perm filename SAVED.MSG[ESS,JMC] blob
sn#189795 filedate 1975-12-02 generic text, type T, neo UTF8
∂26-NOV-75 0510 S,LES
Samet is still running big compute-bound jobs in the early morning.
JBR claims that the drawing program doesn't run decently as a result.
Unless you believe that Samet is doing something good, I think we ought
to request that he go away, or at least substantially reduce his
load.
Another hanger-on is YYY. I have no idea why he is still running here.
Do you?
∂20-NOV-75 0153 S,LES
Any reaction to the Somalvico proposition?
∂19-NOV-75 2036 FTP: host CMU
From: SIGART(X180LE03)@CMUB
Date: 19 Nov 1975 2336 EST
Subject: your note of 15-Nov.
To: JMC@SU-AI
cc: SIGART, LES@SU-AI
- - - -
Thank you for the note re Minker's letter and also for the postscript
re I. Goldstein's draft notice. I would like to print your note
along with Minker's letter and the letter from the Goldstein's -- do
you have any objections to that? If you do, or you if you would like
to expand on it, please let me know within a week or so -- we hope to
go to press by around November 17.
thanx again for the note,
Lee Erman
-------
∂19-NOV-75 2036 FTP: host CMU
From: SIGART(X180LE03)@CMUB
Date: 19 Nov 1975 2336 EST
Subject: your note of 15-Nov.
To: JMC@SU-AI
cc: SIGART, LES@SU-AI
- - - -
Thank you for the note re Minker's letter and also for the postscript
re I. Goldstein's draft notice. I would like to print your note
along with Minker's letter and the letter from the Goldstein's -- do
you have any objections to that? If you do, or you if you would like
to expand on it, please let me know within a week or so -- we hope to
go to press by around November 17.
thanx again for the note,
Lee Erman
-------
∂19-NOV-75 0439 S,LES
To: JMC
CC: PAW
I'm trying to organize a register of reference materials owned by people
in the Lab. See Page 1 of REFER[BIB,DOC] for specifics.
If you wish to participate, Patte could organize you entries.
∂19-NOV-75 0057 S,LES
Received phone message about a telegram from Marco Somalvico
requesting permission to come here on a NATO fellowship. Says their
deadline is Nov. 23. I don't have the full text, but will pick it
up while I am on campus tomorrow morning (i.e. this morning).
My recollection of his performance on the earlier two-year stay is that
he published a Comm. ACM article on something unrelated to anything here,
generally had a good time, and contributed nothing to any of our projects.
∂18-NOV-75 2239 VCG,DCL
VERIFICATION GROUP MEETING
CORRECTED SCHEDULE
Thursday Nov.20 3.00pm:
R. Cartwright and R. Weyrauch: Problems with Pascal Semantics
Thursday Nov.27, 3.oopm: Thanksgiving!
Thursday Dec.4, 3.oopm:
John McCarthy: Practical and Theoretical Considerations in
Programming Language Design.
(ignore all previous messages!)
∂18-NOV-75 2213 VCG,DCL
@MEET.MSG[IBM,DCL]
∂04-NOV-75 2322 S,LES
OK, I'll do that if you edit FR[R,LES].
∂21-NOV-75 1748 S,LES
There is a version of the bellnet proposal in BELL[R,LES]. It has
an elaborated scenario on page 3 that reflects my prejudices on the
list of capabilities that ought to be developed first. See what you think.
∂17-NOV-75 0033 S,LES
Regarding IJCAI, have you decided where to print you review yet?
You could, of course, put it in SIGART Newsletter, but a wider forum
would probably be better.
On the idea of protesting the treatment of Goldstein: I could suspend
the distribution of free SIGART Newsletters to the USSR, with a
suitable note of explanation. This would have the beneficial side
effect of saving a few bucks, SIGART funds being low currently.
How does that strike you?
∂15-NOV-75 0328 S,LES GRIPE Handling
To: ME, TAG, BPM, ALS, SUPER[S,LES]:;
This note describes a proposed change in our procedures for handling
GRIPEs and an accompanying change in MAIL that hopefully will not
be very hard to implement.
The problem is that from the viewpoint of the user community, sending
a GRIPE is roughly equivalent to shouting into a hole in the ground,
except that if you choose your hole right you at least get an echo.
Feedback from gripes appears to be the exception rather than the rule.
From the user's standpoint, that is very frustrating.
Admittedly, a lot of gripes are based on ignorance or apparent stupidity,
but even these often convey indirect information about documentation
shortcomings or the need for an education effort. It isn't necessary
to type a lot in order to provide a response; a small set of standard
response files will handle a large fraction of the cases, e.g.
(Putdown) "You clearly haven't read the relevant documentation. If
you read it and still don't understand, then please come talk to me
and waste some more of my time."
(Arcane knowledge) "Your suggestion cannot be implemented for reasons
that are too complicated to describe here. Please come see me if you
need to know more."
(Subtile evasion) "Thank you for your suggestion. We will give it
careful consideration in future implementations of the program."
(Positive indefinite) "We should take care of that, but with current
priorities, it probably won't be very soon."
(Positive immediate) "Thank you for calling that to my attention.
I will take care of it right away."
The problem with the current arrangement is apparently NOT that there
is a large volume of gripes to deal with, which may be partly a result
of the way they aren't dealt with. The reason why so many go unanswered,
especially the stupid ones, is that nobody quite feels responsible
for doing it.
I discussed (or, more accurately, attempted to discuss) the need for
feedback with REG, who as usual started his "Don't try to tell me
how to do my job" act at full volume and didn't settle down for 10
minutes. In the course of this "discussion" it became clear that
Ralph doesn't wish to either handle most gripes or act as a switch,
to route them to the appropriate person.
JBR happened by and made a suggestion: put the initial switching
function into the GRIPE command by taking a topic and routing the
gripe to the poor soul who is responsible for that topic. I like
the idea and believe we can use it to deal with more than just
system problems. For example, if you type "GRIPE", it could
respond with something like the following:
"Future gripes may be entered in the form "GRIPE <topic>".
Please select one of the following topics and type enough characters
to identify it unambiguously.
BUILDING (temperature, lights, furnishings, cleaning, etc.)
COMPILERS (FAIL, SAIL, FORTRAN, SNOBOL, PALX)
COMPUTER (CPUs, peripherals, terminals, etc.)
DEBUGGERS (RAID, DDT, BAIL)
EDITORS (E, SOS, TECO, ...)
