLW Proposals and Requests-for-No-Cost-Extensions
To: JMC, LES, wiederhold at SUMEX-AIM
CC: LCW, BS, LLW
In my message to you on this subject of mid-August, I noted that time was
short with respect to both proposals for the upcoming academic year (or
Federal fiscal year) and requests-for-no-cost-extensions of your present
contracts with LLL for S-1 Project R & D. I now must call your attention
to the fact that time is running out on both of these matters, and that I
have no input from either of your groups, on either score.
If this situation does not change significantly, on/about Sept. 15, the
good folks in the LLL Budget Office will commence (over my protests, to be
sure, but will do so nonetheless) withdrawing all those funds from both
Stanford CS contracts which they, in their sole and exclusive judgment,
project to be unused by the end of the contract period (which is now Sept.
30). If this occurs, on Oct. 1, neither Stanford CS effort will have
LLL/S-1 funding, because it is by now impossible to get a new proposal
routed through Stanford, LLL, DoE/SAN, LLL, Stanford and finally through
LLL, before the present ones expire.
It is therefore now IMPERATIVE that each of your groups write to Milt
Eaton, your LLL contract administrator, requesting extensions of your
present contracts, moreover in a form that I can credibly countersign from
this end (e.g., that Curt has gone over with you beforehand). This
request should state the period for which the no-cost extension is sought,
the reason why the work could not be completed on time (e.g., late start,
staffing delays, etc), the reason why it is expected that it can be
completed within the period of the requested extension at no additional
cost, and so forth. It needn't be over 200-400 words--it just needs to
refer to the contract number, and say the necessary things.
I also once again urge you to submit your proposals for the Federal fiscal
year upcoming very soon, so that you can plausibly propose for the entire
period. As I noted previously, we'll be back to the 'Perils of Pauline'
mode of contract renewals and funding gaps, if you let this 'pad' of time
which we've developed slip away, through present inaction. [Betty, please
note that I'm attempting to preserve you from the problems of the past
and, if they do develop again, it's not because I'm not making every
effort to forestall them.]
I've asked Curt to coordinate these matters with both of your groups while
I'm away this coming week, so that he and I can have proposal and
extension-request material over here at LLL with which to work to keep
your groups funded, when I return in a week.
Thanks for your timely cooperation.
Lowell
∂05-Sep-78 2250 Wiederhold at SUMEX-AIM S1 meetings
To: S1
Date: 5 Sep 1978 2239-PDT
From: Wiederhold at SUMEX-AIM
Subject: S1 meetings
To: s1 at SAIL
cc: pickering
Report on the last two S1 meeyings follow.
ALSO : LCW, EJG,PN have received copies of
the proposal being prepared for livermore. You are all
welcome to read it and comment rapidly.. Gio
←←←←←←←←←←←←←←←
S1 meeting 3 august 1978
attending: alp, ark, blh, ejg, fac, fc, hal, lcw, peg, pn, ptz, gio,
(ko was not picked up)
l. LOGS (lcw) - There have not been many log entries recently. This
may reflect the lack of major advance. In the future logs
should be maintained. Major changes should be documented
with around 200 lines of entries.
2. Architecture update this month (lcw). Simulator will be up to
date in the next couple of weeks.
3. New version of SLAC PASCAL documentation has been brought over
(ejg). The documentation lists switches, etc. 3 versions
of line lengths 72, 80, 132 will exist. People approaching
long lines should read the documentation.
4. New PDP10 Pascal is up (peg). This version does not ignore
underscores and will count them in 10 character id lengths.
It does however expand tabs correctly. "It's wonderful."
(PCPH1 handles E-directories ok.)
5. Users of Pascal should restrict themselves to standard subset of
Pascal so that program will run on the various versions here
(lcw). PASCAL conversion document exists on PASCNV.0 [sid,
pmf].
6. Optimization (ptz) - (1) Change subroutine linkage to new linkage.
That entails first few vriables into RSpacefor PCPASC. This
creates a factor of 2 saving in the FORTRAN kernels.
(2) Peephole optimizations are still in the future.
7. FORTRAN (pn) - Functions and IO should be working by tomorrow.
Within the next 2 weeks there will be documentation testing.
There are still no trig functions due to the lack of external
linkage.
8. Numerical Routines (lcw) - Work is progressing well but is very
slow, due to the complexity of the routines.
!
9. SOPA priorities (lcw,gio,peg)
PCPASC/SOPA CLEANUP SUMMARY: 20JUL78
*** Prioritized Items
done:
Reimport PCPASC and associated routines. PCPASC should not blow
up when it encounters errors. Flag all SLAC-to-PDP10
transporation changes and all PDP10-to-S1 transportation changes.
in progress:
1. Fix the SET of CHAR problem in PCPASC and SOPA.
2. Make PCPASC run on the simulator.
do:
1. Use the up-to-date opcode strings and values from SMA2.CHG.
Eliminate use of ALW and NEV. Use MOVMQ for constant-length
moves. Use the new block-descriptors. Don't output alignment
parameter for MOV instruction in PCode.
2. Conform to the new subroutine linkage convention, including
the new traceback convention. Use 16-character routine names.
Output new LDI format.
3. External procedures should be added to PASCAL and PCode
(according to the LASL definition). SOPA should use the
same calling convention for external procedures as for
standard procedures.
4. New PASCAL runtime (coded mostly in PASCAL) conforming to new
linkage conventions.
5. Get the statement-counting and other new features (see Sassan)
working in PCPASC and SOPA.
6. Put PCPASC and SOPA together as one job. Fix the user interface
to ask for file names in a better way.
7. Improve PCPASC error reporting mechanism. Improve SOPA error
reporting to give much more context about a fatal error, including
traceback. Add more options, as necessary.
8. Add double- and half-word reals to PASCAL/PCode syntax. Fix
PCPASC and SOPA to handle these precisions.
9. Modify PASCAL syntax, PCPASC, and SOPA to allow packing into
multiple integer sizes. Convert to bit-addressed PCode.
10. Fix various bugs in SOPA and PCPASC which are documented in the
mail.
11. Change all field widths for integers output from PCPASC to
use FLDW(I) instead of fixed width.
12. Use new numerical routines.
13. Fix BNDCHK by putting cause in a temporary register.
14. Define a standard character set. Put out a formal document
defining the character set.
*** Miscellaneous Items:
1. Does SOPA pay attention to the 25-bit displacement limit for variable-base
addressing? This could be a bug.
2. Fix the disassembly to put out FASM-style listings which can
actually be assembled. This listing should be symbolic when
symbols are passed through PCode.
3. Pass symbol table through PCode.
4. [13MAR78 EJG] (LCP's bug): If PCPASC is fed the normal PASCAL
statement "for CH := 'A' to 'Z' do <something>", it generates
P-code which treats integers and characters interchangeably.
This causes SOPA to stop with a user error message since it
is not strictly kosher P-code. The best fix would be to fix
5. [02MAR78 EJG] Make a fairly detailed hand pass through
ASMNXTINST checking for bugs of the following form:
Freeing register(s) prior to their actually being free in
order to gain better use of temporaries, assuming
that they will then actually be free after exactly one
instruction is emitted, BUT then emitting an extra
intervening instruction (as had been the case with a call to
MOVE_AND_FREE_RTB in four instances).
!
10. Linker (ark) - still coding, getting rid of funny names.
11. Testing (lcw, pn) - Arrange tests in one file (after debugging)
so that "everything" can be tested at once. Every line in a
program should be tested on all (?∞ input combinations
- use assertions on correct entry conditions. Check at least
extreme and boundary cases. Use both "real" input and
manufactured input for testing. (Source - Fortran log.)
12. Register designations as M, D, R (gio)
FORTRAN - HOLD.
PASCAL - Parallel to subroutine linkage. SOPA will demote
R to M but Sites is indicating that he can't demote stuff.
This might be solved by an intermediate level D (between M,R).
PCPASC would output D, D-level could have no aliases. SOPA
would promote D to R. An optimizer could also promote D to R.
Unpromoted D would be treated as M. No demotion would be
allowed.
13. Documentation(lcw) As the end of the summer approaches, remember that
we need formal documents describing the summer's work as follows:
A. LINKER implementation manual.
B. SOPADOPE revised and augmented.
C. SMA-3.
D. PAIL revised.
E. PRUN revised.
F. S1FORT implementation manual.
14. All completed routines should be tested to achieve a common level of
confidence. Systematic testing should involve:
A. Integrating all test cases into a single program with a good
user interface.
B. Executing every line of code at least once.
C. Verifying that every routine produces a well-defined output for all
possible valuations of its input parameters and global variables.
D. Checking boundary cases for all routines.
E. Checking against real and manufactured input.
Also : Use standard PASCAL
15. Changing PCPASC/SOPA(lcw). Write down plans for major changes and
consult experts before coding. Log entries should be 100-200
lines for a major a training guide to new programmers after the
end of the summer. They should be accompanied by a differences
file and new test cases.
16. PCPASC/SOPA Coding[LCW]
I trust that everyone coding in PCPASC/SOPA will
1. Fix any existing trivial errors observed (documentation of same
is a matter of judgement).
2. Possibly fix, but ABSOLUTELY DOCUMENT any existing non-trivial
errors observed.
The appropriate place for documenting bugs, improvements
needed, and changes made is IN THE LOG SECTION OF SOPA.LOG, and
possibly in another section of SOPA.LOG as well.
Documenting Sessions[LCW]
I have found that an effective technique for documenting a session
with PCPASC/SOPA is to
1. note all items to be worked on in my personal log before the
session starts,
2. record in my personal log all significant events occurring during
the session,
3. then to have a giant log-writing session immediately after the
work has been essentially completed (perhaps not fully tested),
but within a few (2 or 3) days of its initiation (so my notes are
not too obscure).
4 Of course improvements which you choose not to make on the spot
should be noted IN THE CODE.
Minutes S-1 meeting 20Jul78 Small Conference Room at AI-Lab
Present mlb, ark, hwc, hal, peg, fac, blh, alp, gls, pn, fc, pbk,ejg
ptz, lcw, gio, ko
absent bth
OPERATIONS (peg,ejg)
Backup of UDP will be done Fridays or Thursday evenings. When a new
disk drive comes to the AI-Lab, allocations can be increased.
There are problems in high segment sharing. Two job slots are needed
to run PDP-10 PASCAL jobs. A 9600 baud line between SAIL and LLL
should work soon (hwc) and can be used to transfer data. Program testing
will remain mainly at the AI-Lab using the simulator.
PASCAL (ko etal)
Standard PASCAL from San Diego seems to be established as a minimum
level.
Ko will determine how MODEL gets to have OWN variables, and present
their solution for our consideration.
Expect soon (underscore) instead of (left arrow). New cross reference
does not handle #, else works. Runtime still needs a change. Option
U and bug has been fixed. EOF not recognized.
LINKER (ark)
Parts of code are beginning work. 30/36 bit address is open.
FORTRAN (pn etal)
Complex and conversions (pn), subroutine calls (fac), format conversions
(fc) are in progress. The question of base register allocation by
procedure versus global allocation was raised.
SOPA OPTIMIZATION (pbk)
Analysis of good tasks is in progress. Register allocation is
promising. There is quite a bit of dependency on Sites work on the
PCode-to PCode optimizer in San Diego.
HARDWARE (lcw,pmf)
The machine is doing better, but routine use of the Mark I from here
is quite aways off.
←←←←←←←←
S-1 Meeting 17 August 1978
Attending: ALP, ARK, BLH, BTH, EJG, FAC, FC, GLS, HAL, KO, LCW, PEG,
PN, PBK, PTZ, GIO
1. AI LAB USAGE - [GIO, LES]
The number and intensity of DEC-10 usage is impacting other
users. Please be considerate, working at night is both
efficient and considerate.
2. LINKER - [ARK]
Coding complete, testing in progress.
3. OPTIMIZATION - [PTZ]
Linkage conventions are being written up.
4. MANUAL - Addressing section complete. Preliminary manual will be
available by the end of September.
5. FORMAL INSTRUCTION DESCRIPTIONS - [LCW, BTH]
P-bit coming slowly due to work on LISP compiler.
6. PCPAC - [HAL]
In progress.
7. YUCC - [ALP]
Work on console routines in progress.
8. FORTRAN - [PN]
Most of code is written but IO is not yet working.
9. PCPASC, SOPA-PCPASC - [PEG]
Accepts large sets (144 members). Being debugged.
10. OS - [EJG]
Working with LLL hydrodynamics code that may eventually be
used on an S-1 multiprocessor.
11. HARDWARE - [LCW]
Actual S-1 Report - DDT works, most compiled programs
accessible via DIAL on a 600 baudline, but someone at the
other end required to set up link.
12. FALL TERM PLANS
PTZ - possible part time, depending on school work.
PBK - no.
ARK - not decided but possibly available 1/4 time (O.S.).
BTH - 1/4 tiime on manual.
BLH - possible 1/2 time depending on work at SLAC (manual,
P-code).
KO -possiible 1/2 tiime depending major-switching (model
symbolic debugger.
GLS - return to MIT but informal help with manual. Possibly
available on LISP project next summer.
HAL - 1/4 time available.
ALP - support work only.
FAC - leaving.
FC - full time.
PN - available 1/2 time.
PEG - 1/4 time available.
EJG - Hertz Fellowship to do research in related areas.
13. PROJECTS
a. Global Optimization of P-Code (interface with Sites)
b. Linker (completion)
c. Fortran (completion, maintenance)
d. PCPASC (maintenance)
e. SOPA (maintenance, linkage convention)
f. System Programming Language (model)
g. Manual (completion)
h. Multiprocessor Architecture Design & Operating System
i. Numerical Routines
j. File System
k. Debugger
-------
∂06-Sep-78 0952 Nilsson at SRI-KL (Nils Nilsson) Report on the SRI AI meeting
Date: 6 Sep 1978 0952-PDT
From: Nilsson at SRI-KL (Nils Nilsson)
Subject: Report on the SRI AI meeting
To: Feigenbaum at SUMEX-AIM, JMC at SU-AI, TW at SU-AI
cc: Nilsson
It was concluded at the "AI in the Eighties Meeting" (held at SRI on
Aug 28 and 29) that there is a reasonable possibility for a joint govt
agency sponsored major new AI project. The participating agencies are
likely to be NSF, ONR, and ARPA. E. Barrett of NSF mentioned a new
speech project as a possibility. There was much discussion of the pros
and cons of a new speech project, of the pros and cons of any new, highly
managed project at all, and of the pros and cons of joint sponsorship.
It was more or less the consensus of those attending the meeting that
there might be an excellent opportunity here to redress the imbalance
between basic and applied AI research. Denicoff and Barrett seemed to
agree that a major basic AI research program might be fundable provided
that it was well-defined and had the general support of the AI community.
It was agreed that a major subtask was to sample responsible opinion
of AI researchers about how best to take advantage of the possibility of
getting some major new basic research funds.
Another major subtask is to synthesize these opinions into an
outline of a research program (or programs). We would then take
this program definition (in person) to Fossum to discuss its possibilities.
An output of the Fossum meeting would be an appraisal of whether or not
it would be worth the effort to put together a carefully drawn up
proposal and research plan. The time scale is a bit short, because
other disciplines will also be competing for the possibility of getting
new basic research money. (Denicoff and Barrett agreed that Fossum
leans toward new basic research initiatives in whatever disciplines
appear to be most able to pay off on such an investment.)
There was much discussion of the need for someone to take responsibility
for seeing that "something really happens" (if that's what we all
want). Of course, everyone took 6 steps backward when a call went out
for someone to take the lead. I was persuaded to get things started.
It was agreed that there ought to be an LA area meeting. Balzer has
set a meeting up for Fri. Sept. 8, which I will attend. Denicoff will
organize an eastern meeting sometime in Sept, which I will attend. I
agreed to serve as a focal point for sampling and organizing opinions
until the eastern meeting at which time the leadership question will
be reopened. Probably we should aim for a Fossum audience sometime in
early October.
Since you (TW, JMC, EAF) weren't at the SRI meeting and undoubtedly would have had
valuable comments had you been there, I would like to talk about all of
this with you together or individually at your convenience. Please
respond when you arrive back in the area, and we can set up something.
--Nils
-------
∂07-Sep-78 0934 Sacerdoti at SRI-KL (Earl Sacerdoti) ''white paper'' on distributed robotics
Date: 7 Sep 1978 0933-PDT
From: Sacerdoti at SRI-KL (Earl Sacerdoti)
Subject: "white paper" on distributed robotics
To: Nilsson, Duda, Hart, Hendrix, Moore, Sacerdoti, Tenenbaum,
To: Bobrow at XEROX-PARC, Fikes at XEROX-PARC,
To: Goldstein at XEROX-PARC, Lenat at SUMEX-AIM,
To: Buchanan at SUMEX-AIM, CCG at SU-AI, Balzer at USC-ISI,
To: Woods at BBN-TENEXD, Denicoff at USC-ISI,
To: Feigenbaum at SUMEX-AIM, JMC at SU-AI, TW at SU-AI
Below is the position paper I promised to produce at the recent workshop on
AI in the 80's. It's 13 pages long; you may want to list it. Comments are
encouraged. --Earl
!
The Distributed Robot as a Focus for AI Research
Earl D. Sacerdoti
SRI International
I Introduction
At the workshop on AI in the 80's I suggested the concept of a
distributed robot, as a "first-level application" to drive research in
many aspects of AI. What follows is a more detailed description of the
concept, the research it might drive, and some potential applications.
The text is basically a revision of an internal SRI memo I wrote in May.
