perm filename S83.OUT[LET,JMC]1 blob
sn#717184 filedate 1983-07-02 generic text, type C, neo UTF8
COMMENT ⊗ VALID 00198 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00016 00002 ∂03-Apr-83 1502 JMC
C00017 00003 ∂03-Apr-83 1515 JMC elephant
C00018 00004 ∂03-Apr-83 1930 JMC
C00027 00005 ∂04-Apr-83 0140 JMC kirkla.1
C00028 00006 ∂10-Apr-83 1527 JMC
C00029 00007 ∂11-Apr-83 1224 JMC
C00030 00008 ∂11-Apr-83 1229 JMC
C00031 00009 ∂12-Apr-83 0953 JMC
C00032 00010 ∂12-Apr-83 1437 JMC via Ethernet SU-SCORE Common Lisp
C00034 00011 ∂13-Apr-83 1237 JMC
C00035 00012 ∂13-Apr-83 1509 JMC Mark Todorovich
C00036 00013 ∂13-Apr-83 1554 JMC honor yes, advantage no
C00037 00014 ∂13-Apr-83 1624 JMC
C00038 00015 ∂13-Apr-83 1640 JMC
C00039 00016 ∂13-Apr-83 2300 JMC
C00040 00017 ∂13-Apr-83 2301 JMC
C00041 00018 ∂13-Apr-83 2302 JMC
C00042 00019 ∂14-Apr-83 1452 JMC
C00043 00020 ∂14-Apr-83 1700 JMC
C00044 00021 ∂16-Apr-83 1540 JMC visitors on Monday
C00045 00022 ∂16-Apr-83 1540 JMC Joe Weening
C00046 00023 ∂17-Apr-83 1310 JMC AI qual
C00047 00024 ∂18-Apr-83 0020 JMC dinner on Wednesday?
C00048 00025 ∂18-Apr-83 1640 JMC
C00049 00026 ∂18-Apr-83 2245 JMC
C00050 00027 ∂19-Apr-83 1259 JMC
C00051 00028 ∂21-Apr-83 2137 JMC
C00052 00029 ∂21-Apr-83 2353 JMC IPTO support of S-1
C00054 00030 ∂22-Apr-83 1137 JMC
C00056 00031 ∂22-Apr-83 1158 JMC new account
C00057 00032 ∂22-Apr-83 1158 JMC
C00058 00033 ∂22-Apr-83 1216 JMC TIP phone numbers
C00059 00034 ∂22-Apr-83 1224 JMC Omni
C00060 00035 ∂22-Apr-83 1232 JMC
C00061 00036 ∂22-Apr-83 1534 JMC
C00063 00037 ∂22-Apr-83 1537 JMC
C00064 00038 ∂22-Apr-83 1538 JMC
C00065 00039 ∂22-Apr-83 1540 JMC
C00066 00040 ∂22-Apr-83 1550 JMC user in NYC
C00067 00041 ∂22-Apr-83 2211 JMC
C00068 00042 ∂23-Apr-83 2354 JMC
C00069 00043 ∂25-Apr-83 1152 JMC
C00073 00044 ∂25-Apr-83 2210 JMC
C00074 00045 ∂25-Apr-83 2301 JMC
C00075 00046 ∂27-Apr-83 0257 JMC metaphors
C00076 00047 ∂27-Apr-83 1650 JMC
C00077 00048 ∂27-Apr-83 2100 JMC
C00078 00049 ∂28-Apr-83 1046 JMC
C00079 00050 ∂28-Apr-83 1540 JMC skyhooks
C00080 00051 ∂29-Apr-83 0041 JMC article from library
C00081 00052 ∂29-Apr-83 1510 JMC
C00082 00053 ∂29-Apr-83 1612 JMC
C00083 00054 ∂29-Apr-83 1613 JMC
C00084 00055 ∂29-Apr-83 1721 JMC
C00085 00056 ∂29-Apr-83 1722 JMC dinner?
C00086 00057 ∂30-Apr-83 1143 JMC
C00087 00058 ∂30-Apr-83 1156 JMC
C00088 00059 ∂30-Apr-83 1310 JMC correcting misinformation about WAITS
C00090 00060 ∂30-Apr-83 1611 JMC
C00091 00061 ∂01-May-83 1200 JMC temporary mail routing
C00092 00062 ∂01-May-83 1814 JMC
C00093 00063 ∂02-May-83 1048 JMC
C00094 00064 ∂02-May-83 1545 JMC speak of the devil
C00095 00065 ∂02-May-83 1559 JMC abstract
C00096 00066 ∂02-May-83 1603 JMC MACSYMA
C00097 00067 ∂02-May-83 1658 JMC C.I.T. and lunch
C00098 00068 ∂02-May-83 1948 JMC Abstract for talk "How Flavors Differ from Smalltalk"
C00099 00069 ∂02-May-83 2042 JMC
C00100 00070 ∂02-May-83 2058 JMC
C00101 00071 ∂03-May-83 1148 JMC
C00102 00072 ∂03-May-83 1433 JMC correcting
C00104 00073 ∂03-May-83 1434 JMC
C00105 00074 ∂03-May-83 1600 JMC
C00106 00075 ∂03-May-83 1612 JMC
C00107 00076 ∂04-May-83 0137 JMC Keyworth
C00108 00077 ∂04-May-83 1136 JMC
C00111 00078 ∂04-May-83 2331 JMC
C00112 00079 ∂04-May-83 2338 JMC
C00113 00080 ∂04-May-83 2345 JMC presentation
C00114 00081 ∂05-May-83 0023 JMC temporary mail routing
C00116 00082 ∂05-May-83 0037 JMC
C00117 00083 ∂05-May-83 0050 JMC
C00119 00084 ∂05-May-83 0110 JMC metaphor
C00121 00085 ∂05-May-83 1041 JMC
C00123 00086 ∂05-May-83 1050 JMC
C00124 00087 ∂05-May-83 1347 JMC
C00125 00088 ∂05-May-83 1422 JMC
C00126 00089 ∂05-May-83 1423 JMC
C00127 00090 ∂05-May-83 1742 JMC
C00129 00091 ∂06-May-83 1153 JMC natural kinds
C00133 00092 ∂06-May-83 1154 JMC my mail address
C00134 00093 ∂06-May-83 1331 JMC failed mail return
C00135 00094 ∂06-May-83 1336 JMC
C00136 00095 ∂06-May-83 1342 JMC
C00137 00096 ∂06-May-83 1442 JMC seminar
C00138 00097 ∂06-May-83 1714 JMC
C00139 00098 ∂07-May-83 1014 JMC visit and seminar
C00140 00099 ∂07-May-83 1059 JMC obscurity
C00141 00100 ∂07-May-83 1317 JMC
C00142 00101 ∂07-May-83 1831 JMC
C00143 00102 ∂07-May-83 1837 JMC circumscription and satisfaction
C00144 00103 ∂07-May-83 2342 JMC natural kinds
C00146 00104 ∂08-May-83 1041 JMC mailing to bulletin boards
C00147 00105 ∂08-May-83 1538 JMC
C00150 00106 ∂08-May-83 2108 JMC visit
C00151 00107 ∂08-May-83 2231 JMC natural kinds
C00153 00108 ∂08-May-83 2348 JMC visit
C00154 00109 ∂08-May-83 2351 JMC visit
C00156 00110 ∂09-May-83 0917 JMC message from Ohlander
C00157 00111 ∂09-May-83 0937 JMC dates for Keyworth
C00158 00112 ∂09-May-83 0938 JMC dates for Keyworth
C00159 00113 ∂09-May-83 1322 JMC Keyworth
C00161 00114 ∂09-May-83 1422 JMC
C00162 00115 ∂09-May-83 1621 JMC
C00163 00116 ∂09-May-83 2157 JMC <ctrl>z
C00164 00117 ∂09-May-83 2237 JMC
C00165 00118 ∂09-May-83 2312 JMC
C00169 00119 ∂10-May-83 1044 JMC paradigms
C00171 00120 ∂10-May-83 1335 JMC
C00172 00121 ∂10-May-83 1601 JMC
C00173 00122 ∂10-May-83 1602 JMC
C00174 00123 ∂10-May-83 1611 JMC
C00175 00124 ∂10-May-83 1655 JMC DARPA scope document
C00177 00125 ∂10-May-83 1705 JMC scope
C00178 00126 ∂10-May-83 1754 JMC
C00179 00127 ∂10-May-83 1815 JMC
C00180 00128 ∂10-May-83 1906 JMC revised scope information.
C00196 00129 ∂10-May-83 2201 JMC
C00197 00130 ∂11-May-83 0015 JMC
C00198 00131 ∂11-May-83 0019 JMC Classical duets, anyone?
C00200 00132 ∂11-May-83 0031 JMC letter to Col. Gordon
C00201 00133 ∂11-May-83 0717 JMC
C00202 00134 ∂13-May-83 1639 JMC
C00203 00135 ∂13-May-83 1652 JMC via mail rms
C00204 00136 ∂13-May-83 1739 JMC
C00205 00137 ∂16-May-83 1038 JMC
C00206 00138 ∂16-May-83 1109 JMC line inoperative
C00207 00139 ∂17-May-83 1847 JMC Data types and natural kinds
C00209 00140 ∂17-May-83 1940 JMC
C00210 00141 ∂17-May-83 2008 JMC
C00211 00142 ∂17-May-83 2142 JMC Meeting
C00212 00143 ∂17-May-83 2224 JMC common.msg
C00213 00144 I just got back from the Far East. Here is the message referred to.
C00219 00145 ∂18-May-83 0855 JMC SOW
C00220 00146 ∂18-May-83 0939 JMC role of logic in AI
C00221 00147 ∂18-May-83 1224 JMC Keyworth is out
C00222 00148 ∂18-May-83 1307 JMC
C00223 00149 ∂18-May-83 2138 JMC
C00224 00150 ∂18-May-83 2248 JMC
C00236 00151 ∂18-May-83 2311 JMC Keyworth alternative
C00237 00152 ∂19-May-83 0023 JMC Second Choice
C00239 00153 ∂19-May-83 0140 JMC letter from Seitz
C00240 00154 ∂19-May-83 1026 JMC
C00241 00155 ∂19-May-83 1337 JMC representing sequences by sets
C00245 00156 ∂19-May-83 1718 JMC data types and natural kinds
C00247 00157 ∂19-May-83 1720 JMC
C00249 00158 ∂19-May-83 2008 JMC no learning systems yet?
C00250 00159 ∂20-May-83 0116 JMC reducing number of Altos
C00251 00160 ∂20-May-83 0121 JMC
C00252 00161 ∂20-May-83 1143 JMC SAIL hardware schedule
C00253 00162 ∂20-May-83 1157 JMC
C00254 00163 ∂20-May-83 1147 JMC sail hardware schedule
C00255 00164 ∂20-May-83 1403 JMC
C00256 00165 ∂20-May-83 1652 JMC phone conversation
C00257 00166 ∂20-May-83 1655 JMC
C00258 00167 ∂20-May-83 1655 JMC
C00259 00168 ∂21-May-83 0153 JMC (→15677 13-Jun-83)
C00260 00169 ∂13-Jun-83 0000 JMC Expired plan
C00261 00170 ∂16-Jun-83 1232 JMC Facilities committee
C00262 00171 ∂16-Jun-83 1237 JMC
C00263 00172 ∂16-Jun-83 1240 JMC
C00264 00173 ∂16-Jun-83 1241 JMC Re: role of logic in AI
C00265 00174 ∂16-Jun-83 1245 JMC
C00266 00175 ∂16-Jun-83 1246 JMC prolog
C00268 00176 ∂16-Jun-83 1303 JMC
C00269 00177 ∂16-Jun-83 1950 JMC
C00270 00178 ∂16-Jun-83 2027 JMC
C00271 00179 ∂17-Jun-83 2326 JMC
C00272 00180 ∂17-Jun-83 2339 JMC
C00278 00181 ∂17-Jun-83 2348 JMC
C00279 00182 ∂18-Jun-83 0000 JMC
C00280 00183 ∂18-Jun-83 0046 JMC AAAI speaker
C00281 00184 ∂18-Jun-83 0114 JMC
C00282 00185 ∂18-Jun-83 0117 JMC
C00283 00186 ∂18-Jun-83 1103 JMC circumscription
C00284 00187 ∂18-Jun-83 1538 JMC
C00285 00188 ∂18-Jun-83 2153 JMC Please send reprint
C00286 00189 ∂19-Jun-83 0741 JMC Ramsey to Sato
C00287 00190 ∂19-Jun-83 1220 JMC Salamin co-ordinates
C00288 00191 ∂19-Jun-83 1223 JMC correctness of program for computing pi or e
C00289 00192 ∂19-Jun-83 1418 JMC
C00290 00193 ∂19-Jun-83 1703 JMC
C00292 00194 ∂19-Jun-83 2334 JMC lisp on perq
C00293 00195 ∂20-Jun-83 0207 JMC Universality of first order logic
C00303 00196 ∂20-Jun-83 0208 JMC
C00304 00197 ∂20-Jun-83 0215 JMC
C00305 00198 ∂20-Jun-83 1302 JMC
C00334 ENDMK
C⊗;
∂03-Apr-83 1502 JMC
To: CLT
UNDATED - Rain dampens the Easter Parade in Manhattan and snow hides
the spring flowers in much of the country while Christians stream to
church in observance of their most sacred holiday.
∂03-Apr-83 1515 JMC elephant
To: KLC
"He meant what he said, and he said what he meant, for an elephant is
faithful one hundred percent".
or
"He said what he meant, and he meant what he said, for an elephant is
faithful until he is dead".
∂03-Apr-83 1930 JMC
To: LLW@SU-AI
After writing the last paragraph of this, I thought of the
Hertz fellowships. See the last sentence.
eugeni[s83,jmc] What if the eugenicists were right?
Around the turn of the century, the eugenics movement came
into existence in Britain. The intellectual basis for the movement
was (1) the Darwin theory of natural selection, (2) the ability to
measure intelligence (3) the ideas of deliberate social policy.
The movement quickly came to believe that less intelligent members
of society were reproducing more rapidly than the more intelligent
and that this would eventually lead to a decline in intelligence.
They proposed various measures to reverse this trend. To my knowledge
none were adopted.
The eugenicists were mainly elitists, and after many years,
they died without recruiting enough younger people to sustain a
movement. The reasons for this failure was probably the increasing
dominance of equalitarian ideas. To equalitarians, the idea that
dominance in human society was associated with inherited qualities
was distasteful. The main means of refuting the eugenicists and
other hereditarians was to require very high standards of proof -
to regard the hypothesis that there were no such inherited qualities
as established unless overthrown by arguments that had no answers.
Now we discover that school scores are declining and have
been declining for many years. There are various scapegoats.
1. Really determined equalitarians say that the tests measure
the wrong thing and should be abolished.
2. The educational establishment claims that not enough
money is being spent, ignoring the fact that education consumes
a larger fraction of the GNP than when the scores were higher.
3. Another popular suggestion is that talented people are
being attracted away from education, especially education in science,
by high industrial salaries. However, verbal aptitude scores have
declined even faster than mathematical.
4. It is also stated that the reforms that responded to
the agitation of the 1960s were harmful.
5. We propose consideration of the eugenics hypothesis.
The lower fertility of the intelligent has finally taken its toll.
The effect may have been masked for many decades by overcoming
barriers to education in the lower classes of society.
i.e. the low fertility of the existing people in intellectual
occupations was overcome by picking new recruits out of the working
class. Once picked out, these people were subject to the same
influences that lowered the fertility of their predecessors.