LIST PROCESSORS (LISP 1.6, ILISP, MLISP, ...)
MONITOR (timesharing system)
OFFICE EQUIPMENT (Xerox, Velobind, etc.)
PROCEDURES (X isn't being done right)
TELEPHONES (malfunctions, changes)
TELNET (ARPA net, FTP)
UTILITY PROGRAMS (e.g. COPY, DART, MAIL, SNAIL, SPOOL, XSPOOL)
OTHER (anything not covered above)
*"
After the person selects a topic, it says "This will be sent to xxx"
then accepts the message and sends it both to the selected person
and to the GRIPE file.
Clearly a lot of these topics cannot be fully handled by a single
person, but he can perform a subswitching function to get it to
the right place and make sure that a response is generated.
I will volunteer to take the PROCEDURES and OTHER topics.
We can adjust the topics and recipients as we like to deal with the
flow. In addition to the modification to GRIPE, the conversion of
MAIL to E format would greatly aid in the message handling process
outlined above.
Well, what do you think of that?
∂13-NOV-75 1216 1,PAW
Bill Leland, ARLO Office, Tressider Union, would like to be added to the
mailing list of announcements for future SENSE seminars.
∂12-NOV-75 2110 IBM,DCL
VERIFICATION GROUP MEETING
Thursday Nov.13 3.00pm:
R. Cartwright: Problems with Pascal and its Semantics
Thursday Nov.20, 3.oopm:
John McCarthy: Practical and Theoretical Considerations in
Programming Language Design.
∂12-NOV-75 1646 1,TOB
Rybak did send a letter before IJCAI.
I remember some hesitation on your part,
and I wonder whether you want to relate
this to the Goldstein incident?
(Rybak is from Kiev)
Tom
∂12-NOV-75 1638 1,TOB
What do you want to do about the invitation
which Rybak requested to visit the lab?
Tom
∂11-NOV-75 0314 100,100: REM @ AMET
I believe you are the person who has voiced the philosophy
that side effects are not important, the only important thing in a
computation is what is outputted at the top level?? The continued
fraction and continued logarithm algorithms that I am working on
with Gosper offer the possibility of directly implementing this.
Suppose the user has a FORMAT statement in FORTRAN, or the equivalent
in SAIL or whatever, which specifies the output field of some datum
to be a certain number of characters long. The continued-mumble
can be set up to compute exactly that many characters, and being
output-driven only those intermediate results that are needed to
compute the output will ever be actually computed. (Of course,
on an interactive system, you might start with one or two digits
per output datum and ask for more digits in any particular datum
as you choose, at which point the system would squeeze out the
new digit you requested, doing only as much internal squeezing
as was needed to compute it, and if interval-data was specified
for input it would request more accurate input at exactly those
points where needed.)
∂30-OCT-75 1746 FTP: host ISI
Date: 30 OCT 1975 1744-PST
From: RUSSELL at USC-ISI
Subject: AI Program Plan
To: LES at SU-AI
cc: RUSSELL, BLUE
Les,
I still haven't received the natural language milestones
for the AI program plan from Stanford. Please give me your comments on this
subject as soon as possible.
Thanks
Dave
-------
∂02-NOV-75 2316 S,LES
Is anything happening on the Natural Language Milestone front?
Also, did you send Russell something earlier relating to his AI Plan?
∂06-NOV-75 1014 1,HVA
Sponsored Projects informed me this a.m. that NASA has approved
proposal and $45,000 will be forthcoming in about 3 weeks.
∂05-NOV-75 1057 FTP: host RAND
M E M O R A N D U M 5 Nov 1975
To: F. Corbato, J. Markowitz, J.McCarthy, M. Pirtle,
T. Stockham, I. Sutherland, B. Woods
From: Peter Weiner (WEINER @ RAND-RCC)
Copies: Hilda Bankert
Subject: Computing Resources Meeting
--------
As a means of reviewing the issues raised in our report as well
as considering other long-term options open to IPTO in the area
of computer resources, Dave Russell has scheduled a meeting at the
Rand Washington Office on 21 Nov 75.
To structure the meeting to some extent he has asked various IPTO
program managers to chair sessions which address each recommendation.
In addition he has proposed to have a general discussion at the
end to probe other computer resource options not covered by the
report which should be considered.
If your schedule will permit, I would ask each of you to attend and
to contribute your views on this important matter.
The meeting will begin at 9:30 AM on Friday 21 November 1975 at the
Rand Washington Office, 2100 "M" Street, N.W., Washington, D.C. 20037.
Please advise me of your plans.
--------
∂04-NOV-75 0914 1,HVA
Mrs. Chertoff, Cttee. for Concerned Scientists, called at 9:05 a.m.,
and is very anxious to speak to you. The collect tel. no. to return
call is (212)679-643l.I've put note in your office, but this is in
case you log in from somewhere else.
∂03-NOV-75 2131 S,LES
There is a draft Formal Reasoning section for the Annual Report in FR[r,LES]
∂31-OCT-75 1529 2,JH
John, I need a story for the campus report by Thursday, November 6.
It would then appear in the Campus Report of November 12. That's
cutting it pretty close. If you, Felix Bloch and John Connolly can't
get together on it, then you should write it and get their agreement on
it. You might also think about how you would like to approach the Stanford
Daily about SENSE. I will see to it that the meeting appears in the Daily's
"Once around the Quad".(great bug in mail if you backspace one too many it gives
what youre seeing above) Let me know if you don't want to write a story so
I can take it off my worry stack.
Jim
∂30-OCT-75 1746 FTP: host ISI
Date: 30 OCT 1975 1744-PST
From: RUSSELL at USC-ISI
Subject: AI Program Plan
To: LES at SU-AI
cc: RUSSELL, BLUE
Les,
I still haven't received the natural language milestones
for the AI program plan from Stanford. Please give me your comments on this
subject as soon as possible.
Thanks
Dave
-------
∂29-OCT-75 2357 100,100: jmc
call Gunkel about Gilfillan
∂29-OCT-75 2147 S,LES
I recently wrote a Sail program called POKE.SAI[M,LES] which you
might want to use occasionally. It looks at disk areas, compares
with the phone lists, and shows peculiarities of three types:
1) legitimate people who have no PPN or who haven't logged in for quite awhile
(cutoff date is requested at the beginning of the run; default is a month ago);
2) people whose accounts were flushed this year, but who have been running
since the cutoff date;
3) outlaws.