"It has sometimes been argued that part of the stimulus to
laborious male activity in 'creative' fields of work,
including pure science, is the urge to compensate for
lack of the female capability of giving birth to
children. If this were true, then Building Robots might
be seen as the ideal compensation."
-- Sir James Lighthill
While Lighthill's argument is rather facile (i.e. a cheap shot) it
is the case that robotics as a field has tended to assume implicitly
that near-human-level performance would come from human-like, or at
least mammal-like automata. I will attempt to argue below that we can
drive progress on a wider range of fundamental AI issues, and provide a
1
!
wider range of intermediate-term practical fallout for our government
sponsors, by scuttling this implicit assumption.
A. The Disbot
Here's my view of what a disbot (for DIStributed roBOT) is:
* It consists of a fairly large number of specialist
machines, perhaps 6-10 in an experimental environment, 50-
100 or more in a first application.
* Most of the specialists are no bigger than a breadbox.
* Most of the specialists can be immobile (but all are
portable).
* The machines are specialized to particular suites of
sensors and/or effectors.
* All the machines can communicate over very short distances
(needs not be more than line-of-sight) (the bandwidth will
probably need to be pretty high).
* Each machine can be presumed to carry the computing power
of, say, half to most of a KA-10. We might find we can get
away with much less for most of the specialists.
* Particular specialists would include: lifters, pushers,
range sensors, various types of visual sensors (at
different levels of acuity and speed, and perhaps for
different portions of the spectrum), auditory sensors
(including listeners for particular sounds and commercial
voice recognition devices), displays, keyboards, clamps,
sensor- and effector-less communications relays, and task-
specific specialists of all sorts.
* The most serious pragmatic problem I can see is how these
things can be powered without making the specialists too
bulky or too short-lived.
Any particular collection of individual specialists will almost
certainly be cheaper than a monolithic automaton with the same
2
!
capabilities. Furthermore, because specialists can be included
redundantly and the overall system can reorganize to respond to the
available resources, the disbot is fail-soft and can be made fixable in
the field.
II What Research Will a Disbot Project Drive?
Because the concept of a disbot violates some nearly universal
assumptions, there is good reason to think that its development will
induce work in novel directions for a number of specialties within the
field of AI.
A. Vision
This is first because I'm least sure about it. Various specialists
can extract different images of intrinsic properties of a particular
scene asynchronously. Some of the ambiguities that remain after the
information in the intrinsic images is combined might be resolvable with
a second picture or pictures from a different perspective. We can tie
in some problem solving and execution monitoring by dynamically
determining when an ambiguity calls for a second picture, with what
sensor and from what position.
Another cut at distributing the work of scene understanding is to
have alternative sensor/processor combinations that provide different
3
!
levels of detail or accuracy for different costs or utilization of
resources.
Since the disbot will have multiple eyes in changing
configurations, it might be interesting to push on what can be done with
multiple, poorly-calibrated inputs.
The disbot's vision system will provide an excellent testbed for
developing and trying out ideas about "knowledge fusion" (see Subsection
D below).
B. Distributed Artificial Intelligence
The computational environment of the disbot is inherently novel,
since the level of coupling of the various processors is simultaneously
tighter than that of resources on a distributed network, and yet looser
than that of ganged-multiprocessor configurations such as c.mmp or cm*.
The development of a distributed robot will drive new work in factoring
AI processes among multiple, asynchronous machines. Work to date in
distributed AI has not yet faced the issue of factoring the intelligent
processing in a "natural" way among the available processors (Reid
Smith's contract nets basically implement a standard top-down control
strategy on similar processors with different data; Erman & Lesser's
distributed HEARSAY factors the speech understanding problem by time
segments, which is not a natural factoring). Because the specialists
are functionally different, we will be forced to work on more natural
methods for breaking down problems.
4
!
The disbot will provide a context for the investigation of issues
related to the centralization and distribution of control. The
coordination of asynchronous processors at the level of goals and
actions (rather than machine primitives and function calls) will require
a (hopefully synergistic) collaboration between AIers and more
traditional computer scientists.
Since different representations of the perceived "state of the
world" will be appropriate for specialists with different capabilities,
a disbot project will drive work on the very basic issue of the
integration of multiple representations.
These problems could be worked on in the abstract, but the disbot
will provide a concrete environment to try out our ideas.
I think this is probably the aspect of the research that will be of
highest technical risk. If the specialists can't communicate with a
relatively small portion of their computational resources, the whole
concept won't work. On the other hand, the payoff from the distribution
of intelligent processing will be very high, in terms of both increased
capabilities for AI systems and novel problems of theoretical interest
to work on.
5
!
C. Problem Solving and Execution Monitoring
Driving the set of specialists around, taking into account the
differing capabilities and requirements of each type and the redundancy
arising from multiple specialists of the same type will force new work
in problem solving, focusing on planning for asynchronous activity and
on issues related to the deferral of detailed decision-making.
In many situations, no individual specialist will have all the
information necessary to create a plan of action to achieve a given set
of goals. The disbot in the aggregate might well have that information.
Thus the disbot will provide a context in which to study issues relating
to the development and integration of partially specified plans, and to
the development of plans that partially achieve a set of goals. The
distributed, asynchronous nature of the activity these specialists will
display will force novel work in execution monitoring and efficient
replanning as well. The need for scheduling as well as simply planning
may provide the impetus for the development (finally!) of an integrated
system for generating plans and taking time into account in scheduling
their execution.
6
!
D. Knowledge Fusion
One of the subject areas of highest interest simultaneously to the
AI research community and the AI funding community is knowledge fusion
-- the integration of information drawn from diverse sources and input
modalities into a comprehensive view of "what's out there." The disbot
provides an excellent testbed for dealing with these issues. The
specialists will acquire independent, partial views of the world based
on their particular locations and suites of sensors. The task of
integrating these partial views in an asynchronous, robust, non-top-down
fashion is a significant long-term challenge for AI research.
Aspects of the knowledge fusion problem that a disbot project would
have to face include: perception of sensory data in noise; communication
of sensory data in noise; integration of information from different
locations and perspectives; integration of information from different
sensory modalitites; integration of information expressed in different
representations; dynamic reconfiguration of communication and control
paths to deal with a changing mix of sensory data; dynamic, cost-
effective tasking of sensors and effectors to resolve ambiguities and
fill in gaps; modeling of the accuracy of sensors under varying
conditions; creating and manipulating partial interpretations from
partial data; and performing all of the above in a fail-soft manner so
that degradation of the overall model of the world is gradual as
elements of the system are removed.
7
!
I think it would be hard to find a knowledge fusion task that was
better subject to experimentation than the disbot. Since the sensors
and effectors will be rather well understood and under control in the
laboratory, it will be possible to model them well. Furthermore, it
will be easy to set up experiments under controlled conditions. The
functioning of the disbot will provide a rather precise measure of how
well the fusion process is being done.
E. Coping with Uncertainty
That the disbot is tasking its distributed sensors to deal with a
dynamically changing environment sounds a lot like SRI's application of
rule-based inference systems to EW. Maybe the executive for the disbot
looks like a distributed expert system?
F. Language Understanding
Perhaps surprisingly, the disbot will even drive research in some
important aspects of language understanding. Language understanding can
be viewed as the process of communicating between intelligent entities.
Issues such as planning and interpreting "speech acts" and focusing on
particular subjects must be faced in developing the high-level
strategies that the disbot specialists will use to communicate
information among themselves. The disbot will provide a relatively
simple context for dealing with these issues independently of the other
aspects of language understanding, and will generate useful hypotheses
for mechanisms to be integrated into a language understanding system.
8
!
III Potential Applications
Because of the novelty of the disbot concept, there are a large
number of new capabilities that the research we would do could support.
In fact, even a partial success with the disbot might create the "robot
gap" that McCarthy is fond of mentioning. Below are some musings on
applications, some rather near-term, some further out. Almost all of
them fit into the "90%-of-the-way" approach to automation, in which the
last (toughest, most generalist, most judgemental) 10% of the process
remains under human control or supervision.
A. Reconfigurable Communications
A string of machines, each within line of sight of two or more
neighbors, can provide a communications link. Since the machines will
need to be able to address one another in any case, it would not be a
difficult addition to allow a human at one machine to address a human at
another within the local network. The link need not be restricted to
two-way only.
9
!
B. Fail-Soft Performance
Because the disbot is spatially and logically dispersed,
reconfigurable, and has independent power sources at each machine, the
disbot will not fail in an all-or-nothing fashion. Individual
specialists can fail and be replaced by others. This is of critical
importance to the military. And, yet once again, there is a new way of
looking at things once the implicit anthropomorphic assumption is
discarded: since the functionality of the robot is redundant and fail-
soft, the individual components can be made somewhat less reliable and
hence much cheaper. This has major implications for the salability of
the whole concept.
C. Applications for Planetary Exploration
Because of the fail-soft characteristic described above, a disbot
is a much better bet than a monolithic robot for planetary exploration.
What's more, improved specialists can be shipped in increments years
apart (assuming a second shipment can land near enough to the first).
While it's tough to make a big lander mobile, it's much easier to make a
distributed one mobile. Even if it takes a couple of mover specialists
six months to set up the rest of the specialists for a particular
experiment, the addition of mobility would have a major impact on the
scientific value of each mission.
10
!
D. Applications for Space
The reconfigurable nature of the disbot makes it attractive for
activities such as manufacturing in space, where a limited amount of
mass has to be allocated to the manufacturing facilities. A disbot
manufacturer could be altered to provide a wider range of capabilities
than a monolithic device of the same size and cost of development.
Maybe a disbot can build structures which in turn can support more
sophisticated manufacturing.
E. Flocks of Remotely Piloted Vehicles (RPV's)
This application was suggested by Keith Uncapher. The
communication of information among semi-autonomous RPVs could result in
superior performance and fail-softness for a mission undertaken by such
a flock. The approaches to communications protocols and knowledge
fusion developed in studying disbots should provide a foundation for the
development of these RPVs.
F. CAD/CAM
The solutions we develop for controlling multiple specialists will
be directly applicable to the gradual introduction of computer-
controlled automation into the factory environment. Relatively
autonomous manufacturing stations will need to have limited
communication with other stations. Planning for complex assemblies will
11
!
involve the integration of independently developed partial plans just as
for the mobile disbot.
G. Automated Warehousing and Inventory Control
Many of the functions carried out in a warehouse could be automated
by a robot system. A disbot makes particular sense in environments
where space is tight (as in Denicoff's submarine example), since the
individual specialists need not all be together for every subtask.
H. Forget Mobility
A number of potential applications come to mind that don't require
the disbot to be mobile at all. An office environment could be
sprinkled with immobile specialists that could control the office
environment, provide reconfigurable communications, and provide for
after-hours security. New military sensor systems might be developed
within this framework. Man-made hostile environments (e.g. nuclear
power plants) could be sprinkled with specialists (here the fail-soft
nature of the disbot is important). Other complex process control
problems (for example, managing a tank farm, refinery, or power grid)
might also fit into the paradigm of distributed robotics.
12
!
IV Conclusion
A multi-year program aimed at the development of distributed
robotics systems would provide an umbrella under which a great deal of
fundamental research in Artificial Intelligence could be pursued. It
would provide a realistic environment in which to evaulate the research
results produced. It would provide, at the end of the program, a set of
prototypes that would provide models for truly useful applied systems.
Because the concept of a disbot is a novel way to look at a set of old
problems, it is likely that the payoff from a distributed robot project
will be high both in conceptual progress for AI and in potential
practical value.
13
!
-------
∂08-Sep-78 0007 LCW Contracts
SAIL needs to request a contract extension immediately to aviod a funding
gap. Can you help?
Curt
∂08-Sep-78 0010 REM WORKING INTCOM VERSION 2
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
I finally found time to debug the version of my Interval-Refinement
data-compression protram. It now works great (but slow of course).
Here are some files I tried compressing, with original size and compression
ratio. Note that because it is one pass it is more convenient to use than
old methods such as CRU2 and SPINDL=CRU3. Because it uses pseudo-baudot
code suggested by Mike Wilber it can crunch any PDP-10 file regardless of
strange characters, unmatched <cr> or <lf> characters, nulls or rubouts,
low-order-bit nonzero, unexpected sequences of characters...
FILE ORIGINAL SIZE COMPRESSION RATIO = CRUNCHED/ORIGINAL
AIWORD.RF 13800 .437+
MAIL.BH 19840 .338-
POX.XGP 36999 .397+
ILISP.UCI 73984 .241-
NOTICE. 90112 .317-
LSPARC.RPG 92800 .350-
SIMULA.JP 129152 .228+
All above files are on [UP,DOC]. Parameters were default except
250000 word allocation-limit and NFORK=6. Object program, before
allocation, occupies 6 pages (3 k). NFORK=6 is optimum for small
files about 10k, but for larger files like SIMULA.JP NFORK=8 is better,
where the C.R. becomes .226+ which is insignifically better than above.
When crunching SIMULA.JP, it takes 7 minutes 3 seconds, for an
effective reduction rate of 8500 bits per second -- computed from
the formula rate = (original - crunched) / time, thus the speed of
reducing disk consumption is about the same as archiving files on
the datacomputer when the computer isn't busy, with the advantage
that it is somewhat more reliable. Hopefully I'll be able to speed
up the program greatly, but we'll see...
∂08-Sep-78 0233 REM Addenda on new Interval-Refinement compression program
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
I have added accounting of tokens compressed and tokens searched
in the histogram, with division at end to give average search depth.
For a small file, it is 6.2-6.3 but for SIMULA.JP (129k original,
22.6% of that after crunching) the average search depth is 2.8-2.9
which indicates linear search performs almost as well as a more
complicated search would, thus reprogramming the search algorithm
won't be worthwhile. Pruning of left-context tree remains as only
obvious way to improve the algorithm.........
∂09-Sep-78 0341 REM More results from my new data-compression program
To: CRUNCH.DIS[1,REM]: CRUPUB.DIS[1,REM]:
E [CSP,SYS] 184704-->60069 CR=.325+ DEPTH=3.79+
ATSIGN.474 83456-->27609 CR=.331- DEPTH=4.30+
FAIL [CSP,SYS] 76416-->25045 CR=.328- DEPTH=4.26+
WUTHER [LIB,DOC] 173184-->50351 CR=.291- DEPTH=3.17+
GRIMM [LIB,DOC] 313216-->95190 CR=.304- DEPTH=2.96+
CR = CRUNCHEDSIZE/ORIGINALSIZE
DEPTH = AVERAGE DEPTH OF LINEAR SEARCH INSIDE ALGORITHM
WUTHER and GRIMM are English text, others are PDP-10 assembly language
with English-language comments intersperced. Note, this is one-pass,
and all results were better than 3:1 (CR=.333).
∂10-Sep-78 2247 REG via SU-TIP LOTS- If you're back in timee
Lieberman has found us $70,000 for buying the processor and renting the
memory for 3 months.
DEC wants to reent the memory for nine months @$7,000/mo.
Ampex would rent the memory for 3 months.
I want to buy the processor, the memory, a tape amd a disk
from DEC for $200,000. I could sell the disk for $30,000 locally.
But Lieberman doesn't have the $170,000 right now, and he wants to see
that it's effective before he spends it.
We have decided to have one more try at getting DEC to agree to a three
month rental. If that falls thru, we'll go to Ampex. Our decision
by 9/15. (should have been 8/15, but you know how these things
go.) With luck, we'll have the stuff in place before the worst of the
crunch.
∂11-Sep-78 0457 REM Now is the time...
To: JMC, LES, RWG, WD, ES
Now that I have proven both the logical correctness of my crunching method
(program doesn't generate any error messages, and uncrunch inverts crunch)
and the usefulness (program this weekend did some impressive compression
ratios, reported in earlier messages), it is time to contact Neils Reimers
and see about getting my stuff patented. I'll probably start today (monday)...
∂11-Sep-78 1410 MARK SHERMAN at CMU-10A (C410MS40) DFTP Allocation on the Datacomputer
To: S1
Date: 11 Sep 1978 1708-EDT
From: MARK SHERMAN at CMU-10A (C410MS40)
Subject: DFTP Allocation on the Datacomputer
To: S1 @ SAIL
Message-ID: <11Sep78 170846 MS40@CMU-10A>
The S-1 project now has its own node on the datacomputer with 750
megabits of storage. Those who are interested in setting up
directories under this node should contact me, and I will give it to
you as soon as I can. An example script of dftp use is given in
script.txt[1,mss] (though it doesn't show the actual use of the s1
node. This will soon be changed.)
-Mark
∂12-Sep-78 0850 MINSKY at MIT-AI (Marvin Minsky) klaut still lives
Date: 12 SEP 1978 1149-EDT
From: MINSKY at MIT-AI (Marvin Minsky)
Subject: klaut still lives
To: MINSKY at MIT-AI, jmc at SU-AI, cooprider at CMU-10A
To: reid at CMU-10A
This can go on forever. It occurred to me that the whole thing
might be dealt with nicely by writing
a "profile" for the New Yorker. My idea
is simply to annotate the great
Robot file that accumulated ilast fall. Two purposes
would be served:
1. Klatu would be treated lightly but thoroughly;
the "annotation would educate the public gently about
various technical problems in making the household robot,
and would nicely explain a lot about computer science, AI, and how research
is done.
2. The working of our society, the first to
use computer network as a normal social instrument would be
self-documented. Especially interesting is
the way the group spontaneously organized to do the investigations,
debug the reports, and informally disband when the job seemd complete.
It would be some work to debug the article. My idea is that the ARPANET messages
would be the main part, with explanations and introductions of the participants inserted between them.
I would like reactions to this idea. If several people are
interested, my plan would be to create a Draft File
for the article. This would consist of
My own collection of KLATU messages, in chronological
order -- that is, all the correspondence that i collected.
