perm filename W84.OUT[LET,JMC] blob
sn#748938 filedate 1984-04-01 generic text, type C, neo UTF8
COMMENT ⊗ VALID 00002 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002
C00003 ENDMK
C⊗;
∂09-Jan-84 1100 JMC
To: bscott@SU-SCORE
Kahn out of country till Jan 30. I'll call Ohlander.
∂09-Jan-84 1414 JMC last visit and next
To: simonds@CMU-CS-C
On the whole I think we accomplished we needed to. Please keep me
informed about how the budget is coming.
My French bank account is
Societe Generale 000 5 816242 2
Marseille Prado
305 Avenue du Prado
13269 Marseille CEDEX2
tel. 76.45.45
I had additional expenses on this trip as follows:
round trip to SFO by car: $10.00
taxi home from SFO: $58. (I got gypped on this one).
As you will remember CMU paid for our tickets, but I am entitled to
one round trip executive class. We both went economy class.
Also the Centre paid for our entire hotel bill which included some
meals. I signed some expense account, but I suppose it isn't quite
correct.
The present plan is for me to come again in the last week in March.
I would arrive on Sunday March 25 and leave on Saturday March 31.
This is contingent on there being something to discuss, i.e. on there
being a budget. If there is reason to do so, I could also visit
Marseilles on that trip. However, most important in my mind is to
have a meeting in Paris of as many of the principals as possible
including JJSS, Mike, yourself and Raj if he can come.
∂09-Jan-84 1415 JMC Paris trip
To: reddy@CMU-CS-C
CC: simonds@CMU-CS-C
A reasonable plan was agreed to last Wednesday at a meeting that
included me, Mike Griffiths, Todd and Pierre Gravier. Assuming
the budget its approved by then, I would plan to come to Paris
again for the week of March 26-30. Then it would be good to have
a meeting of all principals of electronic project including JJSS
and yourself if you can make it.
∂09-Jan-84 1604 JMC
To: ROD
What is situation concerning Shapiro colloquium?
∂09-Jan-84 1720 JMC
To: stan@SRI-AI
∂09-Jan-84 1617 ROD
I sent a note to Dianna last week saying it is fine. I assumed she
passed it on to whoever needs to know. It will be 4:30pm, Terman
Auditorium, Tuesday Jan 24th. I need to know his citizenship/immigration
status and social security number to arrange for an honorarium.
∂09-Jan-84 1743 JMC
To: stan@SRI-AI
CC: ROD@SU-AI
Please send the info to ROD@SAIL when you get it.
∂10-Jan-84 0855 JMC
To: gravier@CMU-CS-C
Files arrived ok. Many thanks.
∂10-Jan-84 1030 JMC
To: tajnai@SU-SCORE
Wu is Prof. and Ginsberg is Dr., and I have another Guo Weide from
Peking. He isn't Dr.
∂10-Jan-84 1607 JMC industrial lectures
To: faculty@SU-SCORE, su-bboards@SU-AI
The faculty has voted to continue the program next year. Please
encourage applications to teach a one quarter course. The application
should consist of a course description suitable for inclusion in
the Stanford catalog together with as much vita as the applicant
wishes considered. Payment will be 1/16 of the lecturers annual
salary with a maximum of $3,000 for a one quarter course.
∂10-Jan-84 1610 JMC
To: rms@MIT-MC
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 at all, 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 cannot realistically be expected.
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 the PDP-6 in 1963 and 1964. The DEC 20 is still selling 20
years later.
!Here is my list of desiderata. I am prepared to argue for them, but
I will be surprised if we get all of them.
Support large number of users doing large problems using large files.
We still don't really know where all the computer time goes, but there
are indications that many operations take times proportional to some
things that they should not be proportional to. For example, the
time required to start editing a page of a file may depend on the size
of the file and even on the number of other files the computer has open.
Such problems are suspected on both D.E.C. and IBM computers and probably
exist on others.
Personnel requirements:
Machine is optimized for operation without operators. D.E.C. 2060s
are often operated that way, but it is not clear that this was an
actual design objective, and many organizations do provide operators.
However, whenever the actual expenses of a computing center is calculated,
this item can dominate the amortization cost of the machine.
Organizations willing to accept a standard mode of operations should
be able to do without full time system programmers.
No small address hack. Most machines that are likely to be proposed have
adequate address space.
Full duplex. I put this down, so that IBM wouldn't consider that VM stands
a chance of satisfying the community, but it is better to be more explicit
and say that there are a variety of reasons why no flavor 370 or VM is likely
to win with this community.
Operate out of screen editor. The full capabilities of a screen editor
are useful in composing commands. All output should come back in
an editable page so the user can control what remains on his screen
and have access to his output text for composing the next input.
Any exec commands executable from a terminal can be done from file
or other source of a stream of characters. It is also necessary
that programs be able to send input to the operating system.
Manufacturer provides IP/TCP and ARPA net support
Manufacturer provides Common Lisp. Others will doubtless want Pascal, C,
and Prolog and maybe even Ada.
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.
Virtual machine facility. Basing the whole time-sharing systems on
virtual machines down to the level of card equipment has proved losing.
A machine wide file naming system where programs can refer to any
file is required. However, a virtual machine facility that permits
debugging operating system changes while the operating system is running
would be a virtue - though probably not a decisive virtue.
The editor, the displays, and keyboards (as well as possible) support
arbitrary character sets.
Manufacturer provided versions of major document compilers, i.e.
TEX and SCRIBE