For each such person, it shows the date and time of last logout.
Just say "EX POKE[M,LES]".
∂27-OCT-75 1018 L,HJS @ NBST
John,
This note contains several questions, comments, and answers to
questions you raised this summer prior to my leaving.
1. You mentioned to me that Goto, in Japan, had implemented a
LISP system using hashing. In the meantime I have seen several
reports from the University of Tokyo referring to this system.
Specifically, there are references to the following two reports:
Goto, E., "Introduction to LISP", bit vol. 6, No.1-No.13, Kyoritsu
Shuppan, Tokyo, 1974-1975.
Goto, E., "Monocopy and Associative Algorithms in an Extended
LISP", Internal Report, Information Science Laboratory, University
of Tokyo, May 1974.
If you do have these reports I wonder if you would be kind enough
to permit me to have a look at them.
2. Prior to my leaving I gave you a copy of a paper I had written
describing a Normal Form for LISP programs. Could you please give
me some advice on where I should submit it for possible
publication.
3. I would like to know if you will be assigning the compiler
writing project in CS206 this quarter. My proof system is now up
at IMSSS and can be invoked by use of the command VERIFY. Tom
Wolpert has aided in bringing it up. The entire system is
compiled and runs quite fast so that there is no need to worry
about execution time. Will your students generate LAP code for
ILISP or for LISP 1.6. This information is necessary since the
LAP formats are slightly different. If code is being generated
for ILISP, then an appropriate flag must be set in the system. A
manual for using the system is available on my disk area and is
known as PROVE.PUB[L,HJS]. It is in PUB format, so it will have to
be compiled with PUB and the output obtained from the corresponding
XGP file. In the near future, time permitting, I will attempt to
produce a more substantive manual which will demonstrate the use of
the system in debugging programs.
I have also included provisions in the verification system that is
running at IMSSS for recording information about each use of the
system. I believe that this would yield some interesting
information. I am particularly interested in the complexity of the
programs that are presented to the system for proof. Another
equally promising set of data is the type of errors that are made
by students and the success that the system has in detecting and
pinpointing them. Examination of subsequent runs can lead to an
evaluation of the systems use in aiding the debugging of compilers.
This can be most easily seen by determining if the errors that have
been pinpointed by the system are removed from the erroneous
programs, and if the corrected programs are resubmitted to the
verification system. This would be very much along the lines of
the results that I presented in the chapter of my thesis that
showed the system in use in debugging a function from RWWs FOL
system.
My plan of action is as follows: Once the class starts using the
system, I would log in daily at IMSSS and examine the current use
of the system. If this proves to be impractical, then I will
simply sample results. In order to do this I will need an
account at IMSSS and some file space to store these temporary
results. With respect to the latter, I will speak to Bob Smith.
4. Prior to my departure, you inquired as to whether or not my
system could handle a modification to the following example: The
function NEXT, given a list and an atom, determines if the element
is in the list. If yes, then it returns the element that follows
it in the list. Otherwise NIL is returned as the value of the
function.
EXPR NEXT(L,X);
IF NULL L OR NULL CDR(L) THEN NIL
ELSE IF CAR(L) EQ X THEN CADR(L)
ELSE NEXT(CDR(L),X);
The modification is that the recursive call to NEXT is expanded and
in its stead the body of NEXT is inserted. Thus one is only
performing one recursive call for every two elements in the list.
The modified function is given below:
EXPR NEXT(L,X);
IF NULL L OR NULL CDR(L) THEN NIL
ELSE IF CAR(L) EQ X THEN CADR(L)
ELSE IF NULL CDR(L) OR NULL CDDR(L) THEN NIL
ELSE IF CADR(L) EQ X THEN CADDR(L)
ELSE NEXT(CDDR(L),X);
The LISP 1.6 compiler generated the following code sequence for the
function.
(LAP NEXT SUBR)
(PUSH P 1)
(PUSH P 2)
(JUMPE 1 TAG3)
(HRRZ@ 1 1)
(JUMPN 1 TAG2)
TAG3 (MOVEI 1 (QUOTE NIL))
(JRST 0 TAG1)
TAG2 (HLRZ@ 1 -1 P)
(CAME 1 2)
(JRST 0 TAG4)
(HRRZ@ 1 -1 P)
(HLRZ@ 1 1)
(JRST 0 TAG1)
TAG4 (HRRZ@ 1 -1 P)
(JUMPE 1 TAG6)
(HRRZ@ 1 -1 P)
(HRRZ@ 1 1)
(JUMPN 1 TAG5)
TAG6 (MOVEI 1 (QUOTE NIL))
(JRST 0 TAG1)
TAG5 (HRRZ@ 1 -1 P)
(HLRZ@ 1 1)
(CAME 1 2)
(JRST 0 TAG7)
(HRRZ@ 1 -1 P)
(CALL 1 (E CADR))
(JRST 0 TAG1)
TAG7 (MOVE 2 0 P)
(HRRZ@ 1 -1 P)
(HRRZ@ 1 1)
(CALL 2 (E NEXT))
TAG1 (SUB P (C 0 0 2 2))
(POPJ P)
NIL
A hand optimized version of the above LAP program which takes
advantage of redundant conditions by eliminating their testing is
given below.
(LAP NEXT SUBR)
(JUMPE 1 DONE)
PC2 (MOVE 1 3)
(HRRZ 1 0 1)
(JUMPE 1 DONE)
(HLRZ 4 0 3)
(CAIN 4 2)
(JRST 0 TAG1)
(MOVE 3 1)
(HRRZ 1 0 1)
(JUMPE 1 DONE)
(HLRZ 4 0 3)
(CAIE 4 2)
(JRST 0 PC2)
TAG1 (HLRZ 1 0 1)
DONE (POPJ P)
NIL
The optimized encoding has an inner loop of 11 instructions and a
total length of 15 instructions while the compiler generated
encoding has an inner loop of 23 instructions and a total length of
33 instructions. Thus the optimized encoding is at least twice as
fast. The proof system was able to prove that both of these
encodings were indeed correct. The proof for the compiler
generated encoding took 26 seconds while the proof for the hand
optimized encoding took 27 seconds.
Actually while timing the system on the hand optimized encoding
several errors were detected. First of all, the second instruction
should be (MOVE 3 1) rather than (MOVE 1 3). Furthermore, both of
the compare immediate instructions perform indexing incorrectly.