Between these, I would write a few comments, and mark people and
subjects that deserve explanations.
Then, I would collect everyone's comments on which items should be
deleted, further explained, etc. The full
correspondence is surely too long, but I think that the
experience has a large prototype social importance much greater
then Klatu itself. In fact, the best plan might
be to make most of the comments about the Network and about AI. For, the
Klatu itself is amply debunked in the messages.
Actually, I am pretty busy, and would be quite happy if one of you wnated to
supervise this new kind of literary project!
I think that for any of us to write something for that the New Yorker would
print would be a major task; I'm quite sure they wouldn't print our messages
However, Berton Roueche or Jeremy Bernstein who write for them on science
topics might well do a good piece. Vera has met Bernstein, but she won't
be back for months.
∂12-Sep-78 0924 REM Enhancements to my crunch program
To: "@CRUNCH.DIS[1,REM]" at SU-AI
Since search length in histogram was about 2 average, it didn't seem
worthwhile to optimize it from other than the current linear-search.
But after installing average-search-length accounting on context tree
I found that average case was rather bad. So I installed a little-ripple
where each time a non-head cell is referenced it is bubbled one step
closer to the head of the list at that level. Here are typical before
and after statistics for a small file (9k words before crunch):
CONTEXT-LENGTH SEARCH-PER-CONTEXT-LEVEL HISTOGRAM-SEARCH
BEFORE 2.44+ 14.1 5.86+
AFTER 2.44+ 10 5.86+
Note that the code wasn't affected, only the compute time. Hopefully
the extra computation used to bubble is more than offset by the
searching saved... anyway the chunk of code is easy to comment out
if desired.
One remaining thing to change in this program is to prune the context
tree whenever memory fills up.....
∂12-Sep-78 1430 JOSEPH at MIT-ML (Joseph Weizenbaum)
Date: 12 SEP 1978 1731-EDT
From: JOSEPH at MIT-ML (Joseph Weizenbaum)
To: JMC at SU-AI
The list of people Sen Kennedy is getting out of the Soviet Union
does NOT include Alexander Y. Lerner! We know that he and his wife are
very sick. I fear that seeing a window open but not being to get out
will be the final blow for them...I fear for the worst. I have asked
Jerry Wiesner to do what he can with Kennedy and whoever else he thinks
might be useful to get Lerner added to the list of people to whom exit
visas are to be issued. I don't know what if anything you can do from your
end . . . but do whatever it might be. I think there are two lives
immediately at stake here!
Best, Joseph Weizenbaum
∂12-Sep-78 1528 BALZER at USC-ISIB SUMMARY OF L.A. MEETING ON POSSIBLE A.I. INITIATIVE
Date: 12 SEP 1978 1454-PDT
From: BALZER at USC-ISIB
Subject: SUMMARY OF L.A. MEETING ON POSSIBLE A.I. INITIATIVE
To: AI Initiative Group:
!
1
The Los Angeles meeting was organized as a follow-on to the meeting
held at SRI on August 28 and 29 to discuss a possible joint IPTO-NSF
program in A.I. The purpose of this meeting was to open the discussions
to a wider segment of the A.I. community (those based in Southern
California) and to build upon the perceptions developed at the SRI
meeting.
Approximately 35 attendees represented the following organizations
with active A.I. programs: Jet Propulsion Laboratory, Rand Corporation,
U.C.L.A., U.C.S.D., USC-ECL, University of Texas, and ISI. None of the
funding agencies were represented.
The meeting was surprisingly productive for such a large group.
Although no consensus was achieved, the broad outlines of a program
which might achieve a consensus were discussed. Such a program would be
based upon one or more broadly defined "conceptual systems" (e.g.,
computer consultant, language understanding and generation, knowledge
acquisition and retrieval) to be selected by the following two criteria:
1. Application fallout - The ability to construct advanced versions
of such systems would lead to a broad spectrum of high impact
applications.
2. Technological advancement - The systems require further advances
in several core A.I. areas.
Thus, these "conceptual systems" would provide both the
applications potential (application pull) and the technological
advancement (research push) that are the driving forces for research.
There was a great deal of discussion concerning the positive and
negative lessons learned from the Speech effort and how they might be
applicable to a new initiative. In particular, discussions focused on
the types of goals that such an initiative should embrace. The two
major alternatives proposed were a single monolithic (though
multi-dimensional) goal as in the Speech effort, and a purilistic set of
goals, which could be separately pursued, focused on advancement of Core
A.I. research areas. Roughly, these two proposals represented the
applications-pull and the research-push viewpoints.
It was pointed out that a large portion of our science is devoted
to the "integrative" or "systems" aspects of a problem and that efforts
focusing on such issues should be prominent within any major new
initiative.
This goal structure issue was not resolved, but most attendees
seemed to feel comfortable with a program whose goal structure was a
blend of individual science oriented goals and (several different)
integrative systems oriented (but not necessarily complete and all
encompassing) goals. Progress would then be measured not via
achievement of any single goal, but rather through the sum total of
progress toward the many different stated goals.
!
2
It was also rather widely perceived that there would be a need for
the capability to alter the goal structure as progress was made, but no
reasonable mechanism was suggested.
One additional topic receiving major attention was "tool" support;
that is, not support of A.I. research directly, but rather of the tools
used by the research community, specifically hardware languages. It was
generally agreed that lately we have been consuming the capital of
earlier investment, that the lack of facilities was beginning to
seriously hamper further progress, and that scarce A.I. talent was being
diverted to support activities because of this situation. The role that
tool enhancement should play in a new initiative was discussed and,
although no consensus was reached, it was generally agreed that this was
a serious problem which requires some immediate attention.
BLACKBOARD NOTES
!
!
1
Suggested Topics and Issues
1. "Smart" sensors will have impact on AI
o push existing technology
o similar application for A.I. problems
2. "Labor Saving" developments-tools
o e.g. replacement for INTERLISP
o e.g. new hardware
3. Objective for A.I. Initiative: It must have a big impact
4. Applications goals needed to focus basic research
5. What do we "need"? (e.g., language generator, tools)
6. Even basic research program ought to have a system as a product
7. Maybe there isn't "a" key problem (because maybe AI isn't a field)
8. Build a "self-aware" system (that can reprogram itself, etc)
9. Information Retrieval (Question-Answering system)
10. An "A.I. expert" that could answer questions, teach, etc.
11. Robots
12. Language Understanding and Generation
13. Knowledge fusion
14. Knowledge acquiring system (e.g., can read books, "listen" to
people)
15. Man-Machine Cooperation system
16. Develop techniques for building expert systems
!
!
!
1
CORE A.I. ISSUES
(in addition to Strawman list)
1. Inference, deduction
2. Planner and Plan Generation, esp hierarchical
3. Knowledge acquisition
4. Knowledge accommodation (assimilation or integration)
5. Reasoning with uncertain and/or inexact information
6. Analysis and understanding of A.I. algorithms (A.I. theory)
7. Specialized architectures for specialized problem
representations (e.g., Fahlmans set intersection) ironic
computation
8. Analogical reasoning
!
-------
∂12-Sep-78 1735 MANN at USC-ISIB afterthoughts on AI initiative ISI meeting
Date: 12 SEP 1978 1723-PDT
From: MANN at USC-ISIB
Subject: afterthoughts on AI initiative ISI meeting
To: AI Initiative Group:
Afterthoughts from the AI initiative meeting at ISI
BASIC and APPLIED: The issues here were not all
represented at the meeting, and have not been thought
through and exposed. The matter of TIMING is crucial.
The "integrative" portion of the speech effort
predictably did not incorporate results from the more
fragmented "basic" efforts, because BOTH EFFORTS
STARTED AT THE SAME TIME. The integrative portion
could not rely on any particular output from the
"basic" efforts simply because of their timing, not
because of their merits.
I subscribe to the "seasonal" model of AI research.
Attempts to integrate the methods appropriate to a body
of data really do interfere with attempts to develop
basic knowledge of the same data. There is a season
for proliferation ofideas, and a season for combination
of ideas. The two kinds of efforts should be
separated, and the best kind of separation is in time,
not site, because of the transfer problems involved in
site separation.
Right now, it is the season to be BASIC. The
integrative kinds of efforts are important, necessary,
revealing and all that, but NOT TIMELY.
I would suggest a plan in which ALL effort to propose
specific integrative work is is postponed for 3 years,
then emphasized (but not to the exclusion of all else.)
POLITICAL TERMINOLOGY: There are dangerous traps in
two or three of the terms that were thrown around a
great deal at the meeting:
Information Retrieval: even the Science Information
people are trying to avoid this one because of the
difficulties it evokes. The people at the meeting were
using the term in a novel, metaphorical way which can
only lead to improper expectations and disappointment
with AI if continued.
Robot: Aside from the general public discontent with
nonfictional robots, there is no concensus among us as
to what counts as a robot. Before lunch, it was a
conceptual beastie, not necessarily having moving parts
or non-symbolic sensors. After lunch, it was a mobile
multisensor perception-based action machine. We should
drop the term.
Five years: Surely ends up in 1984.
Bill Mann
-------
∂13-Sep-78 0413 REM via SU-TIP Crunch program available to users on limited basis
To: "@CRUNCH.DIS[1,REM]" at SU-AI
I have cleaned up the user interface and am now releasing the
core-image for you daring lusers to play with on a limited basis.
Since the algorithm is subject to patent (maybe), I request you
not to decompile the program nor to us special priviledges to
unprotect or look at the source. Be aware that it may contain
bugs, and if a bug causes lossage of uncrunch-ability of a file
and the original file is deleted and not backed up then there is
no way whatsoever to recover. Also, if you use the optional
encrypt feature, and then forget your encryption keyword, nobody
in the world except the NSA and a few expert cryptologists will be
able to recover your file. Note also that the program is compute
expensive, and if you maximize effectiveness of crunch by specifying
a large core size (like 250000 words) it will also be core-expensive.
(Only run it when system load is very low, like less than 1.00 before
you start the program, less then 2.00 as it runs.) -- Within these
guidelines, if you have a large file (at least 5k, preferably at
least 30k) that you need to compress to smaller size, try my program
and tell me how you like it. RU IC2[1,REM] then type ? for help
the first time.
∂13-Sep-78 0959 PAT Doshita
Doshita arrives on the 17th. I have made reservations for him at the
Tiki Inn. He called from Vancouver to confirm the reservations and to
find out if you would be back.
∂13-Sep-78 1155 BALZER at USC-ISIB MESSAGE LIST FOR A.I. INITIATIVE GROUP
Date: 13 SEP 1978 1156-PDT
From: BALZER at USC-ISIB
Subject: MESSAGE LIST FOR A.I. INITIATIVE GROUP
To: AI Initiative Group:
INCLUDED BELOW IS A COMPOSITE LIST OF PEOPLE WHO WERE INVITED AND/OR ATTENDED
EITHER THE MEETING AT SRI OR ISI CONCERNING A POSSIBLE NEW A.I. INITIATIVE.
AI Initiative Group: Balzer@isib,Bledsoe@UTexas,Mann@isib,London@isib,Wilczynski@isib,Erman@isie,Norman@isi,Rumelhart@isi,Lingard@isib,Friedman@usc-ecl,Barnett@isib,Nevatia@usc-ecl,Rick@Rand-Unix,Moore@isib,Crocker@isib,Goldman@isie,Athompson@usc-ecl,Williams@isie,Chiu@isie,Price@usc-ecl,BMoore@sri-kl,
Nilsson@sri-kl,Hart@sri-kl,Hendrix@sri-kl,Sacerdoti@sri-kl,Tenenbaum@sri-kl,bobrow@xerox-parc,Fikes@xerox-parc,Goldstein@xerox-parc,Lenat@sumex-aim,Buchanan@sumex-aim,CCG@su-ai,Denicoff@usc-isi,Feigenbaum@sumex-aim,JMC@su-ai,TW@su-ai,Klahr@rand-unix,Faught@rand-unix,Don@rand-unix,Wesson@rand-unix,gorlin@rand-unix,Burge@rand-unix,stan@rand-unix
KEN COLBY AND JUDEA PEARL FROM UCLA ALSO ATTENDED BUT DON'T HAVE A NET ADDRESS
AND SHOULD BE COPIED VIA U.S. MAIL.
IN ADDITION THE FOLLOWING REPRESENTATIVES OF THE RELEVANT FUNDING AGENCIES ARE
INTERESTED IN THE PROGRESS OF THESE DISCUSSIONS. PLEASE USE YOUR OWN
JUDGEMENT IN DETERMINING WHICH MESSAGES WOULD BE OF INTEREST TO THEM.
RUSSELL@ISI
KAHN@ISI
CARLSON@ISI
DENICOFF@ISI
BARRETT(AT NSF, NO NET ADDRESS)
-------
∂13-Sep-78 1234 MRC
To: JMC, LES, REG
∂13-Sep-78 1227 BEN at MIT-AI (Benjamin Kuipers)
Date: 13 SEP 1978 1526-EDT
From: BEN at MIT-AI (Benjamin Kuipers)
To: mrc at SU-AI
Mark,
Can you convey to the appropriate people at LOTS a
more-or-less official request for any information available
on that system? The computer science department which is
being planned at Tufts, where I am now teaching, is trying
to figure out what kind of student computer facilities
should be available. A detailed description of LOTS would
be very useful to that end. Please have them send the
information to:
Prof. Benjamin Kuipers
Department of Mathematics
Tufts University
Medford, MA 02155
Thanks very much for the help. I'm still exploring about
support for dialnet, and will be back in touch when anything
develops.
Ben
∂13-Sep-78 1445 HEDRICK at RUTGERS (Response to message)
Date: 13 Sep 1978 1742-EDT
From: HEDRICK at RUTGERS
Subject: (Response to message)
To: JMC at SU-AI
cc: HEDRICK
In response to your message sent 13 Sep 1978 1359-PDT
You have had an account for some time. I guess somehow the message
didn't get through. I have changed the password to the one you requested.
Good luck!
-------
Your message got through. It's just that after I returned from my
foreign trip, I read and answered my net mail chronologically. Thanks
for the password change.
∂13-Sep-78 1452 HEDRICK at RUTGERS sloppy message reading
Date: 13 Sep 1978 1747-EDT
From: HEDRICK at RUTGERS
Subject: sloppy message reading
To: jmc at SU-AI
Well, I did it too: I replied to your first reply before reading
you second one. Hope it doesn't bother you that I changed your
password to the one you asked for. I have no easy way to undo
that, as I don't remember what it was before.
Regards,
Chuck Hedrick
-------
∂13-Sep-78 1822 DPB
To: JMC
CC: JB
Do you have a TA for CS206 yet? Who? I will ask for volunteers if you need one.
∂13-Sep-78 2333 MARG at PARC-MAXC (Response to message)
Date: 13 SEP 1978 2334-PDT
From: MARG at PARC-MAXC
Subject: (Response to message)
To: JMC at SU-AI
cc: MARG
In response to your message sent 13 Sep 1978 1344-PDT
Sure, lets go to louies sometime next week. I will actually be around til
the 29th. cu, marg
-------
∂14-Sep-78 0758 JMC robot
To: HPM
CC: JMC
An outfit called Android Amusements is trying to build a robot
for shows, etc. They seem honest, but I think they will find it more
difficult than they imagine - even the mechanical part. You might be
interested in a letter and a spec sheet.
∂14-Sep-78 0909 JMC
To: feigenbaum at SUMEX-AIM
CC: JMC at SU-AI
Is Ershov still coming in October? Is Michie coming?
∂14-Sep-78 0911 Feigenbaum at SUMEX-AIM (Response to message)
Date: 14 Sep 1978 0911-PDT
From: Feigenbaum at SUMEX-AIM
Subject: (Response to message)
To: JMC at SU-AI
In response to your message sent 14 Sep 1978 0909-PDT
Havn't heard a word re Ershov. Michie will be here on September 23
for approximately 3 mos. He will be teaching CS225, advising students,
being a colleqgue, and all those good things.
Ed
-------
∂14-Sep-78 1120 JRA
I asked mcgrae-hill to send a copy of my book to stoyan. i hope they did.
also hope they supplied you with one; if not i've got a few left.
after three
frustrating years at hp, trying to make them understand the difference
between batch-processing and interaction, i gave up. Sigh, it now
seems as if the yelling might have had an effect since they're toying
with a dec 10 purchase; there's no justice.
have you heard of any warm compatible spots for lisp hacking, interactive
programming, architecture, applied ai, ... in the area?
john
I have been abroad, but according to Patte, no copy of your book has
arrived, and I would very much like one. I know of no LISP jobs at
the moment, but I will keep my ears open.
∂14-Sep-78 1651 Tajnai at SUMEX-AIM Update on New Ph.D. Support
Date: 14 Sep 1978 1650-PDT
From: Tajnai at SUMEX-AIM
Subject: Update on New Ph.D. Support
To: faculty.list:
COMMENT ⊗ VALID 00003 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 To: CSD & CSL faculty and research supervisors
C00005 00003 NAME PROG. INTERESTS ADVISOR SUPPORT
C00011 ENDMK
C⊗;
!To: CSD & CSL faculty and research supervisors
From: Denny Brown
Subject: Assignment of new students to research projects (Pass 3)
Response to my previous memo(s) asking for support commitments has been
encouraging. On the next page I summarize the requests that I have received
and suggest one way to resolve most of the conflicts.
If you would like to see (review) correspondence on this issue, see
NEWPHD[PUB,DPB] or call me.
NSF Fellows can be supported quarter time for each of 3 or 4 quarters. Other
fellowships have their own rules.