Now there are only slim pickings left in the non-intellectual
classes in America. Presumably there is still a big gene pool
for intelligence in the immigrants from areas where educational
opportunity was not available to the lower classes or where the
upper classes had high fertility. Thus we may expect, and indeed
see, intellectual contributions from Vietnamese refugees and
other immigrants far out of proportion to their number.
I don't believe explanations 1 and 2, but I
incline to think that explanations 3, 4, and 5 all contain part
of the truth. Therefore, the following remedies are suggested.
1. Devise methods of teaching bright students with less
manpower. Better books, better computer programs, concentration
of able students, early admission to colleges.
2. Reverse the reforms of the 60s in so far as these lessen
demands on the students to study. I also believe that the "new mathematics"
was a mistake, somewhat related to the ideology of the 60s, i.e.
the ideology of superficial relevance. When Euclidean plane geometry
was de-emphasized, it wasn't replaced by anything that gave comparable
opportunity for developing mathematical creativity. Example: My high
school geometry book included a proof of Pythagoras's theorem attributed
to President Garfield (who apparently wasn't the first to discover it).
Garfield was presumably a lawyer and politician by inclination, so
his exposure to mathematics that would excite his creativity must
have occurred in school. New math contains nothing comparable so
far as I know.
3. Adopt some measures to increase the fertility of the
intelligent. Shockley proposes measures to reduce the fertility
of the unintelligent, and they may have some virtues. However,
he is accused of having bad motives, and there seems to be no
likelihood that this ideology could be overcome sufficiently
overwhelmingly to permit adoption of any of his suggestions
by the Government, and they all require action at the State or
Federal level. Actions to increase the fertility of the intelligent
might be taken at the government level, but this is not likely soon.
However, there is much that can be done at lower levels. Here are
some suggestions.
a. Propaganda in favor of the intelligent having more children.
Since having children is an individual decision, this is the
single most important measure. Articles and books in the right
media are important. It is also important to counteract the
propaganda in favor of family limitation in so far as it affects
this group.
b. Measures to help people in prolonged educational processes
have children. For example, universities could improve their support
of housing for graduate students. The stipends could contain
allowances for children.
c. Direct subsidies of children. Every kind of privately
offered fellowship should contain allowances for support of children.
Where eugenic arguments are unacceptable, equity arguments will serve.
∂04-Apr-83 0140 JMC kirkla.1
To: DFH
Please xgp kirkla.xgp[let,jmc] and send it. Tom Donahue's address
is the same as Kirkland's, and Chudnovsky's is in PHON.
∂10-Apr-83 1527 JMC
To: DFH
Please inform Suppes's secretary that I'll miss meeting.
∂11-Apr-83 1224 JMC
To: GIO@SU-AI
proposed title is fine.
∂11-Apr-83 1229 JMC
To: DEK@SU-AI, DKE@SU-AI
CC: RPG@SU-AI, GIO@SU-AI
I''m for including you.
∂12-Apr-83 0953 JMC
To: RPG
tomorrow
∂12-Apr-83 1437 JMC via Ethernet SU-SCORE Common Lisp
To: RPG
There was a discussion at the PI's meeting about it that went
rather well. There was some sentiment, initiated by Feigenbaum
citing Gordon Bell, that it would soon be time
to discuss ANSIizing Common Lisp. It was agreed that this was
probably premature and anyway depended on the Common Lisp group.
Ron Ohlander hoped the group would meet again in the not too
distant future, and would like to know if there is any present
plan to meet. Is there? There is favorable sentiment towards
Common Lisp even from people whose present use emphasizes
Interlisp. The only doubts about DARPA's intention to
emphasize Common Lisp were expressed by Tony Hearn. There was
general agreement with DARPA's present intention to
support directly financially only the 68000 implementation.
The blue and yellow pages ideas received nods but probably not
very widespread understanding. The discussion was regarded as
successful by all.
∂13-Apr-83 1237 JMC
To: CLT
I'm back.
∂13-Apr-83 1509 JMC Mark Todorovich
To: llw@S1-A
Mark Todorovich, son of Miro Todorovich of SE2, is a senior at Caltech
and is interested in large parallel computers. He is coming up Sunday
May 8, and I'd like to bring him to S-1 on Monday. I think he might
be interested and suitable for the S-1 Project.
∂13-Apr-83 1554 JMC honor yes, advantage no
To: csd.bscott@SU-SCORE
I have been elected President of the American Association for Artificial
Intelligence. They may already know.
∂13-Apr-83 1624 JMC
To: DFH
rosens.re1
∂13-Apr-83 1640 JMC
To: DFH
pearl.1
∂13-Apr-83 2300 JMC
May I have bboard for a moment?
∂13-Apr-83 2301 JMC
To: DON%SU-AI@USC-ECL
May I have bboard for a moment?
∂13-Apr-83 2302 JMC
To: DON%SU-AI@USC-ECL
Done, thanks.
∂14-Apr-83 1452 JMC
To: nan@SU-SCORE
Thanks.
∂14-Apr-83 1700 JMC
To: CLT
I'm home.
∂16-Apr-83 1540 JMC visitors on Monday
To: llw@S1-A
Joe Weening and Mark Todorovich are coming with me and Dick Gabriel on Monday.
Tom McWilliams will take care of Mark who is interested in S-1 class
computers and is a senior at Caltech who might be a potential employee.
I have suggested that Jeff Rubin talk to Joe about system performance and
its measurement. Dick Gabriel and I will be glad to talk about LISP
multi-processing if you are available, otherwise we'll talk to each other
about it.
∂16-Apr-83 1540 JMC Joe Weening
To: jbr@S1-A
Joe Weening is one of my graduate students. He is interested in a possible
thesis on performance of time-sharing systems and Lowell has offered
him a summer job. I think it would be worthwhile for you to talk to him.
He should also talk to the Amber people.
∂17-Apr-83 1310 JMC AI qual
To: su-bboards@SU-AI, SID@SU-AI
So far two people have signed up for the AI qual this Spring. If no
more sign up in two weeks, I will negotiate a convenient date with
these two and the potential examiners.
∂18-Apr-83 0020 JMC dinner on Wednesday?
To: kahn%USC-ISI@USC-ECL
I will be in Washington Wednesday for Army conference on AI,
and I would like to suggest various projects if you have time
to discuss them. These are more things that should be done
rather than things I would propose to undertake myself.
Thursday daytime is also possible, but I remember you said
dinner on Wednesday was more likely to be feasible.
∂18-Apr-83 1640 JMC
To: ALS
The Dartmouth Summer Research Project on Artificial Intelligence occupied
the summer of 1956. However, only a few people were there the whole
time - perhaps Minsky and me, perhaps only me. Some came for only a few
days. I can't imagine any way
to get a complete list. Here is a partial list, and you might try to supplement
it by asking Minsky.
John McCarthy
Marvin Minsky
Oliver Selfridge
Nathaniel Rochester
Claude Shannon
Raymond Solomonoff
Julian Bigelow
Arthur Samuel
Alex Bernstein
Allen Newell
Herbert Simon
∂18-Apr-83 2245 JMC
To: kahn%USC-ISI@USC-ECL
I would be glad to have Ron too.
∂19-Apr-83 1259 JMC
To: DFH
I took your last batch of blank transparencies.
∂21-Apr-83 2137 JMC
To: ALS
You are right about Gelernter and Selfridge, but I am pretty sure
that neither Church nor Kleene was there. I my state of mind at that
time, I wouldn't have invited either.
∂21-Apr-83 2353 JMC IPTO support of S-1
To: LLW@SU-AI
I had dinner with Bob Kahn last night, and he brought up that he wanted
to support S-1 but on a smaller scale than the Navy and DoE and wanted
to make a distinctive contribution. Since we were talking about LISP
I suggested that he support making MACSYMA available in Common Lisp
for the S-1, and he liked that. Trying to phone you I got Mike
Farmwald who remarked that that would be nice but it might be rather
trivial in cost, because Symbolics was putting Macsyma into Zetalisp
which is very close. Perhaps Bob Kahn deserves the honor of a larger
contribution and I trust that my off-hand remarks won't deprive him
of that honor.
∂22-Apr-83 1137 JMC
To: ullman@SU-HNV
.arpa,feigenbaum%sumex,rindfleisch%sumex,
csl.jlh%score,rpg,reg,pattermann%sumex,ohlander%isi/cc,pattermann%sumex/cc
I am quite unhappy about having to spend all the money at once, because
I think we don't have enough experience with the work station philosophy
and with the particular proposed work stations - whether LISP machines,
SUNs or VAXes. Also neither the 2080 nor the S-1 are ready to be
bought. Also we don't have enough experience with expanded file
size to support or refute my contention that we will require very
much larger file sizes than presently contemplated.
∂22-Apr-83 1158 JMC new account
To: gotelli@SU-SCORE
CC: DFH@SU-AI, DAC@SU-AI
I have created an account DAC for David Chudnovsky. It is for messages,etc.
and not for major computing. Its costs should be paid from one of my
unrestricted accounts.
Chudnovsky, David and Gregory (Chris is Gregory's wife)
423 West 120th St., Apt. 88
NY 10027
864-5320
∂22-Apr-83 1158 JMC
To: DFH
CC: DAC
Please send David Chudnovsky Short Waits and Essential E.
∂22-Apr-83 1216 JMC TIP phone numbers
To: admin.crispin@SU-SCORE
How does one get them? In particular I am interested in one in
New York City for someone to whom I have given an account on SAIL.
∂22-Apr-83 1224 JMC Omni
To: NCR@SU-AI
I thought they would send me a copy of the issue with the interview
with me, but they didn't and it isn't on the newstands any more. Do
you still have your copy, and may I borrow it.
∂22-Apr-83 1232 JMC
To: bledsoe%UTEXAS-20@USC-ECL
I accept centennial invitation, forgetting whether I already did.
∂22-Apr-83 1534 JMC
To: rindfleisch@SUMEX-AIM
CC: RPG@SU-AI, REG@SU-AI, feigenbaum@SUMEX-AIM
I would be quite unhappy about having to spend all the money at once, because
I think we don't have enough experience with the work station philosophy
and with the particular proposed work stations - whether LISP machines,
SUNs or VAXes. Also neither the 2080 nor the S-1 are ready to be
bought. Also we don't have enough experience with expanded file
size to support or refute my contention that we will require very
much larger file sizes than presently contemplated.
Ohlander's latest about the spreadout of the money seems ok to me.
Please forward this to other people concerned, because your list of
recipients doesn't work from SAIL.
∂22-Apr-83 1537 JMC
To: admin.library@SU-SCORE
Thanks for the information about Stoyan's book. I assume that this is
the Dresden published edition, and it's interesting that the East Germans
continue to make this available. I have a copy Stoyan sent me.
∂22-Apr-83 1538 JMC
To: mrc@SU-SCORE
Thanks.
∂22-Apr-83 1540 JMC
To: RPG
CC: DFH
I'll try to get DARPA approval for the foreign travel. I don't think
it's impossible.
∂22-Apr-83 1550 JMC user in NYC
To: nic@SRI-NIC
I have a user of SAIL who lives in New York City. What is the best
way for him to get network access to SAIL. E.g. is there a TIP, and
what is the telephone number? I believe he has a 1200 baud terminal
and can also use 300 baud.
∂22-Apr-83 2211 JMC
To: other-su-bboards@SU-AI
jmc - I note that the entire discussion avoids the issue of whether
defense of the country is worthwhile and whether it requires nuclear
weapons research. My opinion is that both questions deserve a yes
answer and therefore SLAC should help if it conveniently can.
I believe that unilateral disarmament is what the issue is really
about and that it would be disastrous.
∂23-Apr-83 2354 JMC
To: llw@S1-A, RPG@SU-AI
I think DARPA will accept the honor of supporting the foreign part.
∂25-Apr-83 1152 JMC
To: cheriton@SU-HNV, reid@SU-SHASTA, feigenbaum@SUMEX-AIM, ALS@SU-AI,
JEF@SU-AI
I presently intend to make C.I.T. my main concern in the Academic
Senate if there is leverage to do so. However, I don't suppose my
election is effective till Fall, and there is currently a Senate
committee on C.I.T. Perhaps it will report then, and we can react
to the report. I say "we" with the intent of trying to solicit CSD
reactions. It would be helpful if people would spell out their
criticisms and desires relative to C.I.T.
My own present opinion, subject to modification by study, is more
along the lines of Brian's than of Art's. While there is often
a technological advantage in centralization (more bang for the hardware
buck), it can be outweighed by the effects of Parkinson's law
in an organization that expands. I think this has happened with
C.I.T., and I doubt that it can be substantially corrected in
the administrative environment of Stanford where all kinds of
bureaucracies have been permitted to grow and impose costs.
However, the technology permits decentralization, and if C.I.T.
were abolished its clients would find other ways of meeting their
computing needs given that its budget from general funds were
suitably distributed. The advantage is that a suborganization
of the University meeting its needs out of its own budget will
reach a different balance between money spent on computer personnel
and money spent otherwise. The last time I knew the figures the
amount C.I.T. spent on personnel was many times what it spent on
the computer. At LOTS I would guess that twice as
much goes for amortizing the computer as goes for personnel.
These financial matters are a major reason why computer science
expertise per se is not the key to cost-effective computing.
Incidentally, it would not help our efforts to reform C.I.T.
if we were to make a financial bungle with CSD-CF that forced
us to let the University take over its financial affairs.
∂25-Apr-83 2210 JMC
To: DFH
Please double-check, because I don't remember.
∂25-Apr-83 2301 JMC
To: other-su-bboards@SU-AI
jmc - I have been informed about C.I.T. from time to time over the last 20
years, and every time I have looked carefully my poor opinion has been
confirmed, but I haven't looked recently. Does anyone know how much money
the University spends from general funds for C.I.T. services, how many
employees total C.I.T. has, and what is the ratio of personnel costs to
computer costs?
∂27-Apr-83 0257 JMC metaphors
To: phil-sci%mit-oz%MIT-MC@USC-ECL
The whole discussion of the scientific community metaphor for mind
and the associated collection of metaphors reinforces my conviction
that metaphors are not science but are a mere decoration. No
argument that depends on metaphors for its effect can be regarded
as convincing. In short the last two week's discussion seems empty
to me. Does anyone else think so?
∂27-Apr-83 1650 JMC
To: csd.scott@SU-SCORE
What must I do to promote Ketonen to Senior Research Associate?
∂27-Apr-83 2100 JMC
To: cstacy%MIT-MC@USC-ECL
Yes, I am getting messages including those from the phil-sci list.
I think we will be on TCP in a few days.
∂28-Apr-83 1046 JMC
To: buchanan@SUMEX-AIM
I have a fair number of corrections if I can find my copy.
∂28-Apr-83 1540 JMC skyhooks
To: king@KESTREL
There was a recent article by Hans Moravec (note spelling) in the L-5
Society magazine. It contains references. Unfortunately, I didn't
keep it. Moravec is HPM at SAIL and is either that or Moravec
at CMUA. He can give you references. SAIL hasn't yet switched
to TCP, so messages have to be forwarded through ECLC. However,
it has been announced that we will switch this weekend if the
remaining debugging goes well.
∂29-Apr-83 0041 JMC article from library
To: DFH
Please get me a copy of an article by Einstein published in Science
vol. 84, starting on page 506 (1936). You need to telephone to find out what
Stanford library has Science for 1936.
∂29-Apr-83 1510 JMC
To: csd.ullman@SU-SCORE
I am willing to do it, but Vaughan Pratt has been paying much more
detailed attention to the candidates for admission than I have.