(CAIN 4 2) and (CAIE 4 2) should be replaced by (CAIN 4 0 2) and
(CAIE 4 0 2) respectively. Thus the system can be said to pay for
itself. The correct encoding is given below:
(LAP NEXT SUBR)
(JUMPE 1 DONE)
PC2 (MOVE 3 1)
(HRRZ 1 0 1)
(JUMPE 1 DONE)
(HLRZ 4 0 3)
(CAIN 4 0 2)
(JRST 0 TAG1)
(MOVE 3 1)
(HRRZ 1 0 1)
(JUMPE 1 DONE)
(HLRZ 4 0 3)
(CAIE 4 0 2)
(JRST 0 PC2)
TAG1 (HLRZ 1 0 1)
DONE (POPJ P)
NIL
Incidentally, this message is stored on HJSJMC.LTR[L,HJS] .
Hanan,
∂24-OCT-75 1624 1,HVA
I have added some notes to memo re David Luckam recomm. for
appointment, and have replaced file folder in left-hand drawer
of table by window. If it's possible for you to produce the
missing parts, I will follow through on Monday.
∂24-OCT-75 1050 S,LES
We are reviewing ISI Tenex accounts with ARPA. Do you wish to continue yours?
∂20-OCT-75 0701 S,LES
Hanan Samet is currently consuming vast quantities of CPU time and is
requesting that his disk allocation be restored to 1 million words.
He claims to be doing something important for you. True?
As you may know, we are rather tight on disk space at present.
∂19-OCT-75 1759 ACT,REG
If you're interested in preserving Nick Littlestone's files over
the NEXT purge, please copy the relevant ones to one of your areas.
∂19-OCT-75 1656 2,DCL
some theorem proving proposals are in AUTDED[ESS,JMC]
∂30-OCT-75 1509 100,100: queenie
Ito called from Pasadena, he is attending USA-Japan Conference, he is coming up
here and would like to meet with you either monday afternoon around 2:30 or
tuesday morning (no specified time). He will call back Monday morning ad
and confirm which day and what time. That's monday(11-3) tuesday(11-4).
∂28-OCT-75 2223 1,FXB @ AMET
Preliminary parts list and prices for the Video Graphics Terminal:
96 2107 dynamic rams 12 3210 drivers 67
1 8080A cpu 150
1 8228 system controller 16
1 8224 clock generator 14
1 8251 usart 26
1 6571l character generator 20
1 5320 sync generator 18
1 5307 baud rate generator 18
4 2708 8K PROMs 6
4 2708 8K PROMs 384
8 4200 4K static rams 136
70 misc TTL 70
1 power supply 100
misc hardware(boards,soc) 200
-------------------------------------
total 1891
Comments?
-Forest
∂28-OCT-75 1503 VCG,DCL
I NEVER EVER LOG IN AS [ESS,JMC] OR PLACE COPIES OF LETTERS
FROM ME ON OTHER PEOPLES AREAS.
AN INTELLIGENT GUESS MIGHT BE THAT THIS IS DUE TO ONE OF THE
SECRETARIES.
HAVE YOU SENT A MSG. TO THEM?
∂22-OCT-75 1130 1,HVA
IS THERE A DECISION ON STAN KUGELL YET - TO EMPLOY (SUPPLEMENTING
YOP FUNDS W/MONEY FROM YOUR UNRESTRICTED) - OR SHOULD I FILE AND
FORGET?
∂22-OCT-75 0517 100,100: AVRA COHN @ LONT
HELLO! JUST DOING SOME NET HACKING. IM FINE AND HAPPY. HOPE YOU ARE
TOO. HI FROM MIKE GORDON AS WELL. ----AVRA
∂20-OCT-75 0940 FTP: host ISI
Date: 20 OCT 1975 0936-PDT
From: RUSSELL at USC-ISI
Subject: Survey articles on AI
To: les at SU-AI
cc: jmc at SU-AI, russell
Les,
I addition to everything else I have asked for, it would be
helpful to have a small number of survey articles or reports which address
various reseqrch problems under investigation at Stanford. In particular
an overview of FOL and the Natural Language work would be helpful.
Highly technical articles full of in-house jargon won't sell to well to
management, hence my suggestion that the focus be on survey or
overview articles suitable for a more general audience.
Will appreciate what you can provide.
Dave
-------
∂19-OCT-75 2000 FTP: host ISI
Date: 19 OCT 1975 1806-PDT
From: RUSSELL at USC-ISI
Subject: BASIC AI PROGRAM
To: JMC at SU-AI
cc: RUSSELL
John,
Thanks for your answer.
I ment the request for one or two lines or bullets to
apply to each individual entry at the fy level.
Dave
-------
∂19-OCT-75 1750 ACT,REG
Despite the purge message you have recieved, none of your files have been
deleted. However, you are exceeding your disk allocation. Please review
your files and delete any that you don't need.
∂19-OCT-75 1019 FTP: host ISI
Date: 19 OCT 1975 1015-PDT
From: RUSSELL at USC-ISI
Subject: OUTLINE OF PLANNED BASIC AI WORK
To: NEWELL at CMU-10A, PHW at MIT-AI, LES at SU-AI, FEIGENBAUM
cc: BLUE, RUSSELL, CARLSTROM, CARLSON
I am constructing an AI program plan that requires a section
devoted to proposed work in some structered format. What
I plan to show is the basic AI program broken into three major
categories:
Expert problem solving
Natural language
General reasoning
I recognize that there is considerable overlap between the
three sections, and one could construct a more accurate division
containing two or three times as many divisions, however, it
will be considerably easier to explain and defend a manageable division
of the program. What I would ask each of you to do is to fill in the
outline that follows with the appropriate information on the work
for which you are responsible. I would prefer one or two sentence
milestone descriptions of your program plans showing a logical
progression of work as you see it now. Obviously, near-term
goals can be described with resaonable accuracy, while longer-term
milestones are more general.
EXPERT PROBLEM SOLVING
MIT
FY76 and FY7T program
FY77 program
FY78-FY80 program
Feigenbaum
FY76 and FY7T program
FY77 program
FY78-FY80 program
CMU
FY76 and FY7T program
FY77 program
FY78-FY80 program
NATURAL LANGUAGE
MIT
FY76 and FY7T program
FY77 program
FY78-FY80 program
Stanford
FY76 and FY7T program
FY77 program
FY78-FY80 program
Yale
FY76 AND FY7T program
FY77 program
FY78-FY80 program
FORMAL REASONING
Stanford
FY76 and FY7T program
FY77 program
FY78-FY80 program
MIT
FY76 and FY7T program
FY77 program
FY78-FY80 program
Please note that vision work and automatic programming work
are now elements of the Image Understanding and Software Technology
programs respectively and are not considered as elements of the basic
AI program. Also please avoid as much as possible jargon and the use
of the first person in your comments. If you think that your work should be
included in other categories than are listed, please feel free
to add your work to appropriate category.