We have also offered support to some of the entering Masters students. This
is intended to fill the TA gap which we experienced last year. Some of these
students might be appropriate for a short term research position. A list of
the Masters students who have been offered support appears below.
-Denny
NAME PROG. INTERESTS ADVISOR SUPPORT
_______________________________________________________________________________
Fay, Charles R. CSCE DCS PSL Baskett TA HALF ??
Gilbert, John C. MS AI PSL Brown TA HALF ??
Heartney, Francis J CSCE DCS OR Baskett TA HALF ??
Lin, Bertrand I-pen MS MTC AA FYao TA HALF ??
London, Robert MS CAI PSL Lenat TA HALF ??
Mackinlay, Jock D. MS AI PSL Brown TA HALF ??
Tang, Benjamin Ben- CSCE DCS McCluskey NSF Fellow ??
!NAME PROG. INTERESTS ADVISOR SUPPORT
___________________________________________________________________
Beetem, Anne France PhD MTC AI McCarthy NSF FELLOW (25% ?)
Boyce, James PhD AA MTC FYao NSF FELLOW (FFY 25% ?)
Chapiro, Daniel PhD AI CL MTC Winograd
Chapiro is an Argentinian student with a Mechanical Engineering
background. Most interested in Natural Language/AI.
So far nobody has picked him up. He has arrived, so is available
for interview. He would prefer to defer TAing to get better
accustomed to Stanford. Maybe TOB or HPP could pick him up.
Clanton, Charles IPHD AI Med Buchanan RA HPP HALF
Cooper, Greg IPHD AI Med Buchanan MSTP
Finger, Joseph PhD DCS PSL Baskett (FB, SSO, DCL, MJF)
Finger has been claimed by everybody in sight. See below for
suggestions.
Fuchs, David PhD AI PSL Lenat (GIO, TOB)
Gio asked for either Fuchs or Rowe. TOB asked for either
Fuchs or Lowry. See below.
Gropp, William PhD NA Symbolic Golub RA NA Half
Kenniston, Michael PhD PSL DCS Wiederhold (SSO, GIO, JLH)
SSO and/or JLH asked for Kenniston. Gio would like to put him
on the S1 grant.
Konolige, Kurt PhD AI, DB Lenat Honors Coop - SRI
Kunz, John IPHD AI Med Buchanan RA HALF HPP
Larsen, Howard PhD PSL MTC AA Knuth (DCL, DEK?)
DCL expressed direct interest. DEK would like to compete for
one of the new students to do TEX work. Larsen might appropriate.
Lowe, David PhD AI CL Winograd (TOB)
Claimed first, only, and unequivocally by TOB. An easy case?
Lowry, Michael PhD AI Lenat (TOB)
TOB asked for Lowry or Fuchs.
Mairson, Harry PhD AA MTC Knuth (DEK?,FFY?)
Another candidate for TEX, or for the Yao's project.
Paulson, Charles PhD DCS PSL Baskett NSF FELLOW
A quarter time of Paulson should be available to one of the systems
projects.
Rowe, Neil PhD AI PSL Lenat (GIO)
Gio asked for Rowe or Fuchs for KBMS.
Ryan, Doris PhD NA OR Golub Bell Labs
Schairer, Martin PhD AI CL Winograd German National Fellow
Talcott, Carolyn PhD AI Formal R McCarthy IBM Fellowship
Tong, Christopher PhD AA PSL FYao NSF FELLOW (FFY 25% ?)
Wolper, Pierre PhD MTC AI McCarthy CRD Belgian Graduate F
________________________________________________________________________________
I propose the following partial resolution.
1. Owicki, Hennessy, Flynn, Baskett fight over Kenniston, Paulson, Finger.
2. Gio takes both Rowe and Fuchs (who both have AI/Systems interest) and
staffs the KBMS and the S1 slots appropriately.
3. Luckham takes Larsen.
4. Knuth takes Mairson.
5. Binford gets Lowe and Lowry (Fuchs goes to Gio as above.) Also Binford
should consider Chapiro.
Deficiencies in this solution.
1. Chapiro is unassigned. (HPP, TOB, TW, else TA for fall.)
2. The systems group is understaffed. (Masters students to fill the gap?)
-------
∂15-Sep-78 0000 JMC*
Call Warren Spruil at 7-4195 about auditing LOTS use.
∂15-Sep-78 0040 IAZ Dialnet file transfers at SAIL
We (IAZ and MRC) have just succeeded in transferring a non-trivial file
(ie, the source to SAIL's DLNSER) over the Dialnet debug cable using the
official Dialnet file transfer protocol. The file came through with no
differences in content or size. We're going to put in a datarate meter
so we can determine what the effective baud rate for such transmissions
is at 1200 baud.
This is the file time that this has been done with the official Dialnet
FTP. CRUFT had transferred non-trivial files earlier, but IAZ had left
on vacation before the system bug which made this sort of thing lose had
been fixed.
∂15-Sep-78 0735 JRA
i will get a book for you today or tomorrow.
∂15-Sep-78 1358 RDR via AMES-TIP ComputerWorld network issue To: MRC at
SU-AI, JMC at SU-AI They're seeking contributions for an issue to be
published this November. Maybe a good chance to prosletyze DIALNET.
∂15-Sep-78 1655 REF
How would you like to talk to the Lab Colloq about your China
trip ("Computer Science in China", and any other tamusing stories.
Thanks.
bob
DPB suggested that I give a CSD colloquium, and I'll do that.
∂15-Sep-78 2149 DEK f(x,y,z)←if x≤y then y else f(f(x-1,y,z),f(y-1,z,x),f(z-1,x,y))
If I'm not mistaken, the time needed to evaluate this function, counting
one unit of time per comparison, is:
if x≤y then 1
else if z≥x then 3(x-y)+1
else if z=y then 3(x-y)+1
else if y=z+1 then 3(x-y)+8
else (3y-3z-1)(x-y)+1.
I think you're be mistaken. The numbers of function calls is given by
the recursive definition
ctak(x,y,z) ← if x ≤ y then 0 else 1 + ctak(tak(x-1,y,z),tak(y-1,z,x),
tak(z-1,x,y)) + ctak(x-1,y,z) + ctak(y-1,z,x) + ctak(z-1,x,y)
where tak (for Takeuchi) is the original function. It gives the following
values
x y z ctak(x,y,z)
3 0 3 3
3 0 4 14
3 0 5 52
3 0 6 168
3 0 7 586
which seems exponential, as Takeuchi asserted. LISP programs are in
the file TAK.LSP[E78,JMC].
∂16-Sep-78 0119 DPB colloquium
Ed suggested that you might want to give a colloquium on
Computing in China and the Far East. How about it? You could lead
off the year on 10/3. 10/10 is also open. -Denny
Yes, I'll be glad to. That precise title will be fine.
∂16-Sep-78 0502 MRC FTP data transfer rate
To: JMC, LES
IAZ tells me that FTP's data rate is 647 baud in an experiment he just made.
Hopefully that can be tuned up, but I sorta suspect that there isn't much
that can be done to make it faster.
∂16-Sep-78 1656 MARG at PARC-MAXC dinner
Date: 16 SEP 1978 1658-PDT
From: MARG at PARC-MAXC
Subject: dinner
To: jmc at SAIL
Would you like to join a group of people having Peking duck at Louie's
on monday nite. I dont know exact time, but I will find out and tell you...I
will be there...
-------
Yes, as it happens I was about to invite you to join Sarah and me at
Louie's Monday night, but eating with a larger group is also fine. Count
on me and Sarah.
∂16-Sep-78 1737 REM via SU-TIP Abstract for paper I plan to submit to COMM-ACM or similar. Feedback?
To: "@CRUNCH.DIS[1,REM]" at SU-AI, "@CRUPUB.DIS[1,REM]" at SU-AI
Nested Intervals: Radix-conversion, Mapping Stochastic-processes,
and Data-compression
by Robert Elton Maas, Stanford University
ABSTRACT: The classical mathematical concept of nested intervals,
whose infinite intersection is a real or p-adic number, is treated
in a dynamic (sequential) way. As new information arrives, an
interval is replaced by a subset of itself, approximating some number
more and more closely, converting a stream of digits into a virtual
real number; the inverse algorithm converts a number to its digit
representation. Putting the conceptual algorithms back-to-back as
output-driven coroutines yields an effective algorithm for radix
conversion. Interpreting the number as random yields an algorithm
for mapping random-digit sequences, and this immediately generalizes
to token-sequences resulting from arbitrary stochastic processes.
Given a stochastic (conditional-probability) model for naturally-occurring
text, the problem of optimal data-compression now reduces to the
straightforward application of this algorithm, mapping the statistics
of the input to the statistics that yield maximal entropy rate in
the output medium.
KEYWORDS: data-compression, stochastic-process, Markov-process,
radix-conversion, partial-bits, Huffman-code, nested-intervals,
Shannon-coding-theorem, left-context, conditional-probability.
∂16-Sep-78 2259 LLW DARPA and S-1 Systems
To: JMC
CC: TM, LLW
Tom and I spent an hour and a half with Bob Fossum of DARPA this past
Thursday, updating him and DARPA staff members of his selection on the S-1
Project (I know Bob from his pre-DARPA days, but hadn't discussed the S-1
Project with him since he became DARPA Director). I mentioned to him that
both you and Joel Moses had expressed interest to me in getting an S-1
system for your respective projects, in response to his remark that these
(S-1) processors seemed to be fine AI machines. I suggested that the
cost-effectiveness of the AI research work that these projects could do
for DARPA might be enhanced by the availability of more powerful and
highly cost-effective hardware bases for them; he agreed that this might
well be true, and was certainly worth looking into. (I didn't note that
SAIL was due to get the S-1 Mark I system during this next year, in any
event, as I didn't want to prejudice the possibility of DARPA's buying you
a Mark IIA, as well.)
Unfortunately, Bob is looking into a budget cut (relative to his planning
figures) for this upcoming fiscal year, so his buying even a few 200-300
K$ systems for his favorite AI projects is not presently a terribly bright
prospect. However, I thought this exchange with him was worth mentioning
to you, for whatever follow-up action you might want to take on it, in
your interactions with DARPA types (e.g., including as an topic in your
next proposal to DARPA, in the 'throwaway item' section).
Lowell
Judging from our budget cut, there is no reason to suppose that we are
among DARPA's favorites anyway.
∂16-Sep-78 2311 LLW
To: JMC
CC: LLW
∂16-Sep-78 2304 JMC
Judging from our budget cut, there is no reason to suppose that we are
among DARPA's favorites anyway.
[Dear John;
Don't feel uniquely leaned upon--DARPA in general and Bob Fossum in particular
are not having easy times of it these days. I don't know of any DARPA-sponsored
project that got a real budget increase in the FY78-FY79 transition. Their
seemingly chronic political problems are one reason I have steered clear of
any S-1 Project involvement with DARPA.
Lowell]
∂17-Sep-78 1222 DEK ctak
(small correction to ctak, it should begin "if x≤y then 1 else 1+..." since
the comparison is being made. Yes, I used the so-called "optimal evaluation"
scheme, which doesn't evaluate a parameter until it is used the first time.
My formulas may be off a little, but I doubt if the behavior is more than
quadratic.
∂17-Sep-78 1527 Allen Newell at CMU-10A (A310AN02) Re AIJ editor
Date: Sunday, 17 Sep 1978 1828-EDT
From: Allen Newell at CMU-10A (A310AN02)
Subject: Re AIJ editor
To: JMC at SU-AI (John McCarthy)
CC: feigenbaum @ sumex-aim, newell at CMU-10A
Message-ID: <17Sep78 182804 AN02@CMU-10A>
In-Reply-To: John McCarthy's message of 13 Sep 78 16:09
John and Ed: I think we need a conference call less than we need a
simple letter composed by one of you but signed by all three that
proposes this course of action to the entire editorial board, who I
guess are the parties in power here. I think there is nothing between
us all to negotiate. Ed's concern is to be sure a horizon gets put in
place. We should do that and the two year bit is fine. John's concern
is that we not bind ourselves up tight at that point, and it is far
enough in the future that I can't imagine Ed being unhappy to have
the 2 year thing in as a putative check and sunset point. This should
probably be a physical letter rather than an ephemeral net message.
So either of you has my signature for the signing of that specific
letter. AN
∂18-Sep-78 0120 REM via SU-TIP More IC2 results
To: "@CRUNCH.DIS[1,REM]" at SU-AI, "@CRUPUB.DIS[1,REM]" at SU-AI
Input file is my nutrition database FV.TXT which is 21456 words.
Crunched via old program CRU2 -- 6489 words
New IC2 program, defaults except NFORK=8 -- 4433 words
New IC2 program, same except MAX CONTEXT DEPTH is 8 instead of 5 -- 4054 words.
(One-pass self-adapting crunch using interval-refinement stochastic-mapping)
∂18-Sep-78 0739 Feigenbaum at SUMEX-AIM Re: Re AIJ editor
Date: 18 Sep 1978 0732-PDT
From: Feigenbaum at SUMEX-AIM
Subject: Re: Re AIJ editor
To: Allen Newell at CMU-10A, JMC at SU-AI
cc: newell at CMU-10A
In response to the message sent Sunday, 17 Sep 1978 1828-EDT from Allen Newell at CMU-10A (A310AN02)
Hi, Al!
Joh and I talked about the Meltzer/editor letter. I will write the draft,
and send it to you electronically for inspection.
Ed
-------
∂18-Sep-78 1225 REP via SRI-KL TAing 206
The possibility exists but I have to determine my commitments at
SRI first. I will get back to you as soon as I completely understand
my position. This should be sometime around Wednesday.
Rich Pattis
∂18-Sep-78 1322 CLT book
To: JMC
CC: CLT
i have an assembled copy on my desk. i will be in tonight and will
attempt to get a better copy from the xgp. i made the corrections
you noted in chapter V. is there anything else you want to look at
before we go to press? if not i will go see connie tomorrow am.
how many copies? connie and i thot perhaps 75.
∂18-Sep-78 1942 MRC Klatu
To: JMC at SU-AI, LES at SU-AI, Geoff at SRI-KA,
Lauren at UCLA-SECURITY
Gee, it's on TV now! Sheeit!
∂18-Sep-78 1946 MRC Klatu show
To: JMC, LES
It was on NBC just now on this show called "consumer byline" or something.
The turkey fell for it completely. I'm sure as hell glad I don't watch TV
often, with the utter garbage on it.
Anyway, you might want to chide NBC for their gullibility.
-- m
∂18-Sep-78 2241 IAZ FTP
To: LES, JMC
[from MRC and IAZ]
with minor FTP code improvements, we just got an effective baud rate
of 890 baud, and that was still with some trace messages.
∂19-Sep-78 0004 MINSKY at MIT-AI (Marvin Minsky) (Response to message)
Date: 19 SEP 1978 0302-EDT
From: MINSKY at MIT-AI (Marvin Minsky)
To: jmc at SU-AI
Hi. Here is my idea for a story.
John --
Here is the idea of an article. I don't think it would be too much
work. As you can see the idea is a counterpoint -- explaining the
ARPANET at one level, the QUASAR thing underneath. What do you think.
Near the end, you thank Reid for "the original version. I assume that
this was the Mark Fox report?
-- Marvin
.header notes on QUASAR Fake Robot, --- Minsky PAGE
****************************************************************
Date: 19
OCT 1977 1708-EDT
From: MINSKY at MIT-AI (Marvin Minsky)
To: MINSKY at MIT-AI, HORN at MIT-AI, PHW at MIT-AI, binford at SU-AI
To: winograd at PARC-MAXC, bobrow at PARC-MAXC, nilsson at SRI-KL
To: hart at SRI-KL, reddy at CMU-10A
You have all noticed the publicity about the Quasar Industries -- of
Rutherford, New Jersey -- robot.
At first I thought it was funny. Naturally, I assumed that the robot is
a fake, and that someone was merely attempting a hoax, presumably to
fool businessmen into buying stock and the like. Such activities are,
of course, no more our concern than anyone else's.
However, I find that lots of people see it, and few have healthy,
wholesome, unshakable doubts. And on further thought, I realized, why
should they? Only the AI community is in a position to challenge -
expose - educate - whatever.
So perhaps we must act -- for the first time? Perhaps this is not
merely a matter of meddling or criticising. If we are the only "group"
with the knowledge and authority to expose a fraud -- assuming that
this is the case -- then don't we have a moral obligation to say
something -- write an expose -- call a press conference -- or
something?
Such an action needs two steps:
Someone has to "investigate" and summarize what he found. Some others
have to back him up, sign a joint letter or whatever, so that there is
no one holding a legal bag alone.
Someone around here has a friend in journalism school, who might be
interested in doing this as an "investigative journalism" project, if
AI people will help her. Any other ideas?
-- Marvin
****************************************************************
Date: 15 Oct 77 AM-Robot Sam,250
Here is an AP dispatch obtained over ARPANET.
←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←←
a263 1529 15 Oct 77 AM-Robot Sam,250 PORTLAND, Ore. (AP) - Meet Sam,
a well-educated, male chauvinist robot. Sam was stationed on the
second floor of a downtown department store as a promotional gimmick
Friday using his 4,800-word vocabulary to insult customers and make
passes at young ladies. The $4,000 robot looks like a king-size aerosol
can with a smooth translucent bubble head and flexible arms. He is
guided by a computer hidden in the midsection of his 5-foot-2,
240-pound frame and movber wheels hidden under his polished metal
skirt. His taped responses are triggered by certain keywords uttered by
humans.
''What's your name?'' somebody asked him.
''My name is Sam Struggle Gear,'' the robot answered metallically.
''What's yours, dummy?''