If he is willing, he would be better than I.
∂29-Apr-83 1612 JMC
To: buchanan@SUMEX-AIM
Paul Martin is at SRI.
∂29-Apr-83 1613 JMC
To: csd.ullman@SU-SCORE
OK, I'll do it.
∂29-Apr-83 1721 JMC
To: king%kestrel@SU-SCORE
skyhooks/su
There was a recent article by Hans Moravec (note spelling) in the L-5
Society magazine. It contains references. Unfortunately, I didn't
keep it. Moravec is HPM at SAIL and is either that or Moravec
at CMUA. He can give you references. SAIL hasn't yet switched
to TCP, so messages have to be forwarded through ECLC. However,
it has been announced that we will switch this weekend if the
remaining debugging goes well.
∂29-Apr-83 1722 JMC dinner?
To: llw@S1-A
Are you interested in dinner before or after your panel with Panofsky.
There may be ARPAnet problems in communicating, because Marty plans
to switch us to TCP tonight.
Unsolicited advice:
Keep discussion with Panofsky friendly. Also imagine that there are
potential recruits for helping with your defense research in the audience.
I think the President's appeal for scientific help will have some effect.
∂30-Apr-83 1143 JMC
To: csd.bscott@SU-SCORE
I am entirely satisfied and would perhaps advocate being even more
generous.
∂30-Apr-83 1156 JMC
To: jmc@MIT-MC
test1
∂30-Apr-83 1310 JMC correcting misinformation about WAITS
To: JJW@SU-AI, pratt@SU-NAVAJO, mailhax@SU-HNV,
nethax@SU-SHASTA
At the time WAITS was developed, there was no rest of the world. There is
no system presently extant - not TOPS-10, not TENEX, not ITS that existed
at the time WAITS was developed. Nor was a hardware configuration suitable
for time-sharing available from one manufacturer. WAITS started with the
DEC time-sharing (I forget what it was called) but had to modify to get
a reliable disk service and to serve displays. D.E.C. later made
modifications, some of which paralleled those of WAITS, to get TOPS-10.
The people who said, "the rest of the world be damned " were ITS and
BBN. BBN copied their SDS-940 system with improvements onto the PDP-10.
SAIL lost through not pursuing time-sharing development money and not
writing papers about the features of WAITS. At no time was WAITS
considered a research project.
∂30-Apr-83 1611 JMC
To: RMS@SU-AI
test
∂01-May-83 1200 JMC temporary mail routing
To: phil-sci-request%oz%MIT-MC@SU-SCORE
SAIL is now running TCP, but mail doesn't work yet.
Mail for jmc-lists@sail should be routed through SCORE.
When our mail is working, I'll send another message.
∂01-May-83 1814 JMC
To: CLT
Please phone home.
∂02-May-83 1048 JMC
To: bosack@SU-SCORE
Chudnovsky, David and Gregory (Chris is Gregory's wife)
423 West 120th St., Apt. 88
NY 10027
home: 864-5320, work 280-3950
∂02-May-83 1545 JMC speak of the devil
To: RPG
Next week Richard Stallman will be in the area and has offered to
1. Give a lecture on flavors.
2. Demonstrate latest LISP machine software on our machine. He has
a disk.
3. Tell us the latest about various controversies.
∂02-May-83 1559 JMC abstract
To: RMS%MIT-MC@SU-SCORE
ARPANET MAIL directly to SAIL will be flakey or non-existent for a few
days, so please mail the abstract to MCCARTHY@SCORE, and it will
be forwarded on ethernet.
∂02-May-83 1603 JMC MACSYMA
To: RPG
According to a file that you get referred to when you tn to mc,
MACSYMA will be available to universities from Symbolics at
nominal cost. Why don't you make it happen on SAIL? Perhaps
previous political obstacles are gone.
∂02-May-83 1658 JMC C.I.T. and lunch
To: siegman@SU-SIERRA
Perhaps I expressed myself too vigorously about C.I.T., but I have
been around Stanford more than 20 years and have encountered it in
many incarnations. Perhaps I tend not to update my information.
Anyway if you are familiar with C.I.T.'s current state or plans
to co-ordinate computer facilities at Stanford, you might be agreeable
to having lunch some time to exchange points of view.
∂02-May-83 1948 JMC Abstract for talk "How Flavors Differ from Smalltalk"
To: JJW@SU-AI, ME@SU-AI, mrc@SU-SCORE
I see that SCORE has become very prompt in forwarding mail.
∂02-May-83 1943 RMS%MIT-OZ%MIT-MC.ARPA@SCORE Abstract for talk "How Flavors Differ from Smalltalk"
Received: from SU-SCORE by SU-AI with PUP; 02-May-83 19:42 PDT
Received: from MIT-MC.ARPA by SU-SCORE.ARPA with TCP; Mon 2 May 83 19:46:00-PDT
Date: Monday, May 2, 1983 10:34PM-EDT
From: Richard M. Stallman <RMS%MIT-OZ@MIT-MC>
Subject: Abstract for talk "How Flavors Differ from Smalltalk"
To: mccarthy at SU-SCORE
∂02-May-83 2042 JMC
To: siegman@SU-SIERRA
How about Wedneday, Thursday or Friday of this week?
∂02-May-83 2058 JMC
To: CLT
BERKELEY (AP) - Dr. Joel Hildebrand, a chemist whose research into
liquids led to an understanding of the ''bends,'' a potentially fatal
deep-sea diving affliction, died Saturday. He was 101.
Hildebrand authored a chemistry text used widely by college freshmen
for three decades. He retired at age 70, but remained a professor
emeritus at the University of California at Berkeley campus until his
death.
∂03-May-83 1148 JMC
To: siegman@SU-SIERRA
I have made a Faculty Club reservation for two at 12:15 on Thursday.
∂03-May-83 1433 JMC correcting
To: pratt@SU-NAVAJO, mailhax@SU-HNV, nethax@SU-SHASTA
At the time WAITS was developed, there was no rest of the world. There is
no system presently extant - not TOPS-10, not TENEX, not ITS that existed
at the time WAITS was developed. Nor was a hardware configuration suitable
for time-sharing available from one manufacturer. WAITS started with the
DEC time-sharing (I forget what it was called) but had to modify to get
a reliable disk service and to serve displays. D.E.C. later made
modifications, some of which paralleled those of WAITS, to get TOPS-10.
The people who said, "the rest of the world be damned " were ITS and
BBN. BBN copied their SDS-940 system with improvements onto the PDP-10.
SAIL lost through not pursuing time-sharing development money and not
writing papers about the features of WAITS. At no time was WAITS
considered a research project.
∂03-May-83 1434 JMC
To: bscott@SU-SCORE
I am entirely satisfied and would perhaps advocate being even more
generous.
∂03-May-83 1600 JMC
To: RPG
Dinner at Louie's at 5:45 Wed will LLW before his panel with Panofsky.
∂03-May-83 1612 JMC
To: rindfleisch@SUMEX-AIM
My condolences and thanks over the bureaucratic fuss to
satisfy DSSW.
∂04-May-83 0137 JMC Keyworth
To: genesereth@SU-SCORE
I talked to Lowell Wood and he thinks that the probability of getting
Keyworth as a lunch speaker or the like is quite high. You can
either write directly or I can call someone called Carol Lynch in
his office.
∂04-May-83 1136 JMC
To: other-su-bboards@SU-AI
jmc - There is one more way in which non-research use costs the Department
money. The Government has a prejudice against paying more than its
"fair share" of computer costs and has auditors to enforce this
prejudice. These auditors can cause the Government to refuse to pay
bills they consider unfair. These rules prevent the facility from
giving away computer time.
All time resources used must be paid for on the same basis
by projects and by the Department's own users. Suppose the computer
were idle for an hour and someone said, "Give me that hour to compute
pi; no-one else will be harmed". The effect of granting is this
request is that the total usage is increased and the Government's
share is reduced.
There are two mitigating factors. First, while all rates
must be available to all customers, it is possible to create rates,
e.g. for inconvenient hours, that are likely to be more used by
one group of customers than another. I understand that C.I.T.
can get away with a $2.50 per hour night rate, because its main
customers are determined nine-to-fivers. Nothing so simple would
work here to bias the costs towards Government supported use.
Second work done toward improving computer facilities, e.g.
improving editors, finger, etc., can be charged to CSD-CF itself,
and these charges are ultimately shared by all users including
contracts. Sometimes a creative interpretation
can be put on time spent on behalf of improving the computer facilities.
However, as Arthur pointed out, the total amount that can
be extracted from Government supported use is bounded by the
budgets of the grants and contracts. What you don't know the
difference between a grant and a contract? When you grow up,
you'll have to learn.
∂04-May-83 2331 JMC
To: rms%oz%MIT-MC@SU-SCORE
The abstract is ok, and you will hear from RPG about time of talk and
possibly a place to stay. I haven't heard from Fateman.
∂04-May-83 2338 JMC
To: RPG
Chuck Richards, Pinnacles Climbing Guide, 1974 is in my Concorde bag.
∂04-May-83 2345 JMC presentation
To: llw%S1-C@SU-SCORE
Congratulations! The presentation seemed close to optimal for the one
or two people for whom it was optimized - especially the last slide.
∂05-May-83 0023 JMC temporary mail routing
To: ME
Does the following mean that TCP incoming mail is working, and I can
have the M.I.T. people mail directly to SAIL?
∂05-May-83 0015 @USC-ECLC,@MIT-MC:GAVAN%MIT-OZ@MIT-MC temporary mail routing
Received: from RANDOM-PLACE by SU-AI with TCP/SMTP; 5 May 83 00:14:02 PDT
Received: from MIT-MC by USC-ECLC; Sun 1 May 83 19:15:54-PDT
Date: Sun, 1 May 1983 22:14 EDT
From: GAVAN%MIT-OZ@MIT-MC
To: JMC%SU-AI@USC-ECLC
Cc: jcma%MIT-OZ@MIT-MC, cstacy%MIT-OZ@MIT-MC
Subject: temporary mail routing
In-reply-to: Msg of 01 May 83 1200 PDT from John McCarthy <JMC at SU-AI>
From: John McCarthy <JMC at SU-AI>
SAIL is now running TCP, but mail doesn't work yet.
Mail for jmc-lists@sail should be routed through SCORE.
When our mail is working, I'll send another message.
Right now, we're sending to you through USC-ECLC. Let me know if this
message reaches you, so that we'll know that other messages are
reaching you. Do you prefer that we route through SCORE? Let us know
when SAIL's TCP is working correctly, and we'll switch.
∂05-May-83 0037 JMC
To: gavan%oz%MIT-MC@SU-SCORE
Routing through ECLC is pointless now, because we no longer are on
NCP. This message was received by SMTP from ECLC, but Martin Frost says SMTP
is still flakey. However, I suppose you might as well start sending
the messages directly to SAIL, since this one worked. SCORE
forwards by Ethernet, so that doesn't depend on our SMTP working.
∂05-May-83 0050 JMC
To: other-su-bboards@SU-AI
jmc - Before the AI Lab moved we didn't operated as PB advocated, and it
worked fine as long as there was one main source of funds - namely ARPA.
We charged costs of running the facility directly to the contract and
did no accounting of usage. Les Earnest and I could also let others
use the machine if we judged that their use would benefit the goals
of the contract. When ARPA funding shrank and we got multiple sources
of funding, we had to try to allocate computer costs. This never worked
well, and when we moved, I felt we had to go to a facility with
charges based on use. Maybe the CMU system could be made to work here
if someone wanted to figure out a proposal for allocating costs. I
doubt it, however, considering that some income comes from outside
the department. PB has strange ideas about who has influence on how
the Government does accounting.
∂05-May-83 0110 JMC metaphor
To: gavan%oz%MIT-MC@SU-SCORE, dam%oz%MIT-MC@SU-SCORE,
rickl%oz%MIT-MC@SU-SCORE, phil-sci%oz%MIT-MC@SU-SCORE
I suppose that metaphors are sometimes of use in generating scientific
hypotheses, although I am not conscious of using them myself.
However, the metaphors are not the hypotheses themselves. It seemed
to me that the discussion which I criticized presented the metaphors
themselves rather than going on to definite hypotheses. For example,
the "society of mind" metaphor went on airily without anyone noticing,
until Marvin pointed it out, that the society to which he was imagining
the mind analogous was more like an anthill than like human society.
Marvin's view was that the parts did not individually have minds or
intelligence. Well the anthill analogy is fine as far as it goes,
but to go beyond the analogy requires yet more specific hypotheses.
Total precision is not quite the issue; it seemed to me that the
discussion was becoming ever more airy and that the emphasis on
metaphor was a big part of the problem.
∂05-May-83 1041 JMC
To: RPG
∂05-May-83 0245 @USC-ECLC,@MIT-MC:RJF@MIT-MC
Received: from RANDOM-PLACE by SU-AI with TCP/SMTP; 5 May 83 02:45:09 PDT
Received: from MIT-MC by USC-ECLC; Mon 2 May 83 23:46:20-PDT
Date: 3 May 1983 02:30 EDT
From: Richard J. Fateman <RJF @ MIT-MC>
To: JMC @ MIT-MC
RMS (Richard Stallman) sent me a note indicating you wnated to get
Macsyma. Certainly you should be able to get a copy for a DEC-10 or -20
for Stanford; I hope also for VAX systems, soon (cost $500 from Symbolics
for the latter; if Symbolics can't supply it, or the cost is too high,
UC Berkeley can send a tape.)
Progress on getting a "public domain" copy depends on how MIT stonewalls
the Dept of Energy.
Feel free to give me a call at 415 642-1879.
Richard Fateman
∂05-May-83 1050 JMC
To: ailist-request%SRI-AI@SU-SCORE
Please put jmc-lists@su-ai on the AIList distribution.
∂05-May-83 1347 JMC
To: DFH
OK about LGC phone.
∂05-May-83 1422 JMC
I would be quite unhappy to lose my Library key. About half of my use
is outside of regular Library hours. I have never stolen any books.
∂05-May-83 1423 JMC
To: library@SU-SCORE
I would be quite unhappy to lose my Library key. About half of my use
is outside of regular Library hours. I have never stolen any books.
∂05-May-83 1742 JMC
To: YOM
blocks[w83,jmc] Blocks axioms using circumscription
circum.com[w83,jmc] Making circumscription computable
circum.not[w83,jmc] More: 1. Circumscribing when generalizing
bird[f82,jmc] Another version of "A bird can fly unless ...
bird.2[f82,jmc] Another approach to " A bird can fly unless
bird.3[f82,jmc] The reasoning itself
circum[f82,jmc] Mathematical questions of circumscription
circum.dis[f82,jmc] Mailing list for reports on circumscription
circum.ex[f82,jmc] Examples for circumscription paper
circum.lec[f82,jmc] Notes for lecture on circumscription
circum.mor[f82,jmc] More on circumscription - for IJCAI or AAAI
cross[f82,jmc] Crossing the street and concurrent action
more[f82,jmc] More on Circumscription
∂06-May-83 1153 JMC natural kinds
To: phil-sci%oz%MIT-ML@SU-SCORE, dam%oz%MIT-ML@SU-SCORE,
batali%oz%MIT-ML@SU-SCORE, rickl%oz%MIT-ML@SU-SCORE
I think the best way of thinking about natural kinds is as
a fact of physics. Consider the Durgin-Park slogan: "There is no
place anything like this place anywhere near this place, so this
must be the place". It is a fact about the world that cows aren't
continuous with horses, and 1 isn't continuous with 2. We can
imagine another world in which we would have to establish an
arbitrary boundary between cows and horses, just as, in our
world, we would need to be arbitrary in order to distinguish
hills from mountains. Perhaps there is a planet somewhere in
with terrain in which there is a sharp distinction.