I will provide a copy of the plan for your comments in the
next few days.
I would appreciate your comments by Wednesday, 22 Oct 75.
Dave
-------
∂16-OCT-75 1254 ACT,REG
So, are you going to talk to John Leng on the 24th? Do you want me to be there,
or will a list of questions suffice?
∂16-OCT-75 0234 S,LES
We are slated to have a visitation from the Computer Science Advisory
Committee on Monday morning, Oct. 20. This group of distinguished
visitors provides advice to the Department on various policy questions.
The tentative schedule is
9:00 am Arrival: tour of the Lab., with a few demos
10:00 Brief presentations on research projects by group leaders.
10:45 Coffee & donuts
11:00 More presentations and discussion.
11:45 They head for lunch.
If you have suggestions or scheduling constraints, please let me know.
[I will have to leave about 11:15 to catch a plane.]
∂18-OCT-75 2315 FTP: host SRI
Date: 18 OCT 1975 2314-PDT
From: OMALLEY at SRI-AI
Subject: Bay Area AI Circle Meeting 10/28
To: BAAIC-People:
Members and Friends
Bay Area AI Circle
Greetings
It is time for yet another exciting Bay Area Artificial Intelligence
Circle meeting.
For those of you who are new, the BAAIC is a group of people
interested in all aspects of AI who get together about once a month
for dinner and discussion. While we are an exclusive group, we
welcome anyone interested in AI to join.
This month Earl Sacerdoti of SRI has agreed to talk on his procedural
net representation of knowledge for use in planning and problem
solving. In order to provide balance and entertainment, there will
be a panel of discussants. Unfortunately, the panel members have
asked for anonymity in order to protect themselves from reporters for
the Chronicle and DATAMATION.
The meeting will be held at SRI on Tuesday, October 28. There will be
wine from 6:30 to 7:00, dinner in the SRI international dining room
at 7:00 and the discussion in the SRI auditorium afterwards. Cost
will be $6.50 per person.
Please let us know if you intend to come, either by phone or ARPANET,
before friday, October 24. There will be rides from Berkeley leaving
at 5:15 from Evans Hall.
Mike O'Malley 642-4624 or OMALLEY@SRI-AI
Lotfi Zadeh 642-4959 (auto-answer phone)
A Structure for Plans and Behavior
ABSTRACT
The talk will describe progress that has been made in the
ability of a computer system to understand and reason about actions.
A new method of representing actions within a computer memory has
been developed, and this new representation, called the procedural
←←←←←←←←←←
net, has been employed in developing new strategies for solving
←←←
problems and monitoring the execution of the resulting solutions.
A set of running computer programs, called the NOAH (Nets Of
Action Hierarchies) system, embodies the representation and
strategies mentioned above. Its major goal is to provide a framework
for storing expertise about the actions of a particular task domain,
and to impart that expertise to a human in the cooperative
achievement of nontrivial tasks.
The key ideas that will be explored in the talk include planning
at many levels of detail, representing a plan as a partial ordering
of actions with respect to time, and execution monitoring and error
recovery using hierarchical plans.
-------
∂17-OCT-75 0132 S,LES
Elaine Kant is here now and will finish the Program Understanding part
within an hour or so. I've scanned the first part, which looks good.
I propose to ship it as soon as it is all together.
I forgot to mention that Dave Russell apparently called today and expects
to be out here next week. He plans to be here Thursday afternoon from
about 3 to 5pm.
∂16-OCT-75 0554 S,LES
I just received a message from Russell saying he needs the
Accomplishments as soon as possible. Perhaps we should send what we
have, with whatever polishing can be done straightaway.
∂16-OCT-75 0248 S,LES
Regarding the Bullets: TW says that he has nothing significant to report,
their effort having gone into building foundations for a new system and
in proposal writing. DCL claims that he will emit a blurb any moment now.
I'll push it into the file when it comes. CCG has not been heard from.
I begin to get the impression that he isn't reading his mail.
Did you talk to Floyd about which new students need support? This
question bears on how far we can stretch the budget to cover other
things. There is also the related question of whether we can afford
to continue Martin Frost.
∂15-OCT-75 1312 IBM,DCL
please let me know if you require any further paper work
from me for the adjunct affair. Perhaps i should go over
a checklist of items required with you before it goes
to the next stage (when is that?)
-david
∂15-OCT-75 0031 S,LES
If you would like to look at the ARPA blurbs, they are in BULLET[R,LES].
I still have only the formal reasoning and hand-eye parts.
I have a reservation concerning the latter: it comes on strong for RPV
applications in both the first and third paragraphs but doesn't make
the connection clear. I am not convinced that there is a connection
that is close enough to be worth mentioning.
∂11-OCT-75 2351 FTP: host AI
Date: 12 OCT 1975 0250-EST
From: CARL at MIT-AI
To: jmc at SU-AI
Thanks!
-------
∂11-OCT-75 1554 FTP: host AI
Date: 11 OCT 1975 1852-EST
From: CARL at MIT-AI
To: CARL at MIT-AI, jmc at SU-AI
Hi John,
I sent you a few quibbles about your IJCAI report which in general I liked
very much and think should be published. My only concern was that you not
inadvertantly get some Russians in political hot water. By the way what is a
"refusenik"?
Cheers,
Carl
-------
∂10-OCT-75 1357 AM,DBL
I have read your statement about the 4IJCAI.
It seems to be accurate, and worded quite carefully to convey objectivity.
As for the Goldstein's tape, their only comment that you might want to know
about is their fear of reprisals and increased harassment (by KGB) as soon
as IJCAI ended and everyone went home.
A few tiny suggestions: the sentence about London, Stanford,... IJCAI sites
is hard to parse. The list of conclusions was drenched with technical metaphors,
and is comprehensible only to a scientificly-oriented audience (e.g., a scientific
journal, but NOT the N.Y. Times). You might want to include a few quotes from
Lerner, Goldstein, Chav., Sandewall, or some US scientist who was there.
You might want to give the numbers of people who attented 4IJCAI, broken
into USA/USSR/other, perhaps also broken into student/faculty/professional.