Perhaps emboldened by a friendly comment and a kiss on his plastic
forehead from Portland State University student Alice Ericsson, Sam turned
and announced, ''I like girls.''
Waving his accordion arms, he moved unerringly on the lone woman in a
group.
''I like this one,'' he said. ''What's your name?''
Somewhat startled, the woman replied, ''Sandie Murphy.''
After a machine-whirring pause, Sam asked: ''You fool around?''
''No,'' Mrs. Murphy said firmly. ''I'm married.''
''That's all right with me,'' Sam said. ''Us robots like women with
experience.''
Somebody told Sam he was being a little rude. Sam stared at his critic
from his featureless face and growled: ''How would you like tire marks
on your belt buckle?''
1828pED 10-15
****************************************************************
Date: 19 OCT 1977 1442-PDT
From: WINOGRAD at PARC-MAXC
Subject: (Response to message)
To: MINSKY at MIT-AI, HORN at MIT-AI, PHW at MIT-AI,
To: binford at SU-AI, bobrow, nilsson at SRI-KL,
To: hart at SRI-KL, reddy at CMU-10A
In response to the message sent 19 OCT 1977 1708-EDT from MINSKY@MIT-AI
I talked to Paula Apsell (who is doing the NOVA AI film) about this
guy, and she knows a lot. Her associate Roy Gould) tried for weeks to
get an interview with the guy, since they wanted to use this robot on
their show. After fantastic shenanigans, he managed to get to the
factory, where it turned out (after all sorts of evasion, etc.) that
what they make is remote-controlled publicity devices, with no AI
content whatsoever. The robot in the clipping Marvin sent around was
probably one of them, (as opposed to a person in a robot suit, the
other alternative). The guy more or less confessed to Roy that what he
was doing was "show business". As far as I know, they have no plans to
expose it, since they are busy trying to do a film on legitimate AI.
However, I agree with Marvin that something should be done, and maybe
they can be useful. At the moment they're on the road filming, but
I'll mention it when I see them, and find out what they are interested
in doing. The whole thing is an interesting lesson in the gullibility
of the press, and the general public level of sophistication about AI ,
and if we could use the opportunity to get out bettter information, we
should. --terry
****************************************************************
Date: 20 OCT 1977 0016-EDT
From: MINSKY at MIT-AI (Marvin Minsky)
To: BOBROW at MIT-AI, BOBROW`PARC-MAXC at MIT-AI, HORN at MIT-AI
To: PHW at MIT-AI, MINSKY at MIT-AI, tob at SU-AI, tw at PARC-MAXC
To: hart at SRI-KL, reddy at CMU-10A
Thanks to TW for useful reply about fake robots. I will file my letter
and replies in AI:MINSKY;ROBOT > file and we can decide what to do in a
week or two. Terry thinks it might be a useful public opportunity to do
some good. Who else should be on mailing list? --Minsky
****************************************************************
Date: 19 Oct 1977 2256-PDT
From: JMC at SU-AI (John McCarthy)
Subject: Quasar fake robot
To: FAKE.LIS[ESS,JMC]:;
I agree that it is the AI community's responsibility to expose the fake
robot. Who if not us? If possible, we should also criticize the
gullibility of the press. It is either that or an irresponsible liking
for seeming marvels - like the magician's "prediction" of the headlines
about the New York blackout riots a week in advance. The magician is
blameless; switching envelopes after the prediction is a common stage
trick.
It would be best to have an independent investigation. Since the
Quasar outfit is in New Jersey, maybe some Bell Labs people or Rutgers
people could do the work, but it is very likely that they would be
stone- walled unless they pretended to be customers who wanted it for a
Department Store publicity stunt or something like that. Maybe they
could get the co- operation of Macy's or some other respectable outfit.
Chuck Rieger and/or Saul Amarel at RUTGERS-10 might organize it.
After the investigation there should be a press conference held in New
York. Respectable AI people should be present. Perhaps a foundation
would underwrite the necessary travel. Whoever did the actual
investigation should be featured along with the professors. If not,
short range travel should suffice.
The following points might be made:
1. Quasar's claim (what they claimed should be documented from the news
stories) would, if verified, constitute a major advance over what the
worldwide AI community believes to be the state of the art.
2. Since they never came to scientific meetings or published and we
never heard of them we were skeptical.
3. They turned out to be remotely controlled. A robot depending on a
man in a control room is hardly useful for household chores.
4. The newspapers, wire services, and TV broadcasters have been remiss
in their duty to check marvels. They can ask well-known scientists in
the relevant field, and if they get a skeptical answer they can
investigate further.
****************************************************************
Date: 20 Oct 1977 0923-PDT
From: Hart at SRI-KL
Subject: (Response to message)
To: MINSKY at MIT-AI, BOBROW at MIT-AI,
To: BOBROW`PARC-MAXC at MIT-AI, HORN at MIT-AI, PHW at MIT-AI,
To: tob at SU-AI, tw at PARC-MAXC, reddy at CMU-10A
cc: HART, JMC at SU-AI
In response to the message sent 20 OCT 1977 0016-EDT from MINSKY@MIT-AI
I agree that a little public educuation would be a good thing. It
would, however, be worth paying attention to the pitfalls of appearing
overly sanctimonious or self-serving, particularly when tarring the
media. For example, I was contacted by The Wall Street Journal shortly
after the wire service stories appeared, and I expressed considerable
skepticism about even a small part of the claims being true. To my
knowledge, WSJ did not give any space to the issue.
Mail list suggestion: Keith Uncapher, Josh Lederberg, Ed Feigenbaum
****************************************************************
Date: 20 Oct 1977 1411-PDT
From: Nilsson at SRI-KL
Subject: Fake
To: Minsky at MIT-AI
cc: Nilsson
Marvin, I agree with the need to expose fakes and agree generally with
Winograd's and McCarthy's suggestions. I don't have time now to plaly an
active role but will support a reasonable campaign.
-Nils Nilsson
****************************************************************
Date: 25 Oct 1977 1351-EDT
From: Raj Reddy at CMU-10B
Subject: quasar robot
To: minsky@CMU-10B
CC: raj reddy@CMU-10B
When Newsweek called me a few months back to talk about the Quasar
robot, I said that it had to be fake, although it might be possible to
build an Eliza-like system which pretended to understand speech.
Yesterday the Robot was on display in Pittsburgh. The following
discription by Mark Fox shows that it is indeed a fake.
****************************************************************
The Carnegie-Mellon University Artificial Intelligence Lab meets `The
Ultimate Home Appliance'
(Reported by Mark Fox)
On October 24, 1977, a well-known department store in the heart of
Pittsburgh advertised the appearance of a `domestic robot' named Sam
Strugglegear. Although this robot is not yet offered for sale, its
inventor, Anthony Reichelt of Quasar Industries in New Jersey, claims
that its powers include speech recognition with a 4800-word
vocabulary, sonar-navigated steering, and the ability to do household
chores such as vacuuming, serving of drinks, and babysitting. This
highly-publicized `robot' has been described in Newsweek, Parade, and
other national magazines.
Knowing of CMU's pioneering work in Artificial Intelligence,
particularly in the field of speech recognition, various friends have
called CMU to ask how this robot might be so much better at speech
recognition than our talented and dedicated research team.
Rising to the challenge, four courageous members of our department
went downtown to investigate. They found a frightening sight: in the
men's department, among the three-piece suits, was a 5'2'' image of
an aerosol can on wheels, talking animatedly to the crowd. The robot
seemed able to converse on any subject, to recognize the physical
features of the customers, and to move freely (though slowly) in any
direction. The crowd was quite charmed by the talented machine.
The robot moved on wheels or tracks; we couldn't tell without
upending it. It could move about three inches per second (about one
tenth the normal walking speed of an adult), and always came to a
complete stop before changing direction. We saw the left arm move up
and down without ever grasping or holding anything, but never saw the
right arm move at all. One patron asked to see the robot vacuum a
carpet, but was brushed off with the reply that its batteries were
running low. A more detailed description of the mechanics can be
found at the end of this writeup.
The CMU team set out to investigate the robot's sensory mechanisms.
Pushing and blocking it had no effect; it didn't seem able to tell
that an object was blocking its path. Covering the faceplate didn't
seem to change its behavior either. Since the robot seemed able to
navigate around the room without hitting anything, we found it quite
curious that it had no detectable sensory reactions.
Feeling more dubious, we began looking around the room for evidence
of remote control. Lo and behold, about ten feet from the robot,
standing in the crowd, we spotted a man in a blue suit with his hand
held contemplatively to his mouth like Aristotle contemplating the
bust of Homer in the famous Rembrandt painting. After watching for a
while, we noticed that whenever the robot was talking, the man in the
blue suit could be seen muttering into his hand. Further seeing that
this man had a wire dangling suspiciously from his waist to his shoe,
one of the CMU group screwed up his courage and approached this
stranger. "Do many people figure out what you are doing?", we asked.
"No," he said, "they are usually too busy watching the robot to
notice me."
We then asked him what were the robot's speech and vision abilities,
to which he replied that the machine can see about ten inches, dimly,
and that its speech-understanding ability was about 200 words of
unconnected speech in a quiet environment.
We didn't really believe his statement of the robot's abilities, and
in the light of our discoveries of the robot's poor perceptive
skills, we were convinced that there must be yet another remote
control handling the motion. Time was running out; they needed to
move the machine to a suburban store for an evening demonstration.
We returned to CMU feeling unsatisfied.
When we gave our report to the rest of the lab back at CMU, a second
group of eight immediately set out to the suburban store, determined
to find the source of the robot's control. They found a
furtive-looking and rather disagreeable person loitering in the back
of the room. He was carrying an airline flight bag, with his hand
stuck down inside the bag. We asked him his business, to which he
replied that he was a truck driver. He became extremely agitated
when we asked him what was in the bag, asking if we were police. We
dispatched a person to watch him, in an attempt to find correlations
between movements of his hand and movements of the robot, whereupon
the bag man got very excited and called for store officials to come
get us away from him. We never did get to see in the bag. However,
we did see the man with the microphone say to a store official, "Tell
him we want to take it for a walk," whereupon the store official
wandered over to the `bag man' and whispered something to
him.
The robot is thus, as we suspected, a complete fake. It uses
mechanisms no more sophisticated than remote control and two-way
radio, and is good for not much more than sales promotion. One and a
half centuries ago a fake chess-playing robot , Baron Von Kempeler's
chess playing maching, was built (it had a man concealed inside);
that one had a lot more class.
A More Detailed Mechanical Description (provided by Bill Brantley).
The arms had a joint at the elbow but we never saw it move (probably
no actuator in this model). We saw both arms move being rotated at
the shoulder. The drive system had two large (8-12" dia.) drive
wheels and several small stabilizing wheels -- probably just like the
MIT turtle.
The hands were like clam-shells in design. There was a rod at the
wrist which I assume is for opening and closing the hands. However,
on the model we saw the hands were glued together (literally). So
that, even if there were an actuator inside the hands would be
inoperative.
The actuators for the arms were electric motors attached to the arm
by gears (not belts). When one physically stopped the arm from
moving the motor would stop -- some sort of primitive feedback.
ADDENDUM
At the time of writing, we have been trying to contact the department
store officials as to their knowledge concerning the robot's remote
control. We will also be contacting Quasar and will keep you
informed.
****************************************************************
Date: 4 NOV 1977
From: DHT@MIT-AI (David Taenzer)
11/04/77 08:41:12 Hi. A friend of mine here suggested that the Quasar
people ma be doing these demos simply for the money the department
stores are paying them for attracting customers. It might be
worthwhile finding out how much the stores have to pay for a visit from
the "robot"... ---Dave
Hello. I'm Cathy Shaw, a journalism student at Columbia, putting
together an article on the Quasar robots. I have interviewed Marvin
Minsky on the subject and, if possible, would like to get some other
scientists' comments on the subject. I have a promotional paper from
Quasar detailing the abilities of "robots by Quasar." This paper is on
the file DHT;QUASAR > on MIT-AI. You are cordially invited to respond
to any part--or all--of this memo. I will assume your comments are
quotable. Much thanks for any help you can give me in preparing this
article. Please send replies to DHT@MIT-AI. Thank you. Cathy Shaw
****************************************************************
Date: 14 Nov 1977 1518-PST
From: JMC at SU-AI (John McCarthy)
To: FAKE.LIS[F77,JMC]:;
Sandra Blakeslee who works for the Stanford news service issued a press
release based on what I told her about the Quasar fake robot.
Consequently, various news stories will be based on this release. My
statements to her rely heavily on the CMU investigation and have no
independent source of information except the statement by Reichelt in
response to a San Jose Mercury reporter's telephoned question.
Reichelt said it didn't have to be able to tell a cup from a saucer,
because it didn't take anything off the tray. This blows Reichelt's
claim that his robot can serve a meal and shows that he isn't even
careful to make his claims consistent. It would be nice to have a
careful investigation of Quasar leading to a full report, but since
everyone seems too busy for that, it seemed to me that saying what I
did was better than having no public response at all from the AI
community.
****************************************************************
From: JMC at SU-AI (John McCarthy)
To: reid at CMU-10A
CC: minsky at MIT-AI
Thanks for the copy of your original version. I'll keep it out of the
newspapers. It didn't lose a lot by being toned down, but I think
Minsky is right that it is the responsibility of scientists in a field
to expose frauds in their area. This may require incurring some risk
of being sued, since these days, an over population of lawyers
encourages groundless suits.
****************************************************************
Date: 21 NOV 1977 2135-EST
From: KLH at MIT-AI (Ken Harrenstien)
Subject: Fake robot: a call for help
To: BRIAN.REID at CMU-10A
CC: MINSKY at MIT-AI, DHT at MIT-AI, GLS at MIT-AI
Date: 21 Nov 1977 1950-EST
Sender: BRIAN.REID at CMU-10A
Subject: Fake robot: a call for help
From: BRIAN REID(C410BR10) at CMU-10A
To: Header-People at MIT-MC, MSGRP.DST
A reporter from Business Week magazine is going to Quasar tomorrow
morning (Tuesday 22 Nov) at 10:00 a.m. I want him to take with him a
person who would be able to expose the thing for what it is.
Are any of you folks in New York City? Would any of you be willing to
go along with this reporter tomorrow morning? If so, please let me know
IMMEDIATELY, and I will connect you up.
---- Brian Reid
Indeed. You should get in touch (call long distnace if necessary)
Marvin Minsky or Dave Taenzer, so as to get in touch with Cathy, who is
down in the NYC area intersted in doing this expose... Dave is at
(315) 399-3922. Minsky at (617)734-3363 but that is home and only
advised if really urgent. Don't have Cathy's # unfortunately. Good
luck, keep me posted.
****************************************************************
Date: 22 NOV 1977 0208-EST
From: HENRY at MIT-AI (Henry Lieberman)
To: MINSKY at MIT-AI, BAK at MIT-AI, KLH at MIT-AI, HENRY at MIT-AI
To: jmc at SU-AI, rsmith at RUTGERS-10, lefaibre at RUTGERS-10
To: reid at CMU-10A
The Quasar promotional literature is on DHT;QUASAR >.
****************************************************************
Date: 22 Nov 1977 0209-EST
From: Brian Reid at CMU-10A
Subject: the state of the Business Week visit
To: McCarthy at SU-AI, RSmith at RUTGERS-10, LeFaivre at RUTGERS-10,
To: Henry at MIT-AI, Minsky at MIT-AI
CC: Reid@CMU-10A
I think we seem to have scared the reporter away. When I talked to him
right before I sent out that hue and cry, he seemed interested in
having a technical person along. Now he wants to go it alone. I gave
him a lot of tips for what kind of things to look for in spotting the
hoax. I don't think it would hurt a thing if a person or two just
showed up anyhow. But we shall see. The reporter has called me every
day this week, and I assume that I will hear from him again tomorrow).
I'll keep you posted.
Brian
****************************************************************
Date: 22 Nov 1977 0632-PST
From: JMC at SU-AI (John McCarthy)
To: FAKE.LIS[F77,JMC]:;
HOW QUASAR CAN MAKE MONEY
That the Quasar robot was a fraud has always seemed clear, but it
wasn't clear how they could make money from it.
The file DHT;QUASAR > at MIT-AI is a copy of a promotional document
that Quasar sends in response to inquiries. On the basis of the
following excerpt from the document, here is a conjecture on how they
can swindle people.
"The DA-2 model is available only by reserved production number order
to qualified individuals and corporations. The specialized nature of
these Robots require a compatability program analysis with prospective
owners before delivery, to establish gender, personality, intelligence
level re- quired and specialized internal sub-systems.
"Inquiries are now being received for sponsor application forms to
determine individual qualificationa and requirements."
This provides an excuse for a representative to visit selected
inquirors to check their suitability for swindling.
The ideal swindle, as in most schemes, would be an elderly person who
believe what he or she reads in the newspapers and who has money and no
nosey or suspicious people living with him or her. Since many
newspaper and magazine stories have been printed that accept the claims
without qualification, the Quasar salesman can carry along a convincing
set of clippings. Once it has been determined that the person is
suitable, the salesman can ask for a deposit for the standard model and
as many options as the swindlee's bank account will stand. An
opportunity to invest can also be offered, deposits on dealerships can
be taken, etc. Nothing has been promised for two years, and excuses
for further delay in accomplishing such ambitious goals can easily be
found. People who seem too suspicious, not rich enough, or who seem to
have attorneys, suspicious relatives or friends can be asked to wait
for further compatibility testing.
The newspaper publicity Quasar has received since July can already have
led to a number of swindles. The recent unfavorable publicity will be
missed by many of the people who read the original favorable people.