The key epistemological property of natural kinds is that
the kind has properties that we don't a priori know about and
can subsequently learn. Amusingly, even when objects are not
natural kinds, the concepts and words for them often are.
Children believe very strongly in natural kinds. When a child
hears a new word, he is ready to believe that it designates
something that has many properties he doesn't know about and
can subsequently learn.
Essential properties on the other hand seem irrevocably
epistemological and linguistic; they relate to conventions
whereby people will maintain communication in the face of
hypothetical new information. They can be only partially
successful. Gedanken experiment: Imagine that
Massachusetts and California are to be linguistically separated
for 20 years and that during this twenty years the same but
presently unknown changes are to occur in the world. For example,
there may develop intermediates between cows and horses. Suppose
we try, before the separation, to develop linguistic conventions
about what are the essential properties of various entities,
e.g. species of animals, that will ensure that after the linguistic
separation ends, we will still be using the same terminology.
Success would be quite partial.
∂06-May-83 1154 JMC my mail address
To: phil-sci-request%oz%MIT-ML@SU-SCORE
Our TCP seems to be working well enough so that it is no longer
necessary to go through another computer, and my address should
be jmc-lists@su-ai with whatever decoration is required to get
out of oz.
∂06-May-83 1331 JMC failed mail return
To: phil-sci-request%oz%MIT-ML@SU-SCORE
When I MAIL to phil-sci, I get failed messages back aimed at recipients
I never heard of. I presume this is because of wrong entries on the
phil-sci mailing list. Can I avoid this by only copying phil-sci?
∂06-May-83 1336 JMC
To: grosof@SU-SCORE
Just look for me in the afternoon - in person or electronically.
∂06-May-83 1342 JMC
To: bboard@SU-AI
for merger with page 143
jmc - Stimulated by a discussion with AFL-CIO people, my interest has
revived in fascist computer programs. It will be interesting to see
what Stuart comes up with.
∂06-May-83 1442 JMC seminar
To: hewitt%MIT-XX@SU-SCORE, cobb%oz%MIT-ML@SU-SCORE
speaker: John McCarthy, Stanford
title: A new version of circumscription applied to blocks world
time: 1983 May 13 Friday 2pm
place: AI playroom
Priscilla,
Please reserve the AI Playroom for this time,
check with LCS Headquarters for conflicting
seminars, and make seminar announcement sheets for
distribution (there will be no abstract)
Please arrange for borwnies and coffee after the seminar.
Thanks,
Carl
∂06-May-83 1714 JMC
To: admissions@SU-SHASTA
I don't know its history, but I oppose the exclusion of engineering. I don't
understand what Jeff has changed, but if he has changed it so that the
whole thing reads "in the applicant's major field of study" and the
exclusion of engineering is omitted, that's fine with me.
∂07-May-83 1014 JMC visit and seminar
To: kdf%oz%MIT-ML@SU-SCORE, dam%oz%MIT-ML@SU-SCORE
I will be in Cambridge from Thursday evening till Sunday and will
give a seminar on circumscription and the blocks world on Friday
afternoon. I would like to see both of you to discuss philosophical
and other matters of mutual interest. I plan to spend Friday at the
M.I.T. AI Lab.
∂07-May-83 1059 JMC obscurity
To: REG@SU-AI, gotelli@SU-SCORE
Please fix the budget document so that the disk rates for SAIL and
SCORE refer to the same amount of storage.
∂07-May-83 1317 JMC
To: gavan%oz%MIT-ML@SU-SCORE
Thanks, Gavan. Your message dated Fri, 6 May 1983 22:59 EDT arrived
at 7 May 1983 12:46:56. It would be nice to know the cause of the
delay, but unless the delay is a known one at your end, it is better
to wait a week or two for further debugging of TCP here before chasing
problems.
∂07-May-83 1831 JMC
To: reiter%RUTGERS@SU-SCORE
gricean references
∂07-May-83 1837 JMC circumscription and satisfaction
To: YOM
Remember the problem of their relation in the predicate case, i.e.
not merely propositional.
∂07-May-83 2342 JMC natural kinds
To: batali%oz%MIT-ML@SU-SCORE, phil-sci%oz%MIT-ML@SU-SCORE
I'll try one more argument about natural kinds being objective - at
least more so than essential properties. Can you imagine a computer
program or some other intelligent animal to which there would be
a clear distinction between hills and mountains? There are multiple
objective differences between cows and horses, and different experience
will determine which are taken as defining. A race that classified
solely on the basis of mass and which therefore didn't make
a sharp distinction would have to be feeble minded. There are physical
distinctions and the genetic distinction that cows only come from
cows and horses from horses.
It seems to me that taking a subjective position on cows and horses
leads to taking a subjective position on almost all distinctions and
therefore two an impoverished science and an impoverished robotics.
∂08-May-83 1041 JMC mailing to bulletin boards
To: RMS%SU-AI@SU-SCORE
You can mail from E by <ctrl><meta>XMAIL SU-BBOARDS<cr>.
It is worthwile to make the first line of your message a subject
in which case the command should be <ctrl><meta>XMAIL SU-BBOARDS/SU<cr>.
This will mail the current page to all bulletin boards - currently
several tens and twenties and vaxen.
∂08-May-83 1538 JMC
To: batali%oz%MIT-ML@SU-SCORE, rickl%oz%MIT-ML@SU-SCORE,
phil-sci%oz%MIT-ML@SU-SCORE
I'm pleased with John Batali's latest questions, because a few of
them actually seem to have definite answers. Here's how building the
notion of natural kinds into robots will help them. Digression about
children first: Children aren't told that there are natural kinds. Either
it is built in or they develop it early. Their linguistic prejudice is
that a new name designates an individual or a natural kind.
"Daddy, is that a hill or a mountain?", is the kind of question
that children ask that is hard to answer.
Children don't have to take the observed properties of the first
example pointed out to them as defining - as the essential properties if
you like. Show a child its first lemon, call it a lemon, and then ask it
to go to the refrigerator and get a blue lemon. If there are blue lemons
in the refrigerator, the child will find them, because it doesn't take the
yellow color as part of the definition. It may help to tell a robot that
a new name is of a natural kind, but it shouldn't be any more necessary
than for a child. Since we don't know what is innate in a child, we don't
know what we must build into robots and what we should expect them to be
able to learn from experience using the learning mechanisms we build in.
Telling the robot that cows and horses are discontinuous involves
giving information about pairs of terms. The natural kind idea allows a
linear rather than a quadratic amount of information to be given.
Perhaps the concept of natural kind isn't itself a natural kind,
i.e. natural kind concepts grade off into other concepts. However, I
agree with Rick's characterization of it. I also agree with him that we
can make mistakes about whether a concept is a natural kind.
∂08-May-83 2108 JMC visit
To: minsky%oz%MIT-ML@SU-SCORE
I will be in Boston area from Thursday evening till Sunday. Is it
convenient to have me stay with you?
∂08-May-83 2231 JMC natural kinds
To: gavan%oz%MIT-MC@SU-SCORE, phil-sci%oz%MIT-MC@SU-SCORE
It seems that Gavan is correct that our acceptance of an objective
distinction between cows and horses is related to a correspondence
theory of truth. As to faith, I haven't figured out how to answer
someone who asserts that X is a matter of faith. However, if it is
a matter of faith, rather than say a Pascal type wager, we can build
a science on it if we want, and Gavan can only grumble about it.
My remark about an impoverished science comes to this. To many of
us the fact that the cow-horse distinction is different from the
hill-mountain distinction is important for the design of robots.
A theory that cannot distinguish or even dislikes distinguishing
between these distinctions is impoverished. Saying otherwise should
be supported by examples where (say) replacing "objective" by
"intersubjective" has led to scientific success.
Finally, it is true that genetic engineers may someday produce a
continuum between cows and horses especially if Gavan becomes head
of NSF. That supports the point that natural kinds are a fact about
the world and not just a fact about concepts.
∂08-May-83 2348 JMC visit
To: minsky%oz%MIT-ML@SU-SCORE
Also Carl Hewitt is arranging for me to give a seminar on Friday
at 2pm about circumscription and the blocks world.
∂08-May-83 2351 JMC visit
To: gavan%oz%MIT-ML@SU-SCORE
As you may know, I am coming to Boston for the time-sharing history
session this coming weekend. I will actually arrive Thursday night
and will probably stay at Minsky's. Where and when do I actually
appear for the videotaping? I will give an AI seminar on Friday
about circumscription and the blocks world.
Many thanks for refixing my address in phil-sci. Incidentally,
a guest account at OZ was arranged for me, but it happened just
about the time we went off ARPAnet and now it has disappeared.
If it isn't too much trouble to revive it, I would be grateful.
∂09-May-83 0917 JMC message from Ohlander
To: DFH
DARPA wants changes in our proposal. Please pick up the copy of the
message I have just told the Dover to print and make two xeroxes.
One for me, one for RPG and keep one in your office for others in
the group to look at.
∂09-May-83 0937 JMC dates for Keyworth
To: genesereth@SU-SCORE
I need a preferred date and time and, if possible, alternates.
∂09-May-83 0938 JMC dates for Keyworth
To: genesereth@SUMEX-AIM
I need a preferred date and time an, if possible, alternates.
ASAP., since I want to phone today because of going out of town.
∂09-May-83 1322 JMC Keyworth
To: genesereth@SUMEX-AIM
I called Carol Lynch, 202 456-7116, who is Keyworth's executive secretary,
but she was out of town and I spoke only to Elaine Carson, who answered
the phone. We agreed that I would phone Carol Lynch again next Monday.
However, the procedure is, as might be expected, to write a letter of
invitation which should include information describing the organization
and the number of people who would be in the audience. We could send
a letter right away, but there is certainly no problem in waiting till
Monday. I suppose we should offer flexibility about how long a talk
we want, but I imagine that half an hour would be about right. After
I establish contact (Wood claims that Keyworth has heard of me), I'll
pass the buck to you.
∂09-May-83 1422 JMC
To: DFH
Those reservations are entirely satisfactory.
∂09-May-83 1621 JMC
To: minker.umcp-cs%UDEL-RELAY@SU-SCORE
Please acknowledge receipt of this test mail.
∂09-May-83 2157 JMC <ctrl>z
To: gavan%oz%MIT-MC@SU-SCORE
My head is stuck half way through this new user hole. The program
wants <ctrl>z, but I don't know how to sent it through MC.
It is now 0058 your time.
∂09-May-83 2237 JMC
To: gavan%oz@MIT-ML
I seem to have succeeded in activating the account.
∂09-May-83 2312 JMC
To: minsky%oz@MIT-ML
Marvin, you would find some (perhaps one) of the philosophy papers on
natural kinds interesting. Unfortunately, I don't have the reference, but
I'll bring it if I can find it; it's relatively recent. I was the one who
raised the issue in the phil-sci discussion - contrasting natural kinds
with essential properties. The idea is this:
1. Previously philosophers supposed that concepts had definitions
that determined (for example) what you meant by a lemon.
2. But then suppose some geneticist comes up with a blue lemon.
If yellow was part of the definition of a lemon, how do we account
for the fact that people will accept blue lemons as lemons even
if they don't know genetics.
3. The idea of natural kinds is that people in fact don't recognize
things by definitions or even prototypes. Consider a child. It
hears that we are going to the store to buy a lemon. It decides
that there is something called lemons, and it is ready to learn
more. It sees this small yellow fruit in the store, and says to
itself, "That little yellow fruit is a lemon". It can then
recognize other lemons. However, it is prepared to learn that
some lemons are blue and would be prepared to learn that some
lemons are as large as grapefruits.
4. What makes this work, since it seems that it could be stretched
into calling anything a lemon?
5. It is a fact about the world that many kinds of objects are
discretely separated from other objects. In principle, there
could be a continuous range between lemons and grapefruit or
between horses and cows, but there isn't. Of course, sometimes
there is as between hills and mountains. When there is no
bridge, there are fewer problems, and this situation is so
common that we have evolved and our language has evolved to
take advantage of it.
6. The point at issue in the phil-sci discussion is that natural
kinds are natural; they aren't a linguistic convention. The
world helps us out in this way, and Martians would be equally
pleased to empirically discover that there are no bridge fruit
between lemons and grapefruit or bridge animals between horses
and cows.
7. Robots should also be designed to take advantage of this
phenomenon.
∂10-May-83 1044 JMC paradigms
To: phil-sci%oz@MIT-ML
Once upon a time Thomas Kuhn wrote a book called "The Structure of
Scientific Revolutions" that introduced the notion of "paradigm
shift" with such examples as the transition from medieval to the
Galileo-Newton view of mechanics and the transition from classical
to quantum mechanical views. Now every PhD thesis proposes its
own paradigm shift, and Vaughan Pratt has five new paradigms in
one report dated 1982. It's clear that the technology of making paradigms
has entered an era of expansion like that in semi-conductor
memories. However, note that Pratt has five rather than four
paradigms. My proposed meta-paradigm is that the number of new
paradigms in each paper can be expected to increase by a factor
of five every three years. We will therefore reach the mega-paradigm
paper in the year 2026.
∂10-May-83 1335 JMC
To: dam%oz@MIT-MC
vrp@su-ai Will be forwarded.
∂10-May-83 1601 JMC
To: JCMa%oz@MIT-ML
Your message arrived ok, and my request to use JMC-LISTS@SU-AI for the
phil-sci list seems to have been implemented.
∂10-May-83 1602 JMC
To: gavan%oz@MIT-ML
You can MAIL to SU-AI by ARPA directly. ECLC not needed.
∂10-May-83 1611 JMC
To: ZM, DCL
When I have your lists it will be ready to go.
∂10-May-83 1655 JMC DARPA scope document
To: DEK@SU-AI, golub@SU-SCORE, GIO@SU-AI
CC: bscott@SU-SCORE, ZM@SU-AI
Ron Ohlander requires changes in the DARPA proposal scope document
and noted that Don's and Gene's areas weren't included in the previous scope
document. I have included items in the style that I believe him
to be requesting; they are cribbed from the main proposal.
Don's is F - Analysis of algorithms and Gene's
is G - Relations between algorithms and architecture. I modified
Gene's title for fear that some bureaucrat would find redundancy.
I don't think the details matter much in terms of what work you
can do assuming DARPA will support it, but prompt action was
necesary. Depending on how far the bureaucracy has proceeded
when you see this, revisions can be made. The file is
SCOPE[S83,JMC] at SAIL. Betty Scott will know what has to be
done to effect changes. I have only been able to make trivial
changes in Gio's, but it was one of the parts with which Ohlander
had the fewest problems. While Gio and I are away, Zohar is
in charge of responding to technical questions.
∂10-May-83 1705 JMC scope
To: DCL
The document is done except for your part. Please mail me your
revisions ASAP but before 11; I intend to mail something to Ohlander tonight.
∂10-May-83 1754 JMC
To: RMS@SU-AI
Your manuals are on my file cabinet.
∂10-May-83 1815 JMC
To: DFH
Please send telegram in takasu.3[let,jmc].
∂10-May-83 1906 JMC revised scope information.