Perhaps we should hold a seminar some afternoon, on our reactions/experiences;
all the local IJCAI-goers could come, plus anyone interested in hearing us.
Merle and I may have some of these people over to our house one evening, to
talk and show slides. If so, you'll certainly be invited.
Cheers,
Doug Lenat
∂10-OCT-75 0943 100,100: sgk
Steve Mcreynolds of JPL robotics wants to talk to people on 7-nov. (will
be in area). 213 354-3784.
∂09-OCT-75 1728 FTP: host AI
Date: 9 OCT 1975 2026-EST
From: PHW at MIT-AI
To: JMC at SU-AI
Hi John,
I have just read your Tblisi report and I find it quite good. I make
only the following suggestions:
You say C said the G's could come if they behaved themselves. This
seems stronger than I supposed. In any event I thought the tone was
different from what those word suggest and I thought his feelings were
more along the lines of extending hospitality and hoping that the recipients
would not trod on that hospitality by turning a scientific meeting into a
political forum.
You also said that the G's were picked up a second time on Sunday. Owing
to their Insistence that Bert not take lunch with them and to the
possible divergence of their interests from
those of the community of dissidents, I personally would be inclined to
insert the word allegedly somewhere before picked up.
Yours,
Patrick
-------
∂09-OCT-75 0825 CAR,HPM
POWELL, Conoley and Hahn, Ottfried J.
"Propulsion System Optimization for Interstellar Probes",
Journal of the British Interplanetary Society, Vol. 26, No. 6,
June, 1973, 334-342.
POWELL, Conoley
"Propulsion System Optimization for A Single Stage Constant Thrust
Relativistic Rocket"
Journal of the British Interplanetary Society, Vol. 26, No. 8,
August, 1973, 466-474.
also of interest - there's a copy in your mailbox
POWELL, Conoley and Hahn, Ottfried J. and McNally, J. Rand, Jr.
"Energy Balance in Fusion Rockets",
Astronautica Acta, Vol 18, No. 1,
February, 1973, 59-69.
also of interest - I'll try to get copies tomorrow
POWELL, Conoley
"Parallel Staging for Starships",
Journal of the British Interplanetary Society, Vol. 27, No. 9,
September, 1974, 686-691.
POWELL, Conoley
"Optimal Exhaust Velocity Programming for an Energy-Limited
Single-Stage Relativistic Rocket",
Journal of the British Interplanetary Society, Vol. 27, No. 4,
April, 1974, 273-285.
ANDERSON, Gerald M.
"Optimal Interstellar Trajectories with Acceleration-Limited
Relativistic Rockets",
Journal of the Astronautical Sciences, Vol. 15, No. 6,
November-December, 1968, 313-318.
ANDERSON, Gerald M.
"Optimal Interstellar Trajectories with Thrust-Limited
Relativistic Rockets",
Journal of the Astronautical Sciences, Vol. 17, No. 2,
September-October, 1969, 65-79.
also of interest - perhaps you'd like to try to get copies
POWELL, Thomas C.
"Fusion Power for Interstellar Flight",
Ph.D. Thesis, Kentucky University - Lexington, Kentucky,
1970, 186 pages, available through NTIS (N71-38275 STAR abstracts).
DUNN, F.J.
"Optimal Interstellar Trajectories with Drag for Thrust-Limited Rockets",
AF Institute of Technology MS Thesis, GAM/MC/70-2, March, 1970.
ANDERSON, Gerald M.
"Relativistic Optimal Rocket Trajectories for Interstellar Travel",
Ph.D. Dissertation, Univ. of Michigan, December, 1966.
ANDERSON, Gerald M.
"Optimal Interstellar Relativistic Rocket Trajectories with both Thrust
and Acceleration Constraints",
Paper presented at 24th International Astronautical Congress,
Baku, U.S.S.R. October 1973.
ANDERSON, Gerald M., and Franklin, Rex W.
"Interstellar Drag and Its Effect on Minimum Time,
Acceleration Limited Relativistic Rocket Trajectories",
Paper presented at 20th International Astronautical Congress,
Mar del Plata, Argentina, October 6-10, 1969.
Selected Papers from 2oth Int. Astronautical Congress,
M. Lunc (Ed-in-Chief), Pergammon Press, New York, 1972, 497-505.
that exhausts Powell, Dunn and what's worthwhile of Anderson
in Forward's bibliography (also me).
∂07-OCT-75 1738 S,LES
To: JMC, RWW, CCG, DCL, TW, TOB
1. It currently looks as though ARPA funding will continue at about the
present level for the next 18 months or so.
2. I have received the following message from Dave Russell (ARPA),
requesting a summary of recent accomplishments. He needs them next
week, so I will need your input by Monday night, October 13.
3. We also owe them an annual report covering the period July '74 to
July '75. I suggest that you fiddle the text of the last proposal.
Absolute final deadline for your part is Wednesday, October 22.
-------
Les,
I need a summary of accomplishments for the
SU-AI program covering the current year. Please consider the
following guidelines:
1. Avoid JARGON
2. Use the third person - i.e., the program has accomplished
rather than we have accomplished
3. Pitch the summary at the scientific American level
4. Where an accomplishment is listed, particularly a
highly technical accomplishment, indicate why this accomplishment
is important.
Dave/hcb
-------
∂08-OCT-75 1632 1,DEW
Did Michie have a computer chess conference in England this summer?
If so, do you know how it went and is their anything worth reading
that came out of it? Thanks, Dave
∂03-OCT-75 1053 FTP: host SRI
Date: 3 OCT 1975 1052-PDT
From: RAPHAEL at SRI-AI
Subject: TBILISI REPORT
To: JMC at SU-AI
cc: PHW at MIT-AI, SACERDOTI, RAPHAEL, LES at SU-AI
John,
Your report of IJCAI political events is excellent, and I think
it should receive wide circulation. There are a few additional items
that I think should be added to it to help complete the picture:
1. Several Western participants in the Conference who were
deeply interested in the plight of Lerner and the Goldsteins were unaware
of the Tuesday morning agreement by the Conference organizing committee.
This perhaps unavoidable weakness in communications among concerned parties
led to later confusion and complications.
2. Soviet and Western observers, whatever their personal feelings,
were well aware of the fact that Lerner behaved is a statesman-like manner
and evidenced serious scientific interest in the proceedings of the
conference, while the Goldsteins evidently were more interested in
calling attention to themselves than in the subject matter of the
conference, thereby weakening their potential support from both sides.