Although it is possible that Reichelt is mainly deluding himself, the
probability that he has already swindled some people is such that it is
now necessary to inform the Federal Trade Commission, the part of the
Justice Department that handles mail fraud, and the District Attorney
of the county
where Quasar is located.
********************************************************
This is copied from the Quasar Brochure * * * * * * * *
********************************************************
THE DOMESTIC ANDROID
Quasar's Domestic Android represents the historic advance to the
science of Mechanical Humanoids. For the first time there is available
a highly reliable, multiplex, functional domestic servant. Capable of
receiving, memorizing, and expediting basic household tasks. Operating
within the confines of a residential home or apartment, the Domestic
Android will expedite an ever increasing series of functions. The DA-2
model series is designed to accept new and improved sub-systems as they
are developed without replacing the entire robot. Quasar will not
incorporate any major body style changes for at least 25 years, therein
allowing maximum basic value increase potential.
∂19-Sep-78 0903 MINSKY at MIT-AI (Marvin Minsky)
Date: 19 SEP 1978 1202-EDT
From: MINSKY at MIT-AI (Marvin Minsky)
To: JMC at SU-AI
Sorry. I think I sent you the wrong file --
just the old quasar file.
The example of proposed story is in
minsky; newy >
∂20-Sep-78 1139 DPB Colloq.
∂16-Sep-78 1730 JMC
Yes, I'll be glad to. That precise title will be fine.
Don Knuth expressed strong preference for the 10/3 date. I'll put you down
for the 10/10 lecture. If that doesn't work for you, we (you, Don, & I)
will have to negotiate. For now:
10/10/78 - John McCarthy - Computing in China and the Far East
-Denny
∂20-Sep-78 1602 100 : pat flight to detroit
your flights are:
Sept 24 AA604 SF-DET depart 11am arrive 605pm
Sept 26 UA135 DET-SF depart 5:30pm arrive 8:55pm
∂21-Sep-78 1014 Tajnai at SUMEX-AIM Queenie/LOTS
Date: 21 Sep 1978 1014-PDT
From: Tajnai at SUMEX-AIM
Subject: Queenie/LOTS
To: faculty.list:
Queenette Baur of LOTS is scheduled for surgery 10/2. Dennis Brown
asked me to notify the faculty in case they need anything from LOTS.
Please try to handle it next week.
Queenie will be at the Los Gatos Community Hospital.
Carolyn
p.s. She will be out for a month (gall bladder).
-------
∂21-Sep-78 1514 REP
∂21-Sep-78 0819 CJS
To: REP
CC: CLT
Sorry, Rich, the LISP manual is going to the printer's today, and
they are SWAMPED with work. It will be at LEAST 2 weeks before I can expect
it back from the printer.
Connie
------------------------------------------------------
John:
This is the latest word on the manuals. Looks like they wont be in until
the beginning of October. The MACLISP manuals have been ordered and I
check if they are in at the bookstore.
Rich
∂22-Sep-78 1611 JBR
To: S1 at SU-AI
The S1 library now contains a copy of the Bell System Technical Journal
issue on the UNIX time sharing system.
∂23-Sep-78 1006 REM via AMES-TIP More IC2 results...
To: "@CRUNCH.DIS[1,REM]" at SU-AI, "@CRUPUB.DIS[1,REM]" at SU-AI
FILE CORMAX NLCB NFORK AVG.CXT AVGCSCH HISTSCH
ORIGINAL SIZE CRUNCHED SIZE CR
*.SAI[*,*]part 210000 9 10 2.28+ 28+ 4.88+
156190 63394 .406
ROMAN.SAI[FNT,DEK] " 10 8 5.13+ 31+ 2.87+
41600 8428 .203-
<SAME>(2:99)/N " 10 8 2.92+ 26+ 3.75+
25357 7451 .294-
TEXSEM.SAI[1,GLS] " 5 8 2.65+ 25+ 4.59+
(raw 36736) .../N 35639 12928 .363-
EDSN.SAI[MIX,MUZ] " 5 6 2.84+ 24+ 3.83+
(raw 39936) .../N 32328 10310 .319-
SAIL.JFR[AIM,DOC] " 5 10 2.84+ 23+ 3.99+
(raw 139648) .../N 115835 40702 .351+
Note that since E-format directory and padding of nulls after each page
causes considerable bloat in some files. C.R. is quite good (much better
than 3:1 in some cases, ROMAN.SAI above) when this bloat is counted as
part of the reduction obtained, but when file is stripped of this stuff
before crunching and it isn't counted, the C.R. increases to "about 3:1"
for large files (40k or more).
∂23-Sep-78 1439 Moore at SRI-KL (J Moore) A Mechanical Proof of a Metatheoretic Result
Date: 23 Sep 1978 1441-PDT
From: Moore at SRI-KL (J Moore)
Subject: A Mechanical Proof of a Metatheoretic Result
To: csl:
One perennial problem faced by the builders of mechanical
theorem provers is that of domain specific knowledge. One
cannot anticipate all of the special purpose routines one
might need to prove theorems once the theorem prover has
left the "factory." One could let the users of the theorem
prover build in new routines -- but there is much danger
there since whenever anyone performs "brain surgery" on a
theorem prover, its future behaviour is suspect until proven
innocent (i.e., the factory warranty is invalid if the
product appears to have been tampered with by unauthorized
repairmen). For several years the following neat idea has
been circulating in some theorem proving circles: embed the
semantics of ones theory in the theory and use the factory
version of the theorem prover to prove the correctness of
the modification. Richard Weyhrauch, Frank Brown (of
Edinburgh), and we have all independently toyed with the
idea over the years. Our suggested approach is to add a
feature to the system that efficiently compiles and runs
functions that manipulate terms in a way that has been proved
semantically correct. Justifying such a metatheoretic act
is nontrivial; but actually proving the correctness of
particular metatheorems is hard too: one has to have a
mastery of the elements of symbolic manipulation.
To demonstrate the practicality of the idea, we recently had
our theorem prover carry out the proof of the correctness of
a simple PLUS cancellation routine. The function CANCEL was
defined so that if given a list structure of the form
(EQUAL (PLUS...) (PLUS ...)) (that is, something whose CAR is
the atom "EQUAL", CADAR is "PLUS", etc) it returned a new
list structure with the PLUS nests right associated and common
addends deleted from both sides. Then we had the system
prove:
(EQUAL (EVAL (CANCEL X) A) (EVAL X A))
which establishes that if one interprets X as a form in our
theory CANCEL does not change its value. In our case, EVAL is an
undefined function about which three axioms are assumed:
(EVAL (LIST "PLUS" X Y) A) = (PLUS (EVAL X A) (EVAL Y A))
(EVAL (LIST "EQUAL" X Y) A) = (EQUAL (EVAL X A) (EVAL Y A))
(EVAL 0 A) = 0
One doesn't need any more information about EVAL to do this
particular proof. Once the correctness of CANCEL has been
proved, it is permitted to compile CANCEL into an INTERLISP
function and actually run it to simplify terms in later proofs.
(We have not implemented that aspect yet -- we're trying
to see how hard the theorems you have to prove are).
The proof required the development of a good deal of "bag"
theory, such as that the intersection of two bags is a
subbag of both, and that if x is a subbag of y then the
value of the plus nest constructed for x is less than or
equal to that for y (since we have non-negative arithmetic
only). The proofs are constructed without using any axioms
other than the Peano like axioms for numbers, lists, and
literal atoms and the 3 axioms about EVAL mentioned above.
We hope to write up a commentary on the proofs soon, but
thought that the practicality of producing the mechanical
proofs was of interest in itself.
J and Bob
-------
∂23-Sep-78 1604 TOB collaboration
To: JMC, LES
In addition to the Leningrad contact, Marco
Somalvico at Milano and Jean Paul Barthe have
suggested small collaborative exchange proposals
which would cover a few visits. Are there any
formalities to clear these arrangements, beyond
the proposals themselves. I would like to make
the arrangements. What do you think about it?
Tom
∂23-Sep-78 2146 CARL at MIT-AI (Carl Hewitt)
Date: 24 SEP 1978 0045-EDT
From: CARL at MIT-AI (Carl Hewitt)
To: jmc at SU-AI
CC: CARL at MIT-AI
John,
I will be in Palo Alto from Tuesday Oct. 3 thru Monday Oct. 9
and would like to talk to you about "procedural semantics". We have developed
a mathematical model of actor systems that I believe answers some of your
criticisms to the effect that "procedural semantics" was not well defined
by comparison with the classical model theoretic semantics of first order logic.
Also I would appreciate being able
to give a seminar [see abstract below] to as wide an
audience as possible (not just A.I. types).
Sincerely,
Carl
Some Controversial Conjectures in the
Semantics of Concurrent Systems
Carl Hewitt
M.I.T.
By a concurrent system, I mean a network of digital mechanisms connected
asynchronously, where each mechanism has some processing power and a finite but
possibly unbounded amount of storage. Recently three models of concurrent
computation (Petri Nets, Communicating Sequential Processes [Hoare, Milner], and
Actors) have been given mathematical foundations. On the basis of these
foundations, I will draw some tentative and controversial conclusions.
The hypothesis of bounded nondeterminism asserts that for any concurrent
system which is guaranteed to terminate, there is a bound on the number of
states in which it can terminate. The hypothesis has played a prominent role in
some recent models of computation. An important design decision in
Communicating Sequential Processes [Hoare: CACM Sept. 1978] has been justified
on the grounds of excluding the possibility of unbounded nondeterminism. One
consequence of the decision is that communicating sequential processes cannot be
used to implement some useful public information processing utilities that
guarantee to serve customers on a network. (In defense of communicating
sequential processes, it can be argued that they were never designed to
implement such network utilities.)
The actor model provides a mathematical semantics for concurrent systems
in which the possibility of unbounded nondeterminism is not excluded. We have
found it a useful tool in describing the behavior of concurrent systems such as
network utilities which make strong guarantees of service.
I will be glad to see you, and I have mailed your message to Susan
Owicki who is running a seminar on concurrent systems.
∂23-Sep-78 2219 MRM shoes
foo, i seem to have left my shoes and sox in your car. Could you leave them ay the lab somewhere where I can find them in te next few days? thanks, margaret.
I fear your shoes and sox must have gone to Davis and will be back
this weekend. If you need to collect them on your way to mtns, I
can furnish Sarah's co-ordinates. Davis is on one way to
Sierras. Apart from that, would you like to eat sushi (Japanese)
this Thursday or next Tuesday?
∂24-Sep-78 1016 REM
To: "@CRUNCH.DIS[1,REM]" at SU-AI, "@CRUPUB.DIS[1,REM]" at SU-AI
∂23-Sep-78 2246 RWG on TTY73 (at TV-140) 2246
your abbreviations confound me. NLCB = National Labor Creation
Board? NFORK = Riflemanville? HISTSCH = Histology School? AVG.CXT must
mean average context, but what is that? And AVGCSCH is easy--average
hyperbolic cosecant; but why?
[REM - NLCB = Number (length) of LCB = Left-Context Buffer,
NFORK = Number of instances needed before a context-node forks off
to have a histogram all its own,
Other parameters are average context depth (number deep in left
context), average context search (number of alternatives searched,
i.e. context depth times search-time per level), and histogram
search length (after context found, how far down next-tokens until
correct next-token found). Context depth depends on NLCB and file
statistics and also on both NFORK and CORMAX, typically 0 to 5.
Context search typically 20-30. Histogram search depends on entropy
of data and hence is related to effectiveness of code, short search
means first possiblity is very likely means great compresson, thus
the faster the program runs the better it is doing.]
∂24-Sep-78 1557 RPG Downtime
To: LES at SU-AI, JBR at SU-AI, JMC at SU-AI, ME at SU-AI,
BH at SU-AI, MRC at SU-AI
I've reserved the machine for TED (I was helping him build today)
who will be in to debug the memory. Since it is against policy to log in
as someone else, I've reserved it myself.
-rpg-
∂25-Sep-78 0947 SSO Modal logic/concurrent programming seminar
To: "@MODAL.DIS[1,SSO]" at SU-AI
COMMENT ⊗ VALID 00002 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 EE385E - Concurrent Programming Seminar
C00003 ENDMK
C⊗;
! EE385E (CS319E) - Concurrent Programming Seminar
Autumn Quarter 1978/79
This quarter we will emphasize parallel program verification,
expecially liveness properties. Modal and temporal logic will be
studied as tools for that problem.
Prerequisite: EE386 (CS246)
Instructor: S. Owicki
Time: Thursday, 4:00
ORGANISATIONAL MEETING 9/30/77
Room: ERL 237
Credits: 1-4
∂25-Sep-78 0953 DAVID at USC-ISIB Information request
Date: 25 SEP 1978 0954-PDT
From: DAVID at USC-ISIB
Subject: Information request
To: MCCARTHY at SU-AI
cc: DAVID at ISIB
Mr. McCarthy,
As the technical librarian at Information Sciences Institute, I have
recently received a request from a member of our research staff whose current
research appears to be headed in a direction of some of yours. Would it be
possible for you to send me a list of articles which you have authored in the
last 5-10 years. I wouldn't need the articles themselves, only a list of
them.
Thank you for any assistance you can provide,
David Van de Streek (David@ISIB)
Technical Library
-------
∂25-Sep-78 1215 ARK S-1 Meeting
To: S1 at SU-AI
There will be an S-1 Meeting on Wednesday, 27 Sept. at 3 p.m. in the Small
Conference Room at SAIL. Randoms invited.
Arthur
∂25-Sep-78 1345 PN wednesday meeting
To: S1 at SU-AI
Neither Gio nor I can come at the proposed time on Wednesday. I would
like to be there for the whole meeting. I think the best way to get
this together is for everyone to send ARK their schedule ASAP. This is
for this meeting only; at the meeting we can decide on a permanent time.
∂25-Sep-78 1513 Tajnai at SUMEX-AIM FACULTY MEETING, TUESDAY, SEPTEMBER 26, 2:30 p.m., POLYA 204
Date: 25 Sep 1978 1512-PDT
From: Tajnai at SUMEX-AIM
Subject: FACULTY MEETING, TUESDAY, SEPTEMBER 26, 2:30 p.m., POLYA 204
To: faculty.list:
cc: McGoveran, BS at SAIL
PLEASE NOTE ON YOUR CALENDARS --
FACULTY MEETING, TUESDAY, 26 SEPTEMBER, 2:30 p.m., Polya 204
Please send agenda items to McGoveran@SAIL
-------
∂25-Sep-78 1601 MRC Dialnet at LOTS
To: JMC, LES
After some consultation with the LOTS staff, it appears that I will
not be able to get enough downtime on LOTS until Christmas time at
the earliest. Until that time, I may be able to get one or two
sessions at the most, which wouldn't be enough even if I was
experienced in Tops-20 hacking.
I think we should talk with the business school people about letting
me take their machine at night or the wee hours of the morning. Or
rather, JMC should talk to the business school people; Ralph and JQ
both indicated that it would be better that way since they would have
heard of JMC. As for how much downtime I want, I can make good use
of as much downtime they are willing to give me; in other words, if
they will let me have six hours a night for the quarter, I could use
it (although with that much downtime I should be able to have it done
long before that and spend the rest of the time fine-tuning it).
We will, of course, have to offer them a VADIC modem set. If possible,
it would be preferable to buy a third one NOW instead of stealing one
from LOTS -- they wouldn't be terribly interested in talking to the
AI Lab but they will be interested in talking to LOTS. However, since
the AI Lab has a debugged Dialnet system, I will want the AI Lab to
have their setup.
Another thing that will be necessary will be to snarf a disk for my
stuff; they only have one drive and common sense dictates not running
an untested monitor on the public packs. Ralph says he would be
willing to lend us one on a short-term basis.
So we should get together and talk with the business school people
soon.
-- Mark
∂25-Sep-78 1640 SSO correction of seminar date
To: "@MODAL.DIS[1,SSO]" at SU-AI
The organizational meeting of the modal logic/ concurrent programming
seminar will be 9/28/78.
∂25-Sep-78 1659 Mcgoveran at SUMEX-AIM Faculty meeting
Date: 25 Sep 1978 1659-PDT
From: Mcgoveran at SUMEX-AIM
Subject: Faculty meeting
To: Faculty.list:, faculty.list:
cc: Tajnai
Due to the heat wave we're having, the location of the CS faculty meeting
tomorrow has been changed from Prolya 204 to the Cedar Hall Conference
room, which has an air conditioner.
The main issue will be voting on the degree candidates for this year.
-------
∂25-Sep-78 1954 DSB
To: JMC, LES
CC: RWW
As you may know I graduated from Stanford last June. I would like to
keep my SAIL account and am willing to do some free (or paid)
work in exchange. I realize the system load is high so I would try
to schedule around peak periods. Richard has not come
up with anything for me to do so perhaps you may have some ideas?
Thanks.
∂26-Sep-78 0918 Mcgoveran at SUMEX-AIM Today's Faculty Meeting
Date: 26 Sep 1978 0918-PDT
From: Mcgoveran at SUMEX-AIM
Subject: Today's Faculty Meeting
To: Faculty.list:, faculty.list:
Correction: The faculty meeting will be in Serra House, NOT Cedar Hall.
-------
∂26-Sep-78 0928 JRA lisp
a while ago you mentioned the possibility of having a LISP conference;
is that possibility becoming an actuality? if so, when? if not, I'm
volunteering!
john
∂26-Sep-78 0949 FB via SU-TIP S-1 management and funds
To: LES, JMC, LCW, FB
Since Les can't seem to figure out how to pay my legimate expenses as
a faculty member on the S-1 contract, I propose the following change
in arrangements to solve this problem. I propose that I be the PI on
the next contract and that the paperwork go through SEL instead of AI.