To: ohlander@USC-ISI, bscott@SU-SCORE, ZM@SU-AI, DEK@SU-AI, DCL@SU-AI,
GIO@SU-AI, TOB@SU-AI, golub@SU-SCORE
Here is the revised scope information. It is revised from the copy
mailed by Ohlander to Betty Scott. The revisions were performed as
follows:
A. by McCarthy who stands by it.
B. by Luckham who stands by it.
C. by Manna who stands by it.
D. by McCarthy in the absence of Wiederhold. Only trivial revisions were made.
E. by Binford who stands by it.
F. by McCarthy in the absence of Knuth. It is abstracted from the main proposal.
G. by McCarthy in the absence of Golub. It is abstracted from the main proposal.
Stanford University will perform Research as follows:
A. Basic Research in Artificial Intelligence
--------------------------------------------
1. Develop formalizations of facts about the common sense world including laws
determining the effects of actions including mental actions.
2. Develop formalizations (both procedural and declarative) of the
heuristic knowledge used in common sense reasoning and problem solving.
3. Develop formalization of common sense reasoning including non-monotonic
reasoning.
4. Study representation of facts in the memory of a computer including the
connections between AI representation work and database research.
5. Develop languages for communications among computers and between
computers and people, emphasizing the semantic aspects of such languages
relevant to commercial and military communications.
6. Develop systems for computer reasoning and computer-assisted human reasoning.
7. Improve AI Programming languages including LISP, especially Common
Lisp, and also perform research aimed at new languages.
8. Experiment with techniques of automatically specializing programs to
make them run faster when part of their data is constant.
9. Develop programs that take advice from users and give advice in return,
and programs that decide what to do by formal reasoning.
B. Research and Development for Advanced Programming Environments
-----------------------------------------------------------------
Stanford will investigate advanced techniques for production and maintenance
of software. This research has the goal of developing automated aids for all
stages of the production of software systems (requirements, design,
implementation, and maintenance). New high level languages will be designed.
Tools providing automated support for these languages will be developed and
techniques for integrating them into programming environments will be studied.
Special emphasis is placed on the production of new complex systems that
utilize concurrent and distributed processing. The research will focus
primarily on software, but applicability to hardware design will also be
studied.
Specific projects will be undertaken within each of the following tasks:
1. Design new high level languages for formulation of systems requirements,
design specifications, formal annotation of implementations, and
documentation. Such languages are to be machine processable.
2. Develop techniques and guidelines for specifying systems designs and
implementations in these languages.
3. Design, implement, and test tools supporting software production in these
languages. Such support tools will include tools for (i) testing and
validation of requirements, design specifications, and implementations, (ii)
compilation of specifications for rapid prototyping, and (iii) formal
consistency analysis.
4. Develop techniques and tools for analyzing the parallel activity in
systems; this includes consistency analysis, analysis of runtime behavior, and
analysis of communication errors among parallel threads of control.
5. Study the formalization of programming knowledge and its codification in
the broad spectrum languages (task 1) for use in automated knowledge-based
programming assistance.
6. Develop techniques for computer-supported reasoning as required for
advanced automated tools under tasks 1 through 5 above.
7. Design of advanced programming environments that integrate tools developed
under tasks 1 through 6 above.
C. Software Development Techniques
--------------------
Stanford will:
1. Design a high-level logic programming language using new deductive
techniques.
2. Develop and implement interactive and automatic systems for program
synthesis, program verification and planning.
3. Introduce tools for machine reasoning (automated deduction) specially
directed toward software engineering applications
4. Investigate program manipulation techniques: maintenance, transformation,
and optimization.
5. Develop techniques for the verification and synthesis of concurrent
programs.
6. Formulate techniques for specifying, developing, and manipulating
non-applicative programs.
7. Develop tools for rigorously establishing properties of software and
hardware systems (using logic of programming such as dynamic logic
and temporal logic).
D. DATA MANAGEMENT
------------------
Stanford will investigate advanced techniques to improve database access and
management. Within this objective we will analyze the semantics inherent in
the data and in the operations persformed on data. Where the boundaries of
algorithmic approaches are reached we will develop heuristic techniques. We
will stress methods which are domain-independent so that the technology can
be transferred to a wide variety of applications.
Specifically we will:
1. Specify and use data semantics. We will use a categorization of data
semantics to establish those semantics which can aid in designing and
distributing well-structured databases and those which may be of operational
utility.
2. Investigate ambiguity, view and performance problems in database update.
3. Extend and develop semantic aids to the anlaysis of database contents.
4. Develop communication and audit-trail concepts in design databases.
5. Investigate algorithms for maintaining data on optical disks.
6. Acquire and maintain data resourses and database management systems to
support this research.
E. Image Understanding and Robotics
-----------------------------------
Stanford will conduct research in the area of computer vision to
address the problems of cartography, photointerpretation and robotic sensing
as follows:
1. Analyze, design and implement intelligent systems for interpretation
and planning actions, including intelligent interfaces for users incorporating
natural language and speech I/O.
2. Analyze and implement subsystems for geometric reasoning and reasoning in
the physical world including
representation of the physical world, problem formulation,
space/time reasoning, and general methods for using domain-specific knowledge.
3. Investigate learning and abstraction
in building and using databases for the physical world.
4. Study planning, navigation and path-finding, motion control
map-making and world modeling for mobile robots.
5. Study the segmentation and aggregation of image features to analyze texture
regions, canonical groupings, and figure-ground discrimination;
6. Analyze, design and implement systems for stereo mapping, motion parallax and
object motion in spatial interpretation;
7. Investigate architecture of image algorithms and their
implementation in VLSI.
8. Study mechanisms for interpretation of images including
matching algorithms, generic models, multi-sensor integration,
geometric inference rules, and shadows.
9. Study the design of robot devices, task control, programming and planning of
actions.
F. Analysis of Algorithms
-------------------------
Stanford will perform research in creating new computer algorithms for
practical problems, develop the mathematics required to determine
efficiency of algorithms and extend programming methodology so that
such algorithms can be implemented more quickly and reliably than with
present techniques. More specifically we will:
1. Develop new algorithms for the digital raster graphics, the
combinatorial matching problem and the "all nearest neighbors" problem.
2. Develop applications of these algorithms to a variety of problems.
3. Investigate random mappings which have applications to cryptanalysis.
4. Develop programming methodology based on the WEB programming system
which combines systems for developing programs with document formatting.
G. Relations between Algorithms and Architectures
-------------------------------------------
Stanford will investigate the relations between algorithms and computer
architectures for executing them as follows:
1. Develop algorithm-independent results on the complexity and topology
of various applications.
2. Obtain lower bounds on the optimal ratio of processor pwer to
communication capability for representative problems.
3. Relate algorithmic requirements in terms of topology, data rates,
and processor capability to physical and technological restrictions.
4. Investigate inherent tradeoffs in convergence rates and complexity
with concurrency.
5.Investigate the possibility of a metalanguage for the description
of application problems in terms of standard computational processes
which express both the complexity of the computation and its topology
and which allows accurate simulation of the process for various
architectures.
∂10-May-83 2201 JMC
To: DFH
Stallman's manuals are on my file cabinet.
∂11-May-83 0015 JMC
To: novak@SUMEX-AIM
I shall be out of town so please phone Prof. Alphonse Juilland.
∂11-May-83 0019 JMC Classical duets, anyone?
To: CLT
∂10-May-83 2045 HURD@SCORE Classical duets, anyone?
Received: from SU-SCORE by SU-AI with PUP; 10-May-83 20:45 PDT
Date: Tue 10 May 83 20:48:29-PDT
From: Elizabeth Hurd <HURD@SCORE>
Subject: Classical duets, anyone?
To: su-bboards@SCORE
I am a pianist who would love to get together with fellow musicians to
play music. In the past I have played with, and accompanied, cellists,
flutists, and vocalists. If anyone is interested in playing classical
music (anything but modern) please send me mail or call me at home.
Elizabeth Hurd
961-9104 or Hurd@SCORE
-------
∂11-May-83 0031 JMC letter to Col. Gordon
To: DFH
Note that I have offered to review his paper. If it comes while
I am on my Far East trip, inform him of when I will be back.
∂11-May-83 0717 JMC
To: pourne@MIT-MC
Your message sent at 5:26 arrived at 2:24.
∂13-May-83 1639 JMC
To: DFH
∂13-May-83 1037 DBL ai qual
John,
Is the AI Qual still this Tuesday, May 17?
If so, what times (recall that 10:30-12:15 and 4-5 are bad for me)?
Do the students know where and when?
If not: when is it?
Doug
∂13-May-83 1652 JMC via mail rms
To: rms@MIT-MC, ME@SU-AI
mail to rms here is forwarded to rms at the now non-existent ai.
I believe MC is the correct place to forward it.
∂13-May-83 1739 JMC
To: DFH
make it thursday at 2 then..
∂16-May-83 1038 JMC
To: kanef.hp-hulk@RAND-RELAY
The paper was published in Artificial Intelligence, April 1980.
∂16-May-83 1109 JMC line inoperative
To: ME
The line from my house seems to be inoperative. Is that
part of the rewiring?
∂17-May-83 1847 JMC Data types and natural kinds
To: phil-sci%oz@MIT-ML
According to my philosophical consultant, "natural kind" is
an old term that has become fashionable again within the last 15
years. It makes a very nice contrast with "data type" or "object"
as used in recent programming languages.
Types of objects are introduced by DEFSTRUCTS etc. which give the
components of their structure. Natural kinds are introduced to
a human either by name or by observation. Examples:
A child has never heard of lemons. It is told one of
1. "Lemons are what this lemonade is made from".
2. "That's a lemon".
3. "A lemon is a fruit like an orange only smaller and yellow".
In each case it files the name as designating an entity
about which it is prepared to learn more. Clearly it is desirable
to provide computer programs with this facility, and it doesn't seem
very difficult to do it in a primitive way. Maybe it can even be
done reasonably well with the mechanisms now available. We introduce
the name as designating an object with few properties and in the
course of the operation of the program more properties are added.
There is a certain duality between natural kinds and definitions.
∂17-May-83 1940 JMC
To: RV
Please phone me. 7-4430 or 857-0672
∂17-May-83 2008 JMC
To: ME
My line works now.
∂17-May-83 2142 JMC Meeting
To: karp@SUMEX-AIM
∂16-May-83 1048 DFH Meeting
To: "@FACCOM.[1,DFH]"@SU-AI
To: Facilities Committee
From: Diana Hall (dfh@su-ai)
Subject: Meeting
The meeting will be held Thursday, May 19, from 2:30 - 4:00 pm in the
Chairman's conference room. I am aware that there are a number of you
who cannot make this time, but it has proved impossible to find a time
agreeable to everyone involved.
∂17-May-83 2224 JMC common.msg
To: RPG
I think it's time to start a new common.msg; it now takes between
eight and ten seconds of computer time to get E to look at the last page.
I just got back from the Far East. Here is the message referred to.
∂18-May-83 0031 JMC role of logic in AI
To: phil-sci%oz@MIT-ML
Since my views have been cited by DAM and others, I get to state them.
1. First order logic is logically universal in that any other logic
can be formalized withing first order logic by introducing appropriate
entities as objects. Specifically, second order logic is so formalizable,
and set theory formalized in first order logic is more powerful than
any high order logic. "More powerful" must be taken in a technical
sense that involves similar restrictions on what can be objects in
order for this statement to make sense.
2. This doesn't mean that only first order logic should be used.
Indeed my current formulation of circumscription is in second order
logic. Reformulating it in first order logic would probably be
clumsy.
3. In order to justify using a more elaborate logic than first
order logic, whether it be higher order or modal, there has to be
a good reason. Many of the candidate modal logics fail, because
they don't admit enough meta-reasoning within the logic for
practical purposes, so that certain first order formalisms are
actually more powerful. My paper on First Order Theories of
Individual Concepts and Propositions has examples of this.
4. Logic can be used in AI in four ways.
a. The systems most committed to logic do their computation
by reasoning in logical systems. I don't know any way to do this
very efficiently, but I also know no proof that it is impossible.
(I see it as very important to formulate precise assertions about
the computational power or lack of it in various formalisms).
b. A slightly weaker system uses logic at the outer
level, but can use arbitrary programs to generate answers
that are then converted into sentences by reflexion principles.
This is arbitrarily powerful, because logical deduction could end up
doing nothing but blessing the results of the computations. Naturally,
this isn't the intent of such systems, which I think have considerable
promise.
c. A system which is basically something else, e.g.
productions, could use logical deduction as part of its computation.
Thus STRIPS used deduction for reasoning within a state and
handled state transitions in another way. Presumably the human
mind works this way, since we seem to use what amounts to logical deduction
for a substantial amount of our conscious thought, but it is very
improbable that the intellectual mechanisms we share with animals
are very deductive.
d. As Newell emphasized in his 1980 AAAI presidential address,
and as I emphasize in my Ascribing Mental Qualities to Machines,
even if a system uses neither sentences for representing information
or deduction as computation, it may often be conveniently described
at a "logical level". Daniel Dennett makes similar points in his
Herbert Spencer lectures when he talks about "the intentional stance".
I choose to emphasize the epistemological part of the problem
in my own work, because I do that much better than writing programs.
If DAM disagrees with these views, I would like to read specific
criticisms.
Later added note:
Circumscription uses second order logic, because it seems
clearest for expressing non-monotonic reasoning. However, if we
want a system that reasons ABOUT non-monotonic reasoning, it may
be better to find some way of pushing circumscription down into
first order logic.
∂18-May-83 0855 JMC SOW
To: bscott@SU-SCORE
Please distribute copies of the SOW to my fellow piglets. I have looked at
it, and since it pleases DARPA, it's fine with me.
∂18-May-83 0939 JMC role of logic in AI
To: phil-sci%oz@MIT-ML
Added note:
Circumscription uses second order logic, because it seems
clearest for expressing non-monotonic reasoning. However, if we
want a system that reasons ABOUT non-monotonic reasoning, it may
be better to find some way of pushing circumscription down into
first order logic.
∂18-May-83 1224 JMC Keyworth is out
To: genesereth@SU-SCORE
He will be on vacation for the whole of August. Another possibility is
to try Frank Press who was Carter's Science Adviser and is now President
of the National Academy of Sciences. I don't know him, so if you want
to do it, you should just write to him.
∂18-May-83 1307 JMC
To: genesereth@SUMEX-AIM
Another possibility would be to ask if there is another official from
Keyworth's office who would be appropriate. You could call his executive
secretary Carol Lynch 202 456-7116 and ask.
∂18-May-83 2138 JMC
To: CLT
06-21 Tues. Teller reception and dinner, 6:30 and 7:30, Faculty Club
∂18-May-83 2248 JMC
To: wiederhold@SUMEX-AIM
Ohlander has tinkered further and is now satisfied with what follows.
Betty Scott thinks she can do everything that remains. However, since
I am leaving for Japan and Taiwan for 3 weeks on Saturday, I suggest
you co-ordinate with her. Welcome back, and I hope you had an
interesting trip.
∂18-May-83 0742 OHLANDER@USC-ISI Stanford SOW
Received: from USC-ISI by SU-AI with TCP/SMTP; 18 May 83 07:41:56 PDT
Date: 18 May 1983 0740-PDT
Sender: OHLANDER at USC-ISI
Subject: Stanford SOW
From: OHLANDER at USC-ISI
To: JMC at SU-AI
Cc: BScott at SU-SCORE
Message-ID: <[USC-ISI]18-May-83 07:40:56.OHLANDER>
John,
I have performed considerable work on the draft SOW and I now
believe it to be in reasonable shape. It should be added (modulo any minor
changes that you might want to make) to the proposal. You will note that I
decided to consolidate Luckham's and Manna's work afterall.