3. Aside from these political distractions, a surprisingly open
East-West interchange took place throughout the conference. Although
the amount of hard technical data exchanged was small, the scope of Eastern
activities in our field was an eye-opener to many Westerners, and the
many personal contacts that were established have significant potential
value in the future for both Science and Detente.
4. On the other hand, the operational problems of dealing with
the clumsy, inefficient, bureaucratic, Soviet state travel organizations--
Intourist, Aeroflot, and the Visa office-- were truly incredible to most
Americans. Any organization considering holding an international
meeting in the Soviet Union should seriously consider whether the
experience, which I still believe to be valuable, is worth the almost
unimaginable amount of extra time and energy that conference participants
will find themselves devoting to what should be simple practical
arrangements.
John, if you do not want to include something like the above in
your letter, I would not mind adding it as a postscript over my signature.
Of course, I would also like to hear Pat's opinion. Also, I still have
in my possession a statement written by the Goldsteins chronicling their
version of events in Tbilisi, and would welcome your and Pat's suggestions
as to what to do with it.
Regards, Bert
-------
∂06-OCT-75 1124 1,TOB
I like very much your letter concerning the negotiations with Soviet
scientists for IJCAI.
Tom
∂30-SEP-75 2157 BPM,BPM
John: Lee Erman was wondering whether you had any "short piece on anything
at all" that could be included in the next SIGART Newsletter. He mentioned
your collection of essays and that maybe you could pick one that is already
in fair shape and that would be appropriate to the SIGART readership.
∂30-SEP-75 2257 BPM,BPM
To: JMC, LES, CCG
Pat Suppes and Bob Smith have OKed use of INTERLISP at IMSSS except for
9 am to 9 pm weekdays. For FREE!
∂29-SEP-75 1637 network site AI
Date: 29 SEP 1975 1937-EST
From: RWG at MIT-AI
To: jmc at SU-AI
moses says ok if at most two consoles at once, and after 7pm pdt.
jeff golden (jpg%ml) has teaching material and intimate knowledge of how
novices confuse it. i will speak to him when he returns on 1 oct. ellen lewis
has written some introductory stuff which, on very brief inspection, looked
excellent. i will purloin a few copies. see you 6oct.
-------
∂23-SEP-75 2004 1,RWG @ AI
do you want me to ask moses about macsyma time? if so, just what?
∂23-SEP-75 1157 network site SRI
Date: 23 SEP 1975 1154-PDT
From: BOYER at SRI-AI
Subject: TOPOR'S VISIT
To: PROGRAM REASONING GROUP:
RODNEY HAS ARRIVED AND WILL BE IN THE AREA FROM TODAY THROUGH FRIDAY.
HIS TALK IS STILL SCHEDULED FOR 4:00 ON FRIDAY.
-------
∂22-SEP-75 1241 1,PAW
there is a faculty meeting scheduled for Friday, Sept 26 at 9 am in polya 204 to
discuss Luckham
∂22-SEP-75 1231 network site AI
Date: 22 SEP 1975 1531-EST
From: CARL at MIT-AI
To: CARL at MIT-AI, jmc at SU-AI
Dear John,
I have read your report and in general like it very much.
However, I do have a couple of small quibbles.
First it is not clear that it is fair to report that the Soviet scientists
personally object to the presence of Lerner as chairman of the panel.
These conversations took place in private where the Russians thought that they
could speak privately and frankly. Your report may be a breach of this confidence
even though Lerner somehow got wind of it and put it in one of his reports.
Your report would be official confirmation.
Dr. Chavchanidze is singled out to play a prominent role in your report.
Your seem to give him personal responsibility for the immediate decision to
admit Lerner and to do so with great courtesy. The meeting the morning before
the conference opened was private and may have been considered private by some
of the participants. In any case singling out Chavchanidze in this way could
have unfavorable repercussions for him. Perhaps some reference to the "local organizers"
would be more appropriate. I assume that you have checked your comments about the
Goldstein brothers carefully. Your account differs somewhat with one other
fragmentary report that I received. Again you single out Chavchanidze for special
responsibility. Do you think that he might consider his remarks on this matter to
have been private? This concludes my list of quibbles.
Carl
-------
∂22-SEP-75 1144 network site SRI
Date: 22 SEP 1975 1140-PDT
From: BOYER at SRI-AI
To: PROGRAM-REASONING-GROUP:
SEMINAR
BY
RODNEY TOPOR
PROGRAM VERIFICATION BY CONTINUATION INDUCTION
FRIDAY, SEPTEMBER 26
4:00 P.M.
S.R.I.
INTERNATIONAL BUILDING
ROOM S103 (UEMURA ROOM)
RODNEY TOPOR RECENTLY OBTAINED A PH.D FROM THE UNIVERSITY
OF EDINBURGH WORKING UNDER ROD BURSTALL. HE WILL BE VISITING
THIS AREA FROM SEPTEMBER 25 UNTIL SEPTEMBER 30. HE THEN LEAVES
TO TAKE UP AN ACADEMIC POST IN HIS NATIVE LAND, AUSTRALIA.
CONTINUATION INDUCTION IS BASED ON THE IDEAS OF SYMBOLIC EXECUTION,
THE DESCRIPTION OF A GIVEN PROGRAM BY A VIRTUAL PROGRAM, AND THE
DEMONSTRATION THAT THESE TWO PROGRAMS ARE EQUIVALENT WHENEVER THE
GIVEN PROGRAM TERMINATES. THE MAIN ADVANTAGE OF CONTINUATION INDUCTION
OVER OTHER METHODS IS THAT IT ENABLES PROGRAMS USING A WIDE
VARIETY OF PROGRAMMING CONSTRUCTS SUCH AS RECURSION, INTERATION,
NON-DETERMINISM, PROCEDURES WITH SIDE-EFFECTS AND JUMPS OUT OF
BLOCKS TO BE HANDLED IN A NATURAL AND UNIFORM WAY.
-------
∂22-SEP-75 0859 network site SRI
Date: 22 SEP 1975 0855-PDT
From: BOYER at SRI-AI
Subject: A TALK TO BE GIVEN BY RONEY TOPOR AT SRI
To: WEYHRAUCH at SU-AI, MCCARTHY at SU-AI
WOULD YOU PLEASE ALERT EVERYONE AT SU-AI WHO MIGHT BE INTERESTED
TO THE FOLLOWING:
SEMINAR
BY
RODNEY TOPOR
PROGRAM VERIFICATION BY CONTINUATION INDUCTION
FRIDAY, SEPTEMBER 26
4:00 P.M.