I can figure out how to transfer funds from SEL to AI even if Les can't
do the reverse. I assume that this is a feasible alternative to the
present arrangement since the three of you (and numerous others) have
asked me to take of the management of this project, a task I am happy
to do. On the other hand, if Les could figure out ...
Forest
∂26-Sep-78 1042 RWW
To: DSB, LES, JMC
∂25-Sep-78 1954 DSB
To: JMC, LES
CC: RWW
As you may know I graduated from Stanford last June. I would like to
keep my SAIL account and am willing to do some free (or paid)
work in exchange. I realize the system load is high so I would try
to schedule around peak periods. Richard has not come
up with anything for me to do so perhaps you may have some ideas?
Thanks.
I actually have some stuff but haven't sent it to dsb yet.
∂26-Sep-78 1123 LES Half-cocked response
To: FB
CC: JMC, LCW
You seem to have missed the point. I did not object to paying legitimate
expenses. I did object to making payments that are clearly illicit.
If EE wishes to expose themselves and Stanford to losses from government
auditing, they are welcome to do so, but they will not have my cooperation.
Apparently you have been misinformed. While I can understand how that
might happen, I do not understand your response. If you have a beef with
me, I expect you to try to communicate about the problem before you
broadcast your distress.
∂26-Sep-78 1151 PAT stoyan paper
I called Jean Sammet and she said that you should talk directly to
Wexelblatt, that she thought that he had written Stoyan regarding the
paper and that it was more complicated than just bad English. His
number is (215) 542-5150.
patte
∂26-Sep-78 1315 ARK S-1 Meeting Rescheduled
To: S1 at SU-AI
The S-1 meeting has been rescheduled to Wednesday, 4 Oct. at 1:15 p.m. in
the Serra House Conference Room. Random people interested in the project
are invited.
Arthur
∂26-Sep-78 1411 CH oak tree mazda
PLEASE CALL TOM MOBERLY AT OAK TREE AT 56-247-2212. THANK YOU
∂26-Sep-78 1647 REM at MIT-MC (Robert Elton Maas) Design for a Home-Information-Terminal (HOTER) service
Date: 26 SEP 1978 1948-EDT
From: REM at MIT-MC (Robert Elton Maas)
Subject: Design for a Home-Information-Terminal (HOTER) service
To: JMC at SU-AI
CC: RMS at MIT-MC, REM at SU-AI, PBARAN at USC-ISI
BACKGROUND:
For at least ten years Arpanet and related computer systems
have had various information utilities, including (1) system message
services (2) personal message services (3) INFO programs (SU-AI by
REM, and ITS by RMS) (4) bulletin-board programs (mostly on TENEX).
Until this year, such services have been limited to researchers at
universities and others on government contracts and semi-randoms who
wedged themselves into computer use by doing nice things (programming).
The general public hasn't had any use of these services, not even the
hobby-computer public. Now with the CBBS (Chicago, Atlanta, Santa Clara)
existing and available for free to the computer-terminal portion of
the public, setting a de-facto standard for such services, [my opinion]
the time has come to re-open the question of (1) services provided
(2) interface to user originally discussed in HOTER.ESS[ESS,JMC].
STATUS/EVALUATION:
The principal prototype systems are as follows:
(1) Sequential, with header selection: .MSGS.; at MIT-*
(2) Single keyword search: CBBS
(3) Subject areas with message numbers: BBD on TENEX
(4) Tree of information: INFO at SU-AI, INFO at MIT-*
(5) Multiple keyword search: Resource-1 (defunct c. 1974),
DIALOG at Lockheed (expensive!), SPIRES at Stanford, MARS at
Datacomputer (not yet easy to use).
After my use of the Resource-1 bulletin board system (actually
a genuine keyword-search boolean-combination system, more advanced than
current "bulletin board" systems), I realized that not only is keyword
search easier to maintain and easier to use than tree-of-information,
but tree-of-information can be simulated by funny keywords in a way
that allows a single system to have multiple trees of information within
it as well as lots of uncataloged (actually keyword-cataloged but not
tree-cataloged) documents. I wrote some suggestions about how to add
citation links and reverse citation index to create a truly good system.
To date, as far as I know, no system has been written having my proposed
capabilities. I haven't written it myself because I don't have the
authority to make it available to the public. I hope that with
the cost of computers going down down down to where individual
people can OWN a full dialup-access computer system, somebody will
let me guide him/her into making such a system available for
largescale public experimental-access (experimental in the sense
that we will be testing a new idea, but operational in the sense
that we will keep user-submitted documents forever instead of
flushing them as soon as we decide the experiment is over).
If my sketchy comments here are incomplete, you can't guess
the design of my proposed system, tell me and perhaps I can dredge
up some old SU-AI files containing my proposals.
At least the following features should be available:
(1) DIALOG-style expanding keywords alphabetical to catch
variant spellings.
(2) DIALOG-style selecting keywords and typing out the count
of documents.
(3) DIALOG-style combining via AND/OR/BUTNOT and again telling
how many documents.
(4) User-creation of new documents, with optional "strong
keywords". Weak keywords are those found anywhere in text of
document (title, body, headers, including strong keywords), whereas
strong keywords are just those supplied explicitly qua keywords.
Keyword access (1,2,3) should be able to distinguish them.
(5) User-linking of his document to others already existing.
I.e. citations (like footnotes or bibliography items). Thus an
[CRTSTY BLEW IT HERE, THEN OUTPUT SIDE OF NETWORK CONNECTION BLEW,
CONTINUING TO TYPE THIS MESSAGE...]
index may be created simply by listing subjects with citation-links
attached to each one, a multi-level index (a tree) may be done by
linking an index to other (inferior) indexes by citations. Structure
may be shared of course, perhaps having several toplevel indexes
pointing in different ways into an existing set of nodes.
(6) Random-number selection of elements of a set, to get a
representative sample of a set non-manually.
(7) KWIC search as well as actual document retrieval.
(8) Effective back-up-to-previously-visited-document feature.
(9) Least-recently-referenced garbage-collect, and command
to "ignore this reference" if you accidently visit a document you
don't want to see, or after seeing it you WISH IT WENT AWAY.~ User
should be able to mark a node as "for me, pretend it doesn't exist",
a stronger rejection than just "pretend I didn't see it this time".
(10) Weighting factors, like "Martin Gardner is a good
editor, if he reccommends something, show it to me in preference
to other documents" and "JULES is a cretin, avoid anything written
by him or reccommended by him but don't make it totally invisible".
(11) Author update and delete. If an author wishes to
withdraw a document, he may do so, so the only way to access it
is directly via document number (either type-in or from a list of
documents compiled before the withdrawl), and when so accessed it is
boldly marked as "withdrawn". Also, citation links to such documents
would shown the "withdrawn" indication to warn users of obsolescence
of document before they follow the link. If a document is edited,
a new copy is made, and a pointer is made from old document to new
document, also the old one is "withdrawn".
(12) Garbage-collector wouldn't actually delete anything,
it would swap it to a slower device, and request confirmation
before anyone could access it. Also, its keywords would be
flagged so that users could avoid accessing such documents if
desired. Of course, if such a file were accessed, it would be
swapped back in and would be an active document again.
(13) Each user would have password for login. Anyone
could create a new username just by logging in the first time and
confirming. If user happens to select a username already in
use (detected by: he tries to log in as FOO but doesn't know FOO's
password, system asks him whether he wants to try typing password
again, try typing username again, or create new account; he chooses
latter), it should be possible to EXPLODE the username list to see
an unassigned username he/she might want instead. Of course, if one
person wants to be two uses, for example JMC for official work and
PENGUIN for clandistine activities, there is no objection to it.
Since we aren't allocating a fixed amount of welfare money or computer
time per U.S. citizen, and don't care who a person really is, there
is no objection to dividing a person. And if there is a charge
for service, or voluntary contributions to establish higher than
default service level, the tradeoffs in having separate accounts
are reasonable. Only the user (with the password) who created
a document may withdraw or update it, although proposed-update
links may be established by other users and later co-authored by
the original author. If an account remains idle for longer than
some reasonable time (say 3 months unless user announces he is
going on vacation for longer), it is presumed the person has
died or forgotten his password, and his documents are declared
"author no longer around". Of course, a person who forgets his
password totally may create a new account under a different name
and use search facilities to mark all his documents as co-authored
by his new account, then after old account times out he will automatically
be the primary author for them all, thus avoiding lossage.
(This is an example of my general philosophy that NAKs are useless,
and ACKs are sufficient. Examples: If pointer to document or reference
to document recently it tends to sway around, otherwise garbage-collected.
If user logs in he exists, if lost 7 years presumed dead.)
(14) This is getting longwinded. ↑S (↑O on tops-10) should
always stop output immediately, then ? should briefly tell what
state your job is in, or other command should proceed in normal way.
∂26-Sep-78 1707 FB via SU-TIP S-1 contract and funds
To: LES, JMC, LCW
I have apologized to Les for my previous nasty remarks on his abilities
and we (Les and me) are trying to work things out. Sorry to have troubled
you people with my outburst.
--Forest
∂26-Sep-78 1827 MRC Dialnet at SAIL
To: IAZ, JMC, LES
I have made major incompatable changes to Dialnet to accomodate IAZ's
FTP program. Apparently SAIL choked on the previous incarnation. IAZ,
please see me ASAP to get the list of what has changed.
∂26-Sep-78 1855 MRC change to Dialnet programming at SAIL
To: IAZ, JMC, LES, BH, ME, JBR
Read DLNSER.DOC[S,SYS] which has been extensively revised. The difference
occurs in the status bit handling. The DSTAT word no longer exists, and
has been replaced by bits in DEVIOS. None of the "I/O error" bits are
set by the new DLNSER.
All this was done, by the way, for SAIL's benefit. *Sigh*
∂27-Sep-78 1126 MARG at PARC-MAXC ( Forwarded Mail )
Date: 27 SEP 1978 1127-PDT
From: MARG at PARC-MAXC
Subject: ( Forwarded Mail )
To: jmc at SAIL
Date: 27 SEP 1978 1126-PDT
From: MARG
Subject: (Response to message)
To: John McCarthy <JMC at SAIL
cc: MARG
In response to your message sent 26 Sep 1978 2353-PDT
Whitfield talked to Sarah ansd evidently she said that she left my shoes
in your house somewhere. I will try to connect up with them soon. As for
Sushi, im pretty busy this week. I was thinking of leaving for Bos tuesday
morning, but if it gets postponed to wednesday, I'll probably like to eat
sushi tues nite. btw, restaurant "fukisushi" on el camino way is terrific,
has sushi bar.
-------
-------
∂27-Sep-78 1906 Mcgoveran at SUMEX-AIM Informal Meeting of Faculty, Sen Res Assoc, Res Assoc
Date: 27 Sep 1978 1721-PDT
From: Mcgoveran at SUMEX-AIM
Subject: Informal Meeting of Faculty, Sen Res Assoc, Res Assoc
To: faculty.list:, JMCCARTHY, CARHART, CREARY, ZM at SAIL,
To: NII at SUMEX-AIM
Discussion of Conway survey of Computer Science Departments
As suggested at Tuesday's faculty meeting, we will hold a discussion
of the quality of the various Ph.D. degree granting departments of
Computer Science. This informal (and hopefully informative)
session will be held on Friday at noon in the Polya 204 Conference Room, and
will probably last for one hour. Bring your lunch. This session will provide
some additional input for your personal rankings on the forms I sent around.
Ed Feigenbaum
-------
-------
∂27-Sep-78 1957 SSO Visit and seminar
To: Hewitt at MIT-AI
CC: JMC at SU-AI, DPB at SU-AI
Carl,
Glad you will be coming to Stanford. I have set you up to talk in my
concurrent programming seminar on Thursday, Oct. 5 at 4:00. I will
make a general announcement, so we should get most of the Stanford people
who are interested.
I would be free to talk that afternoon, if you want. If that's not
convenient I'm sure we can find another time.
Sue
∂28-Sep-78 0126 MRC Tops-20 Dialnet
Well, the code is all written now. Sunday morning I will have 5 hours to
see how many different ways it will crash Tops-20 (lots, I know...).
For those of you who are into this sort of stuff, Tops-20 Dialnet requires
1200 words of resident code, 2173 words of resident data (mostly packet
data blocks), 160 words of bug strings + 20 words of bug pointers (not
all that important by anybody's reckoning), 17 words of initialization
code, and 721 words of swappable code. Hence Dialnet takes up about
4 pages in a Tops-20 system (all those numbers are octal of course),
not a severe monitor cost at all considering the benefits.
∂28-Sep-78 0838 AFB
To Professor McCarthy,
I hope I have selected the right initials to send to
as Finger was not enoughhelp. I am one of your as-
signed advisees (Anne Beetem-afb). Since your office is
a bit out of my carless range, I thought we could best
communicate this way,
particularly since we have taught our microsystem how to
be a terminal.
Anyway, I plan to spend this year preparing for the
Comprehensive. As such, I am taking courses this fall in
my least experienced subjects, numerical and systems.
(137A,142). I also signed up for Pascal to get back into
the swing of programming, as well as for CS200 to get
into the swing of the department. Lastly, pending your
permission before I turn in my study list, I have signed
up for 4 units of 390 under you to study for the AI branch of the test.
I intend to at least cover the material on the reading
list and will doubtless try some programming as well.
When I run into trouble I hope I can ask you about it.
I would have asked you about this before, but you weren't
around during registration.
I am pleased you have been selected as my advisor.
Any comments about my plans or any information about what
you are up to these days will be much appreciated.
Oh yes, how do I get a copy of the SAIL Monitor manual
Your's truly,
Anne Beetem
Please telephone me, because discussions are better conducted
by telephone than by message. Sorry I had to be in Michigan
during registration. We can also get together before or
after my 11am class on Tuesdays and Thursdays.
∂28-Sep-78 2027 CLT 206 notes
To: JMC, REP
Connie said they are back from the printer and should go to the library tomorrow
∂28-Sep-78 2247 MRC
To: JMC, LES
∂28-Sep-78 2215 IAZ Speed
1280 bytes were retrieved in 11sec →→ 930 baud!!
∂29-Sep-78 0055 REP NEW 206 CLASSROOM
The new classroom for CS-206 is 380Y. I believe this is in or around
the mathematics building on the quad. I will make arrangements to notify
the students of this change. I suggest you start class a little late on
Tuesday to allow for the room change to complete.
Rich
∂29-Sep-78 0808 RWW shoes
could i pick up my shoes at your house sometime today? or could you bring them
to sail and leave them somewhere accessible e.g. the front office? if you dont know where they are, coukd you give me a pointer to sarah so i could ask her? i
would like to leave for the sierra very early tomorrow.
∂29-Sep-78 1549 CLT
To: JMC, RWW, CG
According to an announcement on the math bulletin board, Dana Scott is to
speak Tuesday Oct. 3rd, 4:15pm, room 380C, on λ-Calculus: History and Models.
Thought you might be interested.
∂29-Sep-78 1647 Feigenbaum at SUMEX-AIM dinner sunday night with the Michies
Date: 29 Sep 1978 1647-PDT
From: Feigenbaum at SUMEX-AIM
Subject: dinner sunday night with the Michies
To: jmc at SAIL
John, yes we're on for dinner on Sunday! (Don't forget about it!)
We'll gather about 6pm, though some will swim for a while. If you want to swim,
bring a suit (or skinny-dip).
Ed
-------
∂30-Sep-78 1010 Jrobinson at SRI-KL (Jane Robinson) brunch
Date: 30 Sep 1978 1010-PDT
From: Jrobinson at SRI-KL (Jane Robinson)
Subject: brunch
To: mccarthy at SU-AI
I checked with Young & Co. and they do have a pretty complete variety
of food for their brunches--bacon, sausage, omelettes, etc. (not all
alfalfa sprouts). It is also very wheel chair accessible. So why
don't we try it tomorrow at 11. Margaret and I will meet you there.
I've had supper there a couple of times and found it good, though
not great. If you know of a great place not too far from here, I'm
open to suggestion. Otherwise, see you at Young and Co., on California
Ave (447 California), first block bayside of El Camino, at 11 am.
Jane
-------
Fine, see you there. Sorry to be troublesome about my diet.
∂30-Sep-78 1726 AFB
Can I make an appointment to see you Monday afternoon
after 2:00?
∂30-Sep-78 1935 JMC note for LISP bulletin
To: PG
CC: JMC
Perhaps you would like to print a short note on Takeuchi's function.
XS TAKEUC.XGP[F78,JMC] will get you a copy. It occurs to me that
your readers might also be interested in Takeuchi's survey of the
speed of LISP systems, but you would have to write to him about that,
because my copy is still being mailed from Japan.
∂30-Sep-78 2301 PG tech-note-for-lisp-bulletin
Thank you for your contribution to the bulletin. The text display
is really beautiful. In the course of reading it, an editorial
consideration came to my mind: you omitted to sign the note.
Is it intentional ?
Also the ref. to (McCarthy 1978) on page 1, is not made explicit,
is it "LISP Programming and Proving" ?
About the fact mentionned at 2., i.e. the value is one of x,y or z.
I came to this conclusion when you showed this function to me, 2 weeks
ago, by the following course of reasoning (terribly informal: I have
NO experience in program proving, but if it is not blatantly false,
perhaps it could be cleaned up).
If x ≤ y then value-is y.
Else x>y and, assuming TAK terminates, the call:
TAK(TAK(x-1,y,z),TAK(y-1,z,x),TAK(z-1,x,y))
reduces to:
TAK(TAK(y,y,z),TAK(z,z,x),TAK(x,x,y))
i.e. to:
TAK(y,z,x)
If y≤z then value-is z.