There are still a few other problems with the proposal. Some of them are
minor. We would like to have all of the names of the coprincipal investigators
removed from the cover and we would like to have a title for the proposal.
Another more serious problem regards the budget. It has to be spelled out
in detail. I will talk to Betty Scott offline about this.
Duane Adams is going to be at Stanford Tomorrow. I believe he plans to talk
to you about these issues.
Regards, Ron
Statement of Work
Stanford University proposes to conduct a broad program of research in the
field of computer science. The specific research will include: fundamentals of
artificial intelligence; advanced programming techniques and environments; data
management; image understanding and robotics; analysis of algorithms; and
relations between algorithms and architectures. The following specific task
areas are proposed:
1. Basic Research in Artificial Intelligence
Stanford University proposes to conduct research in aspects of basic artificial
intelligence technology with the objectives of making significant advances in
machine reasoning capabilities, automatic speedup of programs, and artificial
intelligence language design and implementation. Examples of the kinds of
tasks that will be undertaken in this area include: development of
formalizations of common sense reasoning and about facts and knowledge used in
the common sense world; development of languages for communications among
computers and between computers and people; development of systems for computer
reasoning and computer-assisted human reasoning and systems that take advice
from users and give advice in return; improvement of artificial intelligence
languages and also performance of research aimed at new languages; and
experimentation with techniques of automatically specializing programs to make
them run faster.
2. Research in Advanced Programming Techniques and Environments
Stanford University also proposes to carry out research in the development of
advanced programming environments and software production with the goals of
providing much better programming tools than currently exist. Examples of the
kinds of tasks that will be undertaken in this area include: design and
implementation of new high level languages for formulation of systems
requirements, design specifications, formal annotation of implementations, and
documentation; design and implementation of an advanced programming environment
that supports software production in these languages; development of techniques
and tools for analyzing the parallel activity in systems and for verifying and
synthesizing concurrent programs; study of the formalization and codification
of programming knowledge and introduce machine reasoning techniques for
software production applications; design of a high-level logic programming
language using new deductive techniques; development of tools for rigorously
establishing properties of software and hardware systems (using logic of
programming such as dynamic logic and temporal logic); and investigation of
program manipulation techniques in such areas as maintenance, transformation,
and optimization.
3. Data Management
Stanford University also proposes to carry out a research project in data
management with a goal of improving data management capabilities by the
development of better algorithms and the incorporation of machine intelligence.
Examples of the kinds of tasks that will be undertaken in this area include:
investigation of advanced techniques to improve database access, storage,
update, and management; development of new techniques for the use of databases
in VLSI design; investigation of techniques for maintaining reliability and
integrity in distributed data base systems; and exploration of methods of
inferring new knowledge from data.
4. Image Understanding and Robotics
Stanford University also proposes to carry out research in image understanding
and robotics with the goals of achieving automatic understanding of images and
sensing and intelligent planning by robotic devices. Examples of the kinds of
tasks that will be undertaken in this area include: analysis, design and
implementation of intelligent systems for interpretation and planning actions;
study of planning, navigation and path-finding, motion control map-making and
world modeling for mobile robots; study, design and development of mechanisms
for interpretation of images; and investigation of architecture of image
algorithms and their implementation in VLSI.
5. Analysis of Algorithms
Stanford University also proposes to conduct research in the analysis of
algorithms with the objective of achieving dramatic speedups in certain classes
of algorithms. Examples of the kinds of tasks that will be undertaken in this
area include: development of new computer algorithms for a class of practical
problems in order to explore general issues of efficiency of algorithms;
development of the mathematics required to determine efficiency of algorithms;
and extension of programming methodologies so that such algorithms can be
implemented more quickly and reliably than with present techniques.
6. Relations between Algorithms and Architectures
Stanford University also proposes to carry out research in examining the
relations between algorithms and architectures in order to gain an appreciation
of the processing power required for certain classes of algorithms. Examples
of the kinds of tasks that will be undertaken in this area include;
determination of lower bounds on the optimal ratio of processor power to
communication capability for representative problems; relation of algorithmic
requirements in terms of topology, data rates, and processor capability to
physical and technological restrictions; investigation of inherent tradeoffs in
convergence rates and complexity with concurrency; and investigation of the
possibility of a metalanguage for the description of application problems in
terms of standard computational processes which express both the complexity of
the computation and its topology and which allows accurate simulation of the
process for various architectures.
∂18-May-83 2311 JMC Keyworth alternative
To: llw@S1-A
Keyworth has reserved August for vacation. Is there someone else in that
office worth asking to speak at AAAI?
∂19-May-83 0023 JMC Second Choice
To: genesereth@SUMEX-AIM
∂19-May-83 0006 LLW@S1-A Second Choice
Received: from S1-A by SU-AI with TCP/SMTP; 19 May 83 00:05:52 PDT
Date: 19 May 83 0007 PDT
From: Lowell Wood <LLW@S1-A>
Subject: Second Choice
To: jmc@SU-AI
CC: LLW@S1-A
∂18-May-83 2314 JMC@SU-AI Keyworth alternative
Received: from SU-AI by S1-A with TCP/SMTP; 18 May 83 23:14:35 PDT
Date: 18 May 83 2311 PDT
From: John McCarthy <JMC@SU-AI>
Subject: Keyworth alternative
To: llw@S1-A
Keyworth has reserved August for vacation. Is there someone else in that
office worth asking to speak at AAAI?
[John: Perhaps Doug Pewitt, his Assistant Director for General Science,
who is the OSTP guy in charge of masterminding the Government's response
to the Japanese initiative (among many other matters). Doug is smart,
opinionated, Washington-street-wise, and probably a live-wire speaker
(though I've never heard him give a formal address); he was Deputy
Director of Energy Research in DoE prior to going to OSTP. I'd certainly
recommend him as an alternate to Jay, though he necessarily will speak
with considerably less authority. Lowell]
∂19-May-83 0140 JMC letter from Seitz
To: DFH
Please see if you can find a letter from Frederick Seitz of Rockefeller
University mentioning someone worth seeing in Taiwan.
∂19-May-83 1026 JMC
To: DFH
Please try to reschedule the AI qual for late June.
∂19-May-83 1337 JMC representing sequences by sets
To: dam%oz@MIT-MC
Thanks for your Admissible Set Theory ... .
I am puzzled by its first sentence.
Every text on set theory mentions the Wiener-Kuratowski (early 1920s)
representation of ordered pairs by sets and then goes on to show
how this allows the representation of sequences. It represents
the ordered pair (x,y) by the set {{x},{x,y}}. In case your
terminal doesn't have the curly bracket, I'll rewrite it [[x],[x,y]].
It is necessary to show that given [[x],[x,y]] we can
tell which is x and which is y. There are two cases according
to whether x and y are different or the same, but we don't
have to know which. If x and y are different [[x],[x,y]]
has two elements one of which has one element and the other
has two. x is identified as the element of the set of one element and y
is the other element of the set of two. If x and y are the same
[[x],[x,y]] has one element and its element is x and y is the
same.
Sequences are best handled as in LISP. Use the empty set
for NIL and use the ordered pair operation for cons. However, set
theorists sometimes put the last element of the sequence in the cdr
position.
Evidently this result was considered non-trivial since two
famous mathematicians have their names attached to it.
The set theorists need and use sequences. They don't make
them primitive, because their interests are in metamathematics,
and the more primitive constructions there are in a theory, the
more work it is to do the metamathematics.
An objection to the Wiener-Kuratowski construction is that
the rank of a set (the maximum ordinal corresponding to a descending
chain of elements of elements) does not put the elements of a sequence
on the same level.
Jon Barwise objects to this and to other features of ZF,
and gives a theory of "admissible sets" in a book of that title.
It has urelements also.
You have misspelled Zermelo and clumsy.
I apologize for not getting to the essentials of your
contribution, but perhaps I'll get a chance again after I return
in the middle of June from my trip to the Far East.
∂19-May-83 1718 JMC data types and natural kinds
To: phil-sci%oz@MIT-ML
I don't understand the part about "acknowledges the duality". This doesn't
mean that one can be defined in terms of the other. I also don't understand
about not offering related discoveries. I have argued that a child
learns a name and presumes (often correctly) that it names an entity
about which much more is to be learned. He does not suppose that
the original identification of an instance of the concept is a
definition of it and is ready to learn more about it including the
fact that the properties used to identify it originally don't necessary
characterize it. We believe that machines will have to be programmed
to behave in a similar manner. I do acknowledge that I don't know
that anyone has done it yet.
As to philosophical fashions, natural kinds are "in" now. Kripke's
"Naming and Necessity" mentions many of the relevant phenomena, although
I don't recall that a systematic exposition is given in those lectures.
∂19-May-83 1720 JMC
To: levitt%oz@MIT-MC
CC: dam%oz@MIT-MC
data types and natural kinds
I don't understand the part about "acknowledges the duality". This doesn't
mean that one can be defined in terms of the other. I also don't understand
about not offering related discoveries. I have argued that a child
learns a name and presumes (often correctly) that it names an entity
about which much more is to be learned. He does not suppose that
the original identification of an instance of the concept is a
definition of it and is ready to learn more about it including the
fact that the properties used to identify it originally don't necessary
characterize it. We believe that machines will have to be programmed
to behave in a similar manner. I do acknowledge that I don't know
that anyone has done it yet.
As to philosophical fashions, natural kinds are "in" now. Kripke's
"Naming and Necessity" mentions many of the relevant phenomena, although
I don't recall that a systematic exposition is given in those lectures.
∂19-May-83 2008 JMC no learning systems yet?
To: levitt%oz@MIT-MC, phil-sci%oz@MIT-MC
I have to think about whether Winston's system admits natural kinds,
but at first glance it seems so. Kripke and Putnam aren't interested
in programming. They don't have AI type scenarios, but do have human
type scenarios. Kripke's "Naming and Necessity" is full of scenarios.
Indeed philosophers rely excessively on scenarios.
∂20-May-83 0116 JMC reducing number of Altos
To: su-bboards@SU-AI
It has been proposed to reduce number of Altos to those required for
services not available on other machines. What would be the reaction
to reducing them so much that Bravo was infeasible? There would be
more terminals in offices some more maintainable replacements in
the terminal rooms. Please react to JMC (chairman facilities committee)
by June 15 with cc REG@sail.
∂20-May-83 0121 JMC
To: REG@SU-AI, LB@SU-AI
∂20-May-83 0116 JMC reducing number of Altos
To: su-bboards@SU-AI
It has been proposed to reduce number of Altos to those required for
services not available on other machines. What would be the reaction
to reducing them so much that Bravo was infeasible? There would be
more terminals in offices some more maintainable replacements in
the terminal rooms. Please react to JMC (chairman facilities committee)
by June 15 with cc REG@sail.
∂20-May-83 1143 JMC SAIL hardware schedule
To: REG
sail hardware schedule
Marty informs me that with the cleanup work required on the TCP conversion
and his vacation in July, it may be August before the new disks and tapes
are in use. This is disappointing, and I wonder if he can be helped.
However, the only qualified person I know of is you. Any chance?
∂20-May-83 1157 JMC
To: novak@SUMEX-AIM
Gordon, I don't have these names, and I'm leaving tomorrow for 3 weeks.
Dick Gabriel might be able to help.
∂20-May-83 1147 JMC sail hardware schedule
To: REG
CC: ME
Marty informs me that with the cleanup work required on the TCP conversion
and his vacation in July, it may be August before the new disks and tapes
are in use. This is disappointing, and I wonder if he can be helped.
However, the only qualified person I know of is you. Any chance?
∂20-May-83 1403 JMC
To: sharon@SU-SCORE
I have the tickets, but in the future if you can catch them before they
run off have them take them up one more flight to me or to Diana Hall.
∂20-May-83 1652 JMC phone conversation
To: bscott@SU-SCORE
I had a phone conversation in which I announced my intentions offering
six months from June 1. The time was appreciated. I said that I would
write a letter when I returned. I announced my intention to maintain
confidentiality from my side. Please do the same.
∂20-May-83 1655 JMC
To: feigenbaum@SUMEX-AIM, buchanan@SUMEX-AIM,
rindfleisch@SUMEX-AIM
Please maintain confidentiality.
∂20-May-83 1655 JMC
To: RPG
Please maintain confidentiality.
∂21-May-83 0153 JMC (→15677 13-Jun-83)
To: "#___JMC.PLN[2,2]"
I will be in the Far East until 1983 June 12 or 15. Diana Hall (DFH@SU-AI)
can locate me if necessary.
∂13-Jun-83 0000 JMC Expired plan
To: JMC
Your plan has just expired. You might want to make a new one.
Here is the text of the old plan:
I will be in the Far East until 1983 June 12 or 15. Diana Hall (DFH@SU-AI)
can locate me if necessary.
∂16-Jun-83 1232 JMC Facilities committee
To: DFH
∂24-May-83 1049 KARP@SUMEX-AIM Facilities committee
Received: from SUMEX-AIM by SU-AI with PUP; 24-May-83 10:49 PDT
Date: Tue 24 May 83 10:50:29-PDT
From: Peter Karp <KARP@SUMEX-AIM>
Subject: Facilities committee
To: jmc@SAIL
I believe I am now the student member of the facilities committee. Thus
please add me to the appropriate mailing list. Thanks,
Peter
-------
∂16-Jun-83 1237 JMC
To: roode@SRI-NIC
David: The key thing is the personnel budget. Tymshare charges a lot,
because only 12.5 percent of their costs are for running the machines.
The rest are for handholding, sales, etc. I don't know exactly what is
included in what, and my information is five years old.
∂16-Jun-83 1240 JMC
To: TGD@SU-AI
It's Minsky, and I'll bring in my copy if you haven't already found one.
∂16-Jun-83 1241 JMC Re: role of logic in AI
To: DFH
∂31-May-83 2011 @MIT-MC:BLOCK%MIT-OZ%MIT-MC@SU-DSN Re: role of logic in AI
Received: from MIT-MC by SU-AI with TCP/SMTP; 31 May 83 20:11:24 PDT
Date: 31 May 1983 2306-EDT
From: BLOCK%MIT-OZ%MIT-MC@SU-DSN
Subject: Re: role of logic in AI
To: JMC@SU-AI
In-Reply-To: Your message of 18-May-83 0031-EDT
I notice you refer to your "Ascribbing Mental Qualities to Machines".
How can I get a copy?
Ned Block (ned@MIT-COGS@MIT-MC)
-------
∂16-Jun-83 1245 JMC
To: ullman@SU-SCORE
I'll take charge of successor problem.
∂16-Jun-83 1246 JMC prolog
To: warren@SRI-AI
∂10-Jun-83 1026 DUMAS%SUMEX-AIM.ARPA@SU-SCORE.ARPA prolog
Received: from SU-SCORE by SU-AI with TCP/SMTP; 10 Jun 83 10:26:14 PDT
Received: from SUMEX-AIM by Score with Pup; Fri 10 Jun 83 10:29:43-PDT
Date: Fri 10 Jun 83 10:26:15-PDT
From: Jean-Pierre Dumas <DUMAS@SUMEX-AIM>
Subject: prolog
To: mccarthy@SCORE
Dear Dr. Mc Carthy,
Can you tell me what and where about finding a portable Prolog
interpreter...is there any available (free or almost).
Are the prolog interpreters mainly compatible or not ?
I have heard that the performances are not as good as INTERLISP,
as i think that INTERLISP is aquite mediocre system , i wonder.!!!
Any information on prolog will be very much welcomme.