S.R.I.
INTERNATIONAL BUILDING
ROOM S103 (UEMURA ROOM)
RODNEY TOPOR RECENTLY OBTAINED A PH.D FROM THE UNIVERSITY
OF EDINBURGH WORKING UNDER ROD BURSTALL. HE WILL BE VISITING
THIS AREA FROM SEPTEMBER 25 UNTIL SEPTEMBER 30. HE THEN LEAVES
TO TAKE UP AN ACADEMIC POST IN HIS NATIVE LAND, AUSTRALIA.
CONTINUATION INDUCTION IS BASED ON THE IDEAS OF SYMBOLIC EXECUTION,
THE DESCRIPTION OF A GIVEN PROGRAM BY A VIRTUAL PROGRAM, AND THE
DEMONSTRATION THAT THESE TWO PROGRAMS ARE EQUIVALENT WHENEVER THE
GIVEN PROGRAM TERMINATES. THE MAIN ADVANTAGE OF CONTINUATION INDUCTION
OVER OTHER METHODS IS THAT IT ENABLES PROGRAMS USING A WIDE
VARIETY OF PROGRAMMING CONSTRUCTS SUCH AS RECURSION, INTERATION,
NON-DETERMINISM, PROCEDURES WITH SIDE-EFFECTS AND JUMPS OUT OF
BLOCKS TO BE HANDLED IN A NATURAL AND UNIFORM WAY.
THANKS.
-------
∂21-SEP-75 1914 1,RWG @ AI
anyway, i am having a good time out here, but will come back
before my excursion fare expires if you need me.
∂19-SEP-75 1347 PAP,REF @ CMUA
You're welcome. How was russia?
∂17-SEP-75 1218 BPM,BPM
John: You might be interested in a report at the CS Library entitled "An Overview
of the LBL Socio Economic Environmental Demographic Information System
(SEEDIS)" by D. M. Austin, S. G. Kranz, and C. Quong. Is this the system you
were using the other day?
∂09-SEP-75 1555 100,100: patte wood @ SAIL
a meeting at MIT Study of Future of Computing has been set up for
Monday, Sept 22 from 9-1
∂15-SEP-75 1656 FOL,RWW
the version of FOL on the system now contains the printing routines you
requested. if there are any problems let me know.
rwww
∂12-SEP-75 1633 S,LES
Just received a call from a fellow at the Hoover Inst. named Starr, I believe.
He said that they have a prospective donation of 9 bookshelves full of Soviet
literature on computers, cybernetics, etc. and said that you had talked about
storing it here somewhere. I said that we do indeed have space, but no
bookshelves. He plans to call back next Wednesday (Sept. 17) for confirmation
of space availability.
∂12-SEP-75 1624 100,100: patte @ SAIL
Dr. Hileman called and cancelled Sept 24 at 4:00, he will call again after Oct 8
and reschedule.
∂09-SEP-75 0942 network site DMS
DATE: 9 SEP 75 1053-EDT
FROM: LICK at MIT-DMS
ACTION-TO: McCarthy at SU-AI
CC: Lick at MIT-DMS, Carlstrom at USC-ISI
MESSAGE-ID: <[MIT-DMS]9 SEP 75 12:24:24-EDT.20432>
Very sorry, John, that I got away from ARPA without
communicating adequately with you about your revised
proposal. Yesterday I talked with Dave Carlstrom about it,
offering to do anything I could to facilitate matters. He
said that Dave Russell was working (on) it, correlating the
several parts (related to different program managers in
IPTO), but that Dave was on leave this week. You will
probably hear from one of the Daves next week. If you do
not, I suggest you call Carlstrom, who has the largest parts
of the action. I'll keep in touch with him and Russell, but
the matter is in their hands, now, and I do not want to
meddle much in the new management of the office, however
much interested I remain.
Hope the AI Conference came off all right.
Regards
Lick
∂27-AUG-75 0517 network site ISI
Date: 27 AUG 1975 0517-PDT
From: LICKLIDER at USC-ISI
Subject: Reaction Cubed
To: McCarthy at SU-AI
cc: Licklider
By all means let's talk on the phone.
The problem is not whether DoD-supported AI should have a
basic research core. It is whether some of the support of basic AI
research (all the support was for basic research, in the
view of Lukasik and Heilmeier) should be diverted to make room for
efforts in application and technology transfer. Your putting it
as you did (defending the retention of SOME basic research) just
underscores the fact that the administrators and the researchers see
things in such different ways that they are extremely difficult to
mediate.
As for the importance of formal reasoning: There is no question
that it is important for the development of AI broadly. The question
is whether a large amount of it is what is needed to solve the
DoD problems that AI has some chance of solving in the near term. The
whole situation is to be understood in terms of DoD's wanting to
assess the capability oo the field to solve some of DoD's problems now.
Almost everyone who has thought about the matter understands that it
is penny-wise but pound-foolish to go overboard on
pressing for immediate or near-term applications, that the real
importance of AI to DoD lies in the longer term possibility that ther
may in due course be an advance in thinking to match the
a weapons delivery and in explosive power. But the pressure for
near term applications is nevertheless a fact. Not for 100 percent shift to
applications, but for 30-40 percent.
Finally (for this note), this is a bad time for you to be
thinking (or talking) about ceasing to administer the lab. The
reaction would be, indeed, something like "if the leaders of the field
don't have any more committment to it than that, it is too risky for
DoD to put so much money into".
My own assessment of the situation is that now is a time in
hry that needs the best contributions of a lot of people, including
you, and that the situation might be a lot stronger in a couple of
years. The main chance is to bring NSF into the support of AI in a
good way. Please do not mess up the ARPA community just as we get
started on fashioning some kind of arrangement with NSF. But let us
talk on the phone.
Regards
Lick
-------
∂26-AUG-75 2300 PAP,SUZ
I AM SORRY THAT I COULD NOT SEE YOU AT THE LAB SO THAT I CAN
SAY GOOD BY TO YOU. I WILL LEAVE FOR CARNEGIE TOMORROW. I APPRECIATE
YOUR SUPPORT AND HELP DURING THREE YEARS OF STAY AT STANFORD. I WILL
COME BACK HERE TIME TO TIME, BUT BEST WISHES. NORI