Else y>z ∧ x>y and,the same way,
TAK(y,z,x)
becomes
TAK(z,x,y)
and y>z ∧ x>y ⊃ x>z then value-is x.
The text display is my first attempt at using Don Knuth's TEX.
Naturally, column width can be adjusted to suit your requirements.
I intend to sign the note. Perhaps it would be worthwhile to include
a paragraph explaining the publication notation as well as the reference
to the book which will be made explicit.
I don't understand your proof; I don't understand how you go
from tak(x-1,y,z) to tak(y,y,z), etc. What hours do you keep? Perhaps
you could explain it to me.
∂01-Oct-78 1627 MRC Dialnet bug
To: IAZ
CC: LES, JMC
I found it. It was a deeply imbedded system bug. Basically, if you
were sending a return close by doing a RELEASE on the channel, it would
set up the return close and start it. HOWEVER, there was nothing in
the world which prevented the RELEASE handler from cancelling the
port's output, which in this case was the CLS! The fix was simple: to
hang the close until the output list becomes empty on its own free will.
∂01-Oct-78 1729 RPG Require
To: "@MACLSP.DIS[AID,RPG]" at SU-AI
Let me know if you have trouble with REQUIRE in the compiler.
In particular, if it seems that after a REQUIREd file has been compiled,
some s-expressions are mis-read.
-rpg-
∂02-Oct-78 1530 SSO Seminar for Thursday, Oct. 5 at 2:45. ERL 237
To: "@MODAL.DIS[1,SSO]" at SU-AI
!
Some Controversial Conjectures in the
Semantics of Concurrent Systems
Carl Hewitt
M.I.T.
By a concurrent system, I mean a network of digital mechanisms connected
asynchronously, where each mechanism has some processing power and a finite but
possibly unbounded amount of storage. Recently three models of concurrent
computation (Petri Nets, Communicating Sequential Processes [Hoare, Milner], and
Actors) have been given mathematical foundations. On the basis of these
foundations, I will draw some tentative and controversial conclusions.
The hypothesis of bounded nondeterminism asserts that for any concurrent
system which is guaranteed to terminate, there is a bound on the number of
states in which it can terminate. The hypothesis has played a prominent role in
some recent models of computation. An important design decision in
Communicating Sequential Processes [Hoare: CACM Sept. 1978] has been justified
on the grounds of excluding the possibility of unbounded nondeterminism. One
consequence of the decision is that communicating sequential processes cannot be
used to implement some useful public information processing utilities that
guarantee to serve customers on a network. (In defense of communicating
sequential processes, it can be argued that they were never designed to
implement such network utilities.)
The actor model provides a mathematical semantics for concurrent systems
in which the possibility of unbounded nondeterminism is not excluded. We have
found it a useful tool in describing the behavior of concurrent systems such as
network utilities which make strong guarantees of service.
∂02-Oct-78 1533 SSO Lunch
To: "@MODAL.DIS[1,SSO]" at SU-AI
If you would like to have lunch with Carl Hewitt next Thursday, let me know.
∂03-Oct-78 0557 ZM car
Do you need the DODGE today afternoon ? If not - I have a visitor from Israel
that would like to use it for few hours. Zohar
∂03-Oct-78 0858 SSO TIME CHANGE
To: "@MODAL.DIS[1,SSO]" at SU-AI
Since I forgot to get the time of Carl's talk changed with CS,
an official announcement went out naming 4:00 as the time. So
let's meet at 4:00 this Thursday. After that we can go to 2:45.
∂03-Oct-78 1622 CH phone call
please call Sara, she said you had her number. thank you
∂05-Oct-78 0814 JRA lisp conf
did you get my note about a lisp conference? i think the world needs one
Yes, and I think I agree and would be glad to have you organize it, although
I don't know what resources might be required or whether we have them.
∂05-Oct-78 0923 CH DIALNET USE
JEFF MOORE OF THE BUSINESS SCHOOL CALLED THIS MORNING TO LET YOU KNOW THAT YOU
CAN USE THEIR MACHINE FOR DIALNET FROM 2 AM TO 6 AM. MARC CHRISTIN--CONTACT
DON ESTAVAN AT 7-4265 IN ORDER TO GET KEYS. CHARLIE
∂05-Oct-78 1010 Feigenbaum at SUMEX-AIM copy of message to Carlson
Date: 5 Oct 1978 1010-PDT
From: Feigenbaum at SUMEX-AIM
Subject: copy of message to Carlson
To: jmc at SAIL
Dear Bill,
One more note about the symbolic computing meeting at MIT...
John McCarthy does not seem to have been invited. Is that an
accident I hope so.
John is the originator of LISP, and over the years has thought
more deeply about LISP (its internal machinery and its applications)
than almost anyone else in Computer Science. His presence at a
meeting that will help shape the future of LISP use is therefore
very important, in my opinion.
I urge you to fix the bug.
Best regards,
Ed
-------
∂05-Oct-78 1045 BPM Recognition of human faces
Recalling Hubert Dreyfus' claim that a computer would never be capable
of recognizing human faces, you (and he) might be interested in a recent
thesis which continues Mike Kelly's work. "Computer Recognition of Human
Faces" by Takeo Kanade, Birkhaeuser Verlag, 1977. The work was done
at Kyoto Univ. The book is in the CS Library now.
∂05-Oct-78 1100 MRC*
Please call the Business School irt Dialnet.
∂05-Oct-78 2252 MRC VADIC modem for GSB
To: LES, JMC
Well, we are going to have to go and give them a modem now. It looks
like they are willing to let me have as much downtime as I want between
2 and 6 am; they seem to be very nice and cooperative people. I called
Les this afternoon and hopefully the ball is rolling to get a DAA and a
phone line installed there. Do we have enough money in the remainder of
this year's Dialnet budget to buy another modem, or do we have to wait
until next year? We'll have to figure on about a month before a DAA and
phone line gets installed at GSB (unless we can get a super-rush order
in?), so if we can get another modem ordered now, we might with appropriate
timing be able to set up the LOTS and GSB nodes at much the same time,
having a 3-site rootnetwork!!
-- mark
Please ask them if they can loan us a line until we can get one
installed. Most likely, we will have to move the LOTS
modem to GSB. When Dialnet is working, both LOTS and GSB will
have to buy their own modems. We will probably move both
present modems to AI, because AI will have heavy Dialnet
traffic while other organizations are testing their implementations.
∂06-Oct-78 1100 MRC*
Please call the Business School irt Dialnet if you haven't already.
∂06-Oct-78 1349 MRC via SU-TIP Dialnet
To: JMC, LES
We can get a port, but all the lines are taken. I thought we were
more or less going to permanently leave modems at LOTS and GSB.
I don't think our traffic will that heavy; and I suspect that
Tops-20 sites will want to talk to LOTS and GSB rather than to
us!
It isn't the responsibility of the Dialnet project to supply LOTS
and GSB with modems permanently. LOTS, at least, will buy its own
once Dialnet is working.
∂06-Oct-78 1855 MRC FYI
To: JMC, LES
∂06-Oct-78 1448 DGR via NBS-TIP Dialnet at NIH
I have lost my 1200 baud connection to the ARPAnet so have nt logged
in and around lately. Thanks for your offer to come install dialnet
at NIH. Let me tell you what's happened.
As I had told you before, DIALnet was not top priority here, but I was
trying to work it in. I did get a Vadic 3405 modem pu on the front-end,
but we have no dialer or dialer-code for it. I then passed around the
protocol document to some of the minicomputer users at NIH to get a
reading of their interest in using Dialnet to send NIH
scientific data around. A positive response here would have been
sufficient justification for us to do some more work on it with our -10.
Unfortunately, I got no positive response; people are more interested
in non-error checked data transfer (this gbroup of peopoe apparently
has negligible trou lbe with line noise in their data), DECnet (because
it has clear support from DEC however crud it is), or higher speed
data transfers than Dialnet would offer them (they would like to use
DMC-11s at 9.6Kb+).
Well, this doesn't kill Dialnet at NIH but it does ruin my vehicle.
We will have to give higher priority to other things,
which does mean Dialnet implementation is postponed indefinitely.
I am going to see what kind of support there is for putting up Dialnet as a means
of communication between Washington-D.C. area DEC-10's and DEC-20's.
While most will be extremely reluctant to make monitor changes,
there may still be some support.
In other news, we are spending significant time (just so you can see what
is more important) on the RFP we have to issue to replace our
KI's with KL's, with screen programming for the new NIH7000 CRT
terminals being deliverted now (it has a TMS-9900 built in with 8K ROM/8K RAM)
and Alanthus T1222 hardcopy attachments, and with bringing 4 new
people up to speed so that they can help us.
I'll let you know what happens and I hope the news (from the Dialnet
standpoint) will be better next time.
∂06-Oct-78 2230 PEG S1 meeting notes-- new location
To: "@S1[DIS,S1]" at SU-AI
All meeting notes for full project meetings are now in MEET.NOT[DOC,S1],
including those for 10/04/78.
∂08-Oct-78 0523 RWG at MIT-MC (Bill Gosper)
Date: 8 OCT 1978 0823-EDT
From: RWG at MIT-MC (Bill Gosper)
To: jmc at SU-AI
"we" have chosen an impropitious time for the match.
the prototype lispmachine is being retired and its chaosnet connection
has been moved to a newer machine, but the cheops machine is still
connected to the old one, rendering cheops inaccessible except via
the ka10@ai. (I learn this from moon, since rg has not been answering
my queries.) So unless rg has a fix for these problems, we are stuck
playing the kl@mc without the blunderstopper. if you wish to go
ahead, i'd prefer rather late, e.g. 2230.
2230 seems ok to me.
∂08-Oct-78 1731 LLW Jeff Rubin
To: JMC
CC: LLW
John, I forgot to mention to you when we spoke last weekend that Jeff
Rubin has recently asked me for a professional appointment here at LLL.
Because of the very fine work he has done for the S-1 Project, and based
on his excellent record with you, I felt obliged on behalf of the
Laboratory and the S-1 Project to set the local wheels in motion to get a
competitive offer out to him. I expect that the Lab will be making an
offer to him in about a week, in the 28-30 K$/year range. I understand
from Jeff that LLL is not the only place that he has spoken to about
relocating, and I have thus made quite vigorous efforts to see that the
Lab's offer to him is as attractive as UnivCalif professional research
salary scales permit; I would be extremely sorry to see him abandon the
superb work he's doing in the S-1 Project.
Please believe that I am not attempting to poach in your territory, and
that I have scruplously refrained from recruiting any of your staff; I
highly value the evolving relationship between SAIL and LLL, and recognize
that pirating people is incompatible with such a collaborative research
program. Jeff approached me without any encouragement, and asked for a
job offer without any preamble.
I really don't know why Jeff has expressed this interest in leaving SAIL,
and I guess it's really none of my business. I do have the impression
that he's not really highly enthused about the prospect. I therefore hope
that you and he may be able to eliminate whatever present problems he has,
so that he can remain a happy and productive member of the SAIL staff, and
your lead software man on the SAIL S-1 work.
I'll of course be very willing to discuss this matter with you in as much
more detail as I can, if you wish.
Lowell
I do not have any objection to LLL making Jeff a job offer. In fact I don't
think that one organization has a right to refrain from making job offers out
of courtesy to another. We will try to compete - both financially and in
other ways.
∂09-Oct-78 1649 Lederberg at SUMEX-AIM Corrected directory entry SAIL]: people.dat[per,csd]
Date: 9 Oct 1978 1649-PDT
From: Lederberg at SUMEX-AIM
Subject: Corrected directory entry SAIL]: people.dat[per,csd]
To: cet at SAIL
cc: jmc at SAIL, les at SAIL, FEIGENBAUM at SUMEX-AIM
Lederberg, Joshua
Professor, Genetics, CS by courtesy
President,Rockefeller University, NY NY 10021
JSL
HPHONE 212/360-1022
OFFICE R.U., 1230 York Avenue, NYC 10021
OPHONE 212/ 360-1234
EMAIL JSL ⊗ LEDERBERG@SUMEX
GROUP HPP FACULTY MOLGEN COURTESY
LABPJCT Stanford
LABDISK 200
LABBAMS 0
LABDATE 75:
LABAUTH JMC indefinite
-------
∂09-Oct-78 2308 LGC via AMES-TIP This Afternoon's Discussion
Please take a look at the new, corrected and expanded version of INT.TXT[EP,LGC].
It gives a much better picture of the situation than the sheet you looked at
this afternoon. -- Lew
∂10-Oct-78 0827 CARLSON at USC-ISI Real-Time Symbolic Computing Workshop
Date: 10 OCT 1978 0828-PDT
From: CARLSON at USC-ISI
Subject: Real-Time Symbolic Computing Workshop
To: McCarthy at SAIL
cc: Carlson, Townley at HARV-10
Professor McCarthy:
There is to be a workshop on real-time symbolic computing at
MIT on 28 and 29 November 1978. One purpose of the meeting is to
unveil the MIT LISP machine. We had an administrative problem on
this and, so your invitation is just being mailed today. I
hope you will be able to attend.
Sincerely yours,
Bill Carlson/jagl
-------
∂10-Oct-78 1156 Mcgoveran at SUMEX-AIM Computer Science Visiting Advisory Committee
Date: 10 Oct 1978 1115-PDT
From: Mcgoveran at SUMEX-AIM
Subject: Computer Science Visiting Advisory Committee
To: faculty.list:
Below is a copy of the agenda for the meetings to be held on Monday,
October 16.
Professor Feigenbaum said you are welcome to attend the sessions during
that day.
If you would like to attend the lunch at Durand at noon on Monday, please
let me know by 2 pm on Friday, October 13, and I will order a box lunch
for you.
MONDAY, October 16
8:00 am Continental breakfast, Durand 450
8:30 Plenary sesson, Durand 450
Educational Computing, Gerald Lieberman
LOTS, Ralph Gorin
Allocation
Office Information Systems
11:00 Students
12 noon Lunch will be served in Durand 450
1:00 pm Two concurrent sessions:
Durand 450: Computer Science Department
Status of the Department, Edward Feigenbaum
The Computer Science-Electrical Engineering Liaison,
John Linvill
Information Systems Laboratory, James Meindl
Students will be invited to speak
Durand 301: SCIP
Overview: Selected Bright Spots and Problems,
Chuck Dickens
Campus Facility, Ed Williams
Administrative Data Processing Services, Ced Bennett
Students will be invited to speak
4:30 Wrap-up session in each of the two groups
5:00 Plenary session, Durand 450
6:00 Working dinner at the Faculty Club
-------
∂10-Oct-78 1557 DPB reminder re cs200
DATE 10/19/78
SERIES CS200
SPEAKER John McCarthy
TITLE "Research at the Stanford AI Lab"
TIME 2:45
PLACE ERL 401
Informal description of CS200 for potential lecturers
CS200 is intended to be an introduction to the research being done in the
Stanford CS Department. It is intended for newly entering students, in
both the PhD and Masters programs. The course is organized as a series of
lectures by CS faculty members. One unit of pass is available for
attendance. The course is intended to be (primarily) a forum for new
students to meet the faculty and learn what the faculty is most interested
in.
Each week a faculty member will be invited to present a lecture on what
his/her field is all about, what s/he is doing, and what sort of ongoing
projects s/he is involved in. The style of each lecture will certainly
vary depending on the speaker and his/her field of interest. I will
suggest the following questions as typical of the sort a speaker might
want to address.
Definition/description of the field of interest. What are the classical
questions of the field? What are the classical tools applied to discover
answers to the questions? In what courses, in what books, can one find a
good introduction to the field? What is a representative open problem?
(I.e. what is the speaker working on right now?) What contribution, if
any, can a new grad student make to a research project? Can someone whose
primary interests are not in the field make a contribution to the
research? (E.g. Graphics programming in the NA world.) Where is the
interesting research going on? Who is doing it? How does the Stanford CS
dept. stack up compared to other places?
∂10-Oct-78 1610 PAT my schedule
To: JMC, LES, JC, MUZ, HVA
my probable schedule for the next month:
October 18,19, and possible 20 El Camino Hospital
October 20-26 home (I will be available by phone and
can possibly come in for a couple hours)
October 27-28 lab
October 30-Nov 7 Northwestern - Computer Music Conference Nov 1-5.
November 7 back
∂10-Oct-78 1953 MRC Dialnet metering
To: LES, JMC
I wish to take the system down so that I may meter how much of the
machine Dialnet requires. Basically it involves comparing the KL10
meter for channel 5 when the machine is totally idle with the same
thing with continuous Dialnet traffic (I guess I'd want to try both
no-ops and real data transfers).
Anyway, while I feel this information is important, I don't feel that
5 in the afternoon is a good time for it. It can't really be
considered "system" time in any sense of the word, and additionally
there are political implications about a research project taking the
machine for standalone with no system justifications. Please advise.
-- mark
∂10-Oct-78 2217 Feigenbaum at SUMEX-AIM carlson response to my msg about jmc
Date: 10 Oct 1978 2217-PDT
From: Feigenbaum at SUMEX-AIM
Subject: carlson response to my msg about jmc
To: jmc at SAIL
Mail from USC-ISI rcvd at 10-Oct-78 0900-PDT
Date: 10 OCT 1978 0901-PDT
From: CARLSON at USC-ISI
Subject: Re: More on Symbolic Computing meeting at MIT
To: Feigenbaum at SUMEX-AIM
cc: CARLSON
In response to your message sent 5 Oct 1978 1009-PDT
Thanks for calling the oversight to my attention. I
obviously didn't check the mailing list carefully enough.
Bill/jagl
-------
-------
Thanks. I got a message from Carlson yesterday inviting me.