Jean-Pierre Dumas.
-------
∂16-Jun-83 1303 JMC
To: DFH, JK, RPG, CG, LGC, YOM, JJW
I'm back and will be in tomorrow. Meanwhile 857-0672 if necessary.
∂16-Jun-83 1950 JMC
To: CLT
I'll be there. Forget not Teller reception and dinner that evening.
∂16-Jun-83 2027 JMC
To: JK
Let's talk tomorrow then.
∂17-Jun-83 2326 JMC
To: genesereth@SUMEX-AIM
I forgot when. Was it 11:30 on Wednesday?
∂17-Jun-83 2339 JMC
To: pack@SU-SCORE
I just got back from the Far East. Here is the message referred to.
∂18-May-83 0031 JMC role of logic in AI
To: phil-sci%oz@MIT-ML
Since my views have been cited by DAM and others, I get to state them.
1. First order logic is logically universal in that any other logic
can be formalized withing first order logic by introducing appropriate
entities as objects. Specifically, second order logic is so formalizable,
and set theory formalized in first order logic is more powerful than
any high order logic. "More powerful" must be taken in a technical
sense that involves similar restrictions on what can be objects in
order for this statement to make sense.
2. This doesn't mean that only first order logic should be used.
Indeed my current formulation of circumscription is in second order
logic. Reformulating it in first order logic would probably be
clumsy.
3. In order to justify using a more elaborate logic than first
order logic, whether it be higher order or modal, there has to be
a good reason. Many of the candidate modal logics fail, because
they don't admit enough meta-reasoning within the logic for
practical purposes, so that certain first order formalisms are
actually more powerful. My paper on First Order Theories of
Individual Concepts and Propositions has examples of this.
4. Logic can be used in AI in four ways.
a. The systems most committed to logic do their computation
by reasoning in logical systems. I don't know any way to do this
very efficiently, but I also know no proof that it is impossible.
(I see it as very important to formulate precise assertions about
the computational power or lack of it in various formalisms).
b. A slightly weaker system uses logic at the outer
level, but can use arbitrary programs to generate answers
that are then converted into sentences by reflexion principles.
This is arbitrarily powerful, because logical deduction could end up
doing nothing but blessing the results of the computations. Naturally,
this isn't the intent of such systems, which I think have considerable
promise.
c. A system which is basically something else, e.g.
productions, could use logical deduction as part of its computation.
Thus STRIPS used deduction for reasoning within a state and
handled state transitions in another way. Presumably the human
mind works this way, since we seem to use what amounts to logical deduction
for a substantial amount of our conscious thought, but it is very
improbable that the intellectual mechanisms we share with animals
are very deductive.
d. As Newell emphasized in his 1980 AAAI presidential address,
and as I emphasize in my Ascribing Mental Qualities to Machines,
even if a system uses neither sentences for representing information
or deduction as computation, it may often be conveniently described
at a "logical level". Daniel Dennett makes similar points in his
Herbert Spencer lectures when he talks about "the intentional stance".
I choose to emphasize the epistemological part of the problem
in my own work, because I do that much better than writing programs.
If DAM disagrees with these views, I would like to read specific
criticisms.
Later added note:
Circumscription uses second order logic, because it seems
clearest for expressing non-monotonic reasoning. However, if we
want a system that reasons ABOUT non-monotonic reasoning, it may
be better to find some way of pushing circumscription down into
first order logic.
∂17-Jun-83 2348 JMC
To: pack@SU-SCORE
I don't know what to recommend beyond the references I cited in
my course. You might look at Newell's 1980 presidential address
and papers by Bob Moore. If you are interested in specific topics
come see me. Also if you want to do some (for the present unpaid)
work in the areas.
∂18-Jun-83 0000 JMC
To: ned%mit-cogs@MIT-MC
I just saw your message after return from trip. She'll send you a
paper copy of "Ascribing ... " if you send my secretary DFH@SU-AI
a U.S. mail address".
∂18-Jun-83 0046 JMC AAAI speaker
To: genesereth@SUMEX-AIM
If Kennedy declines, how about Cooper or Delauer?
∂18-Jun-83 0114 JMC
To: minker.umcp-cs@UDEL-RELAY
Panel looks fine to me.
∂18-Jun-83 0117 JMC
To: bundy@RUTGERS
Minker circumscribed correctly.
∂18-Jun-83 1103 JMC circumscription
To: bundy@RUTGERS
It is a rule of conjecture not a rule of inference; therefore confirmation
is often appropriate.
∂18-Jun-83 1538 JMC
To: DFH
7. Ask Diana to send Dialnet papers to Richard C. T. Lee at Tsinghua
University in Taiwan.
∂18-Jun-83 2153 JMC Please send reprint
To: DFH
"Common business communication language" to Takasu. It's in my reprint
drawer and is one of two with a German cover. Incidentally I usually
refer to it as CBCL.
∂19-Jun-83 0741 JMC Ramsey to Sato
To: JK
Masahiko Sato at the Computer Science Department of Tokyo University should
be sent the Ramsey proof together with whatever is required to understand it.
He still thinks that Automath is the height of proof checking. I told him
that checking Landau would be easy in EKL. I have had second thoughts, but
I still think it's true. It would be worth having someone do in order to
make the comparison clear.
∂19-Jun-83 1220 JMC Salamin co-ordinates
To: RWG
I have a paper on computation of pi to 4 million places by Kanada of
Tokyo University. If Gene doesn't have it, I'll send it to him. What
are his co-ordinates? You're welcome also or alternatively.
∂19-Jun-83 1223 JMC correctness of program for computing pi or e
To: boyer@UTEXAS-20
You might consider this problem. A priori it requires a notion of pi or
e as a real number and proving that when the program claims to have
given n places, the answer is within 10↑-n of the real numbe.
∂19-Jun-83 1418 JMC
To: feigenbaum@SUMEX-AIM
Was that Hilts?
∂19-Jun-83 1703 JMC
To: CLT
Branka sends her regards and looks forward to seeing you again in N.Y.
∂19-Jun-83 2334 JMC lisp on perq
To: fahlman@CMU-CS-C
I have no real interest in the matter, but when I was in Singapore last
week I was told by someone working for ICL that they were selling perqs
there. I was asked about Lisp and was told it wasn't yet available. I
could imagine that the pipeline between CMU and Singapore is quite long.
Is Lisp available yet on Perq from CMU point of view and if so when did
it become so. This is mere curiosity on my part.
∂20-Jun-83 0207 JMC Universality of first order logic
To: dam%oz@MIT-MC, phil-sci%mc@MIT-OZ
I have been in the Far East and have just looked at the
discussion of expressibility in first order logic. The point
I thought I was making in my original message seems to have
been mainly missed. DAM's example of the transitive
closure of a relation provides a good example of what I meant.
Suppose R(x,y) is a transitive relation. The transitivity
is expressed by the first order formula
(x y z)(R(x,y) & R(y,z) → R(x,z)).
In second order logic we can define the transitive closure of a relation.
We use the auxiliary second order relation covers(R',R) defined
by
(R R')(covers(R',R) iff (x y)(R(x,y) → R'(x,y)))
and then define tc(R',R) by
(R R')(tc(R,R') iff transitive(R') & covers(R',R)
& (R")(transitive(R") & covers(R",R) → covers(R",R'))).
This asserts that R' is transitive and covers R and is the
least relation that does so.
Suppose we don't want to use second order logic, i.e. we
don't want to quantify over relations. We can introduce a domain
of objects that we will call Relations, i.e. we reify (make things
out of) relations. Relations are first order
objects. We introduce a predicate applies(r,x,y) whose intended
meaning is that the Relation r applies to the pair x and y.
We define a predicate Transitive by
(r)(Transitive(r) iff (x y z)(applies(r,x,y) & applies(r,y,z)
→ applies(r,x,z))
and a predicate Covers by
(r r')(Covers(r',r) iff (x)(applies(r,x,y) → applies(r',x,y))).
We write the analogous formula for the predicate Tc denoting
transitive closure, namely
(r r')(Tc(r,r') iff Transitive(r') & Covers(r',r) &
(r")(Transitive(r") & Covers(r",r) → Covers(r",r'))).
This represents our attempt to make first order objects
out of relations. However, as DAM is no doubt ready to point out,
we haven't quite won. There is nothing that requires that our
domain Relations reifies all relations. Thus for some relation
R, there might not exist r in Relations such that
(x y)(applies(r,x,y) iff R(x,y)).
Therefore, our definition of Tc(r',r) is inadequate in that the
minimum transitive object in Relations covering r may not
correspond to the transitive closure of the relation
R defined by
(x y)(R(x,y) iff applies(r,x,y)).
Thus our first order theory of Tc is incomplete.
However, second order logic isn't such a big win either, because
second order logic itself is incomplete. Not every formula true
in all models of second order logic is provable in second order
logic. This contrasts with the completeness of first order logic.
This tradeoff between incompleteness of theories and incompleteness
of the logic is unavoidable.
Now we come to questions of strategy. Suppose we are
interested in formalizing a robot's ability to compute certain
relations, i.e. we want to introduce a predicate
can-compute(Robby,R) where R is a relation. This is a second
order theory and (for example), not suitable for ordinary resolution
theorem proving. An alternative is to use Can-compute(Robby,r),
where r is taken from the domain of Relations. This may be
advantageous, especially if we are really interested in some
restricted set of binary relations.
A rather powerful choice is to use Zermelo-Frankel
set theory. This is a first order theory in which the objects
are sets and the one relation of the theory is x in y. We can
form ordered pairs and sets of ordered pairs, so relations can
be introduced as sets of ordered pairs. Instead of R(x,y), we
write
(x,y) in R.
The computationally difficult part of ZF for theorem provers
is to make them use the axiom schema of comprehension and the
somewhat more difficult axiom schema of replacement. The schema
of comprehension permits is
(x)(Ey)(z)(z in y iff z in x & Foo),
where Foo is any formula with z as its free variable.
This allows us to form a set out of those elements of the
set x that have the property expressed by the formula Foo.
Frege's original formulation allowed forming a set of those
all objects that had a property Foo, but Russell showed that
if you took Foo to be not(z in z), you could quickly derive
a contradiction. Making the computer invent properties Foo
relevant to solving problems or deriving theorems is difficult -
even more difficult than making the computer invent suitable
instances of the axiom schema of induction in arithmetic or
Lisp theory. Only Boyer and Moore do much with the latter, and,
to my knowledge, no-one has written a program that invents
formulas for doing comprehension. We won't have powerful AI
until someone does.
In the meantime it is may be worthwhile to use systems
weaker than set theory in which only certain kinds of sets
are used as objects. Whether we do it in first or second order
logic is a question of technical convenience, but it may be
quite important in making a program practical.
An example is the blocks world in which above(x,y) is
taken to be the transitive closure of on(x,y). If all we want
is above, we don't need anything fancy. We simply write
(x y)(on(x,y) → above(x,y)),
(x y z)(above(x,y) & above(y,z) → above(x,z))
and use the schema
(x y)(on(x,y) → P(x,y)) & (x y z)(P(x,y) & P(y,z) → P(x,z))
→ (x y)(above(x,y) → P(x,y)).
The schema is more tedious to use than the simple axioms, and
it is often unnecessary. Specifically, if we want to prove
above(a,b) in some case, we can usually get by with the axioms
only. However, if we want to prove not above(a, b) or to
prove formulas of the form (x y)(above(x,y) → . . .), then
we'll have to use the schema.
However, if we want as system that might invent above
as the transitive closure of on, then we need something stronger.
We are forced either make our programs use higher order logic or
to reify a domain of relations that will include those we want,
e.g. will include at least on and above.
On some other occasion, I hope to go on at greater length
and more systematically about when it is convenient to reify
various domains of entity.
∂20-Jun-83 0208 JMC
To: fahlman@CMU-CS-C
Thanks for the information.
∂20-Jun-83 0215 JMC
To: ∧fH
Please U.S. mail commun.ns[s83,jmc] to Larry Tesler at Apple Computer.
∂20-Jun-83 1302 JMC
To: ef@MIT-ML
gunkel[s83,jmc] About Patrick Gunkel's proposed meeting
Patrick Gunkel proposes A CONFERENCE AND WHITE PAPER ON
"ARTIFICIAL INTELLIGENCE AND THE FUTURE OF AMERICA" in a document
dated 1983 May 25. In my opinion the proposed conference is
probably not a good idea on account of the following circumstances.
1. Artificial intelligence (AI) is indeed as potentially revolutionary
as Gunkel says. It is also important to preserve America's lead
in the field.
2. AI is in a state where some useful applications are
possible, and many companies are hoping to cash in. However,
the state of the science limits the applications that can be
developed now. For example, robotic servants wait future
fundamental scientific discoveries. Let us compare artificial
intelligence with nuclear physics. In 1938 Rutherford could say
that he saw no possibility of nuclear energy for military use or
electric power. In 1939 when Fermi's 1933 experiments were correctly
interpreted as having exhibited fission, many scientists in many
countries immediately saw the possibility of a chain reaction
leading to bombs and power plants. There were differences of opinion
about whether a determined effort could develop bombs in time to
affect World War II. Thanks to Szilard's leadership, the U.S. was
the only country that drew the correct conclusion and mobilized
the necessary effort.
The situation in AI is quite different. There is no present
scientific basis for a Manhattan Project in AI, just as there was not before
fission was discovered. We cannot tell whether the situation
corresponds to that of 1938 or that of 1905 just after Einstein
published E = Mc↑2 as a consequence of the special theory of relativity.
(Perhaps the potential of nuclear energy wasn't really confirmed until
the masses of isotopes were know and mass defects were computed).
AI differs from nuclear physics in another respect that may or may
not be relevant in the present context. Namely, so far as we know,
we don't have to observe a natural phenomenon like fission. Instead
AI will be a construction of the human mind; we have only to program
computers sufficiently well. In this respect AI is like mathematics
rather than like physics in that the phenomena to be understood,
the relation between situations and the actions required to achieve
goals, are logical in character. It also resembles engineering in
that the point is to build something that works.
3. Given these facts, I don't like to encourage the establishment
to start Manhattan Projects. They are too likely to resemble the
nuclear airplane project of the 1950s that eventually collapsed in
failure and inhibits new attempts now that the technology has advanced.
Moreover, the ratio of AI people working on applied projects is
already too high. there is no more effort going into basic theoretical
and experimental AI research than there was in 1970, and the basic
theory is advancing rather slowly.
In my opinion the financial state of AI is as follows:
a. Good students who want to work in AI can find places
in graduate school and financial support. Too much of this support
is attached to applied and pseudo-applied projects. These are
projects that promise practical results a very few years but
almost never deliver. The demand for short range payoff that peaked
in Government research support had a more harmful effect in AI than
in almost any field.
b. Good PhDs are getting jobs, but again they have to make
promises of quick payoff in order to get research support.
4. The highest financial priority for AI is more research
money for unsolicited proposals judged solely on scientific merit
and not on adherence to some plan. I regard the DARPA speech
recognition project as having been mainly harmful, because it
focussed almost all American speech recognition research into
trying to satisfy the committee thus stifling diversity. It was
the very best committee, but the resuls were still harmful. A
committee planned research effort in general AI would be even
more harmful no matter who was on the committee. Einstein and
Planck and Schroedinger and Heisenberg weren't planned, and such
scientists are what AI needs.
5. The second priority is for postdoctoral fellowships
that will enable the holders to sit and think or write programs
according to their own choice. Therefore, it would be best if
the fellowships were to be awarded nationally rather than to
give more research associates to existing principal investigators.
Some of them should be awarded to smart people trained in other
fields who want to transfer to AI. AI is an attractive field
and many people with PhDs in other fields want to transfer to it.
6. The third priority is bricks and mortar for computer
science departments and endowed chairs. Because computer science
has expanded recently and will continue to expand, it is far
behind more established fields in its facilities. Providing
computers is also important, but this need has been recognized
and is being met.
7. The current enthusiasm for AI in companies and
governments is based on the expectation of a quick payoff.
Unless some fundamental discoveries are accidentally made,
these hopes are likely to be disappointed and AI will be
regarded as a fad. I oppose establishing institutions that
require quick applied payoff. Scientific discoveries should
be the criterion for continuing support of research projects
and research institutions. AI should be regarded as genetics
was before genetic engineering appeared or as like fundamental
physics.
8. Someone should identify the basic research in AI
that is going on and figure out how to increase it.
Here are some comments about some particular points of
the Gunkel draft.
9. "Its purpose would be to exploit the present critical
moment to reshape the general conception and course of artificial
intelligence research in the United States, . . .". I don't
want my research reshaped by any conference, and I don't want
to take part in reshaping anyone else's. I prefer to influence
other people's research by publishing papers and not by getting
my hands on their sources of financial support.
10. "Those countries that lead the world in AI are apt
to lead the world in other terms as well: in industrial growth,
world trade, per capita wealth, quality of life, science and
technology, education, social and political evolution, military
power, and cultural progress". This is seriously misleading.
Human level AI would indeed have all these effects, but that doesn't
seem to be in the cards immediately. In the meantime, AI should
not be regarded as a cure for all problems. Each of the above
matters requires separate efforts.
11. "AI researchers should seek from the outset to achieve
not merely artificial intelligence but artificial humanity: 'mechanical'
forms of emotion, purpose, imagination, creativity, character,
conscience, and kindness". This is a very bad idea of Gunkel's.
We want tools not artificial slaves who would eventually have
to be liberated. Fortunately, it isn't presently necessary to
harangue on the matter, since the possibility is far off. If it
were close, I would have to abandon my research and crusade against
Gunkel's recipe for disaster.
There are five more pages to comment on, but I'd rather
do research in AI. If the conference is held, I'll come lest
some bad grand plan come out of it. Of course, I would be pleased
if the conference were to recommend the measures proposed in
my points 4, 5 and 6.
∂21-Jun-83 1102 JMC
To: JRP
How about Thursday or Friday?
∂21-Jun-83 1430 JMC
To: JRP
I have made a reservation for three at noon on Friday. Which 1979
paper do you mean?
∂21-Jun-83 1432 JMC
To: JRP@SU-AI, grosz@SRI-AI
Duplicate to JRP, because I forgot Barbara the first time.
I have made a reservation for three at noon on Friday. Which 1979
paper do you mean?
∂21-Jun-83 1655 JMC conditional pass
To: DMC
By the beginning of Fall quarter, please satisfy me of your knowledge
of Boyer-Moore as in Handbook, STRIPS as in Handbook and basic paper,
planning section of Handbook, situation calculus as in McCarthy and
Hayes and its relation to STRIPS, Tom Mitchell's Generalization as
Search.
∂21-Jun-83 2333 JMC
To: cheriton@SU-HNV
jmc - On the one hand, I don't think Kennedy was criticizing either
venture capitalists or software designers. He was merely criticizing
what he sees as a faddish journalistic emphasis on computers, etc.
compared to all other parts of the economy. On the other hand, if
he means to urge more smart people to go into the Government, that's
regrettable. There are enough Government lawyers and bureaucrats.
However, I also think that it wouldn't matter much if the Kennedy
did have strange views about software designers. There is no reason
to get excited if the president of the university doesn't agree
with us about something. The Administration seems to be reasonably
sympathetic to expanding computer science, and that's the most
important thing from the point of the need for software designers,
etc.
∂21-Jun-83 2336 JMC
To: reid@SU-GLACIER
jmc - From what is quoted, I don't think Kennedy was criticizing either
venture capitalists or software designers. He was merely criticizing
what he sees as a faddish journalistic emphasis on computers, etc.
compared to all other parts of the economy. On the other hand, if
he means to urge more smart people to go into the Government, that's
regrettable. There are enough Government lawyers and bureaucrats.
However, I also think that it wouldn't matter much if the Kennedy
did have strange views about software designers. There is no reason
to get excited if the president of the university doesn't agree
with us about something. The Administration seems to be reasonably
sympathetic to expanding computer science, and that's the most
important thing from the point of the need for software designers,
etc.
Serves you right for going to graduation.
∂21-Jun-83 2341 JMC rms
To: RPG
While I suspect your characterization is correct, messages such as yours
should not be part of the Common Lisp discussion file. In the heat of
a controversy such remarks may be reasonable, but remember that they'll
still be in the file a year hence. I would suggest that you edit it
out.
∂22-Jun-83 0720 JMC Shackleton for Festschrift
To: llw@S1-A
I didn't know about the Festschrift or this idea might have occurred
to me sooner, but maybe it's not too late. When Edward spoke of going
to the moon again, it occurred to me that you and Rod and I might write
up Shackleton in its present state for the Festschrift. It would provide
an obscure but respectable place from which to push it, and from which
Edward could push too if his initial favorable reaction holds up.
I could have a draft in two weeks if that were reasonable or a week if
it made the critical difference.
How about the 29th or the 5th or 6th for my next visit to the Lab?
∂22-Jun-83 0841 JMC letter to Creary
To: DFH
When the xgp is working - presumably later today - please spool
creary.xgp[let,jmc]. You may have to try two or three times to get
a clean copy. Please be sure not to leave any copies lying around.
∂22-Jun-83 0901 JMC
To: DFH
Please remind me to dicatate thankyou notes to Lee and Takasu, Sato, and Igarashi.
∂22-Jun-83 0901 JMC
To: DFH
yeo.1
∂22-Jun-83 0913 JMC
To: nilsson@SRI-AI
1983 May 29
Conference on a request for proposals (RFP) for a new computer science
time-sharing system.
The decision by Digital Equipment Corp. to abandon the project to make
a compatible successor to the 2060 presents the computer science community
in general and the AI community in particular with a problem and an opportunity.
The 2060 has become the standard time-sharing system for computer science
research, and this has greatly helped exchange of programs, especially compilers.
This has been especially important recently because of the experimentation with
new languages and new variants of old ones. The prospect is that those
organizations needing new time-sharing systems will often go in a variety of
directions, the results will be unsatisfactory and much of the present
possibilities for exchange will be lost. Of course, some people believe
new time-sharing machines will not be required, and these need read no further.
! The opportunity is to continue the degree of compatibility that has
been achieved and even increase it. We can try for an improved system in
various ways, and we can ask the manufacturer to provide certain commonly
wanted facilities.
Because different organizations will have different amounts of money
and different time-scales, full unity will not have been achieved. Moreover,
some organizations will find deals with specific manufacturers tempting, either for
financial reasons or in order to influence features of the design or even
because of connections between decision makers and manufacturers. However, any
unity that is achieved will make future machines more suited to the needs of the
AI and computer science research communities as well as for student use. Our
community has enough weight to be influential if we use it. D.E.C. won big
by paying attention while designing PDP-6.
!Here is my list of desiderata:
Support large number of users
No small address hack
Full duplex
Operate out of screen editor
Any exec commands executable from a terminal can be done from file
or other source of a stream of characters.
Manufacturer provides IP/TCP and ARPA net support
Manufacturer provides Common Lisp
Manufacturer provides multi-processor support and puts it in Common Lisp
I prefer queue based multi-processing.
Range of compatible machines--down to microprocessor when possible.
∂22-Jun-83 0917 JMC strange mail
To: bledsoe@UTEXAS-20
I received some U.S. mail from you that puzzled me. It is a
collection of memos about recruitment in computer sciences
stapled together on the fronto of which is written in your
handwriting "copies to the chair committee". Perhaps it was
sent to the ATP prize committee by mistake?
∂22-Jun-83 1005 JMC
To: DFH
see.1
∂22-Jun-83 1523 JMC
To: bscott@SU-SCORE
Monday is impossible. I'll be in San Jose all day. Also Tuesday.
∂22-Jun-83 1555 JMC
To: DFH
In view of his inexperience in the area he will be working, I have
decided on the lower rate.
∂23-Jun-83 0301 JMC
To: CT@SU-AI
I'll be in both Thursday and Friday. Catch me if you can.
∂23-Jun-83 0301 JMC
To: tucci@SU-SCORE
I'll be in both Thursday and Friday. Catch me if you can.
∂23-Jun-83 1443 JMC
To: RSC
>Kahn, K.M.\UNIFORM - a language based upon unification which unifies(much of
*LISP, PROLOG, and ACT I\Univ. of Uppsala\1981.
>***Kahn, K.\An Actor-Based Computer Animation Language\MIT\unknown. (GEN)
>Kahn, D.\Intermission-Actors in Prolog\Uppsala Univ.\1981.
∂23-Jun-83 1821 JMC
To: llw@S1-A, rah@S1-A
I'll come the 29th - and the 6th also if it seems worthwhile.
∂23-Jun-83 1824 JMC making life complicated for ourselves
To: faculty@SU-SCORE
Once upon a time, there was a programming problem attached to the
comprehensive. Grading it was some pain for the committee, but the
students passed or didn't as the case may be. In response to a
desire for change, that was abolished and a system was instituted
which I haven't bothered to fully understand, but which seems to
involve the student submitting a project to an advisor who then
endorses it to a committee for decision. (Whether this decision
is final I don't know; perhaps it can be appealed to the Supreme
Court). Recently my advisee Chris Tucci, while I was away submitted
an apparently weak project. Zohar (at the request of my secretary
Diana, did me and the student a favor, and after some scrutiny, endorsed
it over to the Committee, which (I imagine) passed the buck to a student
member, who found it inadequate. I gather that Gio has been pondering
the matter and will come to a statesmanlike decision on Friday. To
my count ten people have been involved in deciding whether Tucci has
met the requirement. Almost none of this could be counted as advising
a student or carrying out any other educational situation.
Let me relate this to the fact that while I delivered about
twice as many lectures per week during my trip to the Orient as I
do when teaching, I nevertheless accomplished more research than
I usually do.
Let's stop complicating our lives unnecesarily. In particular
let's go back to the Comprehensive programming problem that doesn't
require the attention of ten people and many net messages in order to
determine whether one student has met the requirements.
∂23-Jun-83 1840 JMC
To: ullman@SU-SCORE
Sorry I didn't make the meeting today. I was meeting out.
∂24-Jun-83 2008 JMC
To: bledsoe@UTEXAS-20
I have no opinion on AMS vs. M.I.T.
∂24-Jun-83 2009 JMC our discussion
To: CLT
While, as I said, I didn't understand your presentation, I still hope to.
The fact that many others did and liked it is encouraging to me and
I'm sure encourages you. My difficulty in understanding abstract mathematical
presentations is my own problem, not really yours. I think you are ahead of
other people in this area. I worry that my difficulties may discourage you,
and I hope it won't. Don't let what I said at the end about not being
able to get people to work on proving facts about concrete problems worry
you either. Somehow the problem will be solved. The baleful influence
of Perlis and Lipton is still active.
It still may be worth your while to figure out a presentation that I
can follow better. There are many who are no better at it than I am.
∂26-Jun-83 0954 JMC your paper
To: LGC
Another look at your "Commonsense Factual Reasoning: Representation,
Epistemology and Process" has confirmed my opinion that the ideas are
worth pursuing and that they should be pursued elsewhere than in my
research group. I can say good things about it should that become
relevant. Your name did not come up at yesterday's lunch.
∂26-Jun-83 1755 JMC
To: RWW
It will be fine with the footnote according to Penny and Ed.
∂27-Jun-83 0001 JMC
To: HST
I will be at Karlsruhe, and it would be interesting to visit Erlangen.
Just before and just after the meeting are both feasible at present.
∂27-Jun-83 0010 JMC
To: HST
Let me tentatively make it the 15th, i.e. the Monday after IJCAI,
but coming on the weekend. I may want to change it later so let
me know when it should be definite.
∂28-Jun-83 1719 JMC
To: bscott@SU-SCORE
I was never a P.I. on an NSF grant through the University of Oklahoma.
I think they have asked about this before with the same reply.
∂28-Jun-83 1725 JMC
To: nilsson@SRI-AI
I'll be at Alouette
∂28-Jun-83 1738 JMC
To: BOSACK@SU-SCORE
I would like to talk with you about CSD-CF.
∂28-Jun-83 2240 JMC Job: Cantonese-speaking computer programmer
To: jbr@S1-A
∂27-Jun-83 1415 GONSALVES@SU-SIERRA.ARPA Job: Cantonese-speaking computer programmer
Received: from SU-SIERRA by SU-AI with TCP/SMTP; 27 Jun 83 14:15:04 PDT
Date: Mon 27 Jun 83 14:18:50-PDT
From: Tim Gonsalves <Gonsalves@SU-SIERRA.ARPA>
Subject: Job: Cantonese-speaking computer programmer
To: su-bboards@SU-SIERRA.ARPA
A Cantonese-speaking computer programmer is wanted for a short-term job
paying consulting rates. Call Ned Chapin at 854-1567
-------
∂29-Jun-83 1728 JMC
To: DFH
I'll have titles and abstracts for Randell tomorrow (Thursday).
∂30-Jun-83 1149 JMC
To: DFH
Please find out from Betty how much foreign travel we asked for. Make
a budget dividing it evenly between importing Keith Clark from
Imperial College in London
Masahiko Sato from Tokyo University, my attending a meeting of IFIP
working group 2.2 in Europe (location unknown at present) and somone
attending AISB (artificial intelligence and simulation of behavior
conference in Europe). This budget is pro forma only as I understand
it.
∂30-Jun-83 1638 JMC
To: aaai-office@SUMEX-AIM
I would like a single from Sunday night through Friday night. I have
no special food or beverage preferences.
∂30-Jun-83 1949 JMC
To: jmc@S1-A
ferrite
∂30-Jun-83 2339 JMC
To: HST
That would be fine, but in that case it would be best to make it
the weekend after IJCAI. Walking and seeing old castles would
be very interesting, and I thank you much for the invitation.
If it makes a difference, just previous to IJCAI is also possible.
∂01-Jul-83 1345 JMC Parallelizing LISP
To: ullman@SU-HNV
CC: RPG@SU-AI
I be glad to talk about parallelization of LISP next Thursday, and
I'll invite Weyhrauch. Weyrauch I don't know.
It won't take long to tell the little I know, so there will be
plenty of time for discussion.
Abstract: Consider parallel processors (e.g. S-1) all looking at
a single large memory. There are at least two approaches to using
such a system for a single problem. In one a compiler finds opportunities
for parallelism, either because the parallelism is indicated or by
cleverness, and produces a program that explicitly uses a certain
number of processors. In our approach, however, the compiler produces
code that executes forks by starting on one task while putting a
list of the others on a queue. When processors become free they
take tasks from the queue. Not even the object program knows what
processors will execute what code or even how many processors are
available. We consider language extensions to LISP for allowing
and controlling such queue based parallel processing. Not much
has been done, so the presentation will be very short. The files
MULTI[S83,JMC]@SAIL and MULTIP[S83,JMC]@SAIL contain some preliminary
studies emphasizing the S-1.
∂01-Jul-83 1416 JMC
To: ullman@SU-HNV
That's fine.
∂02-Jul-83 1956 JMC
To: JMC
alouette 327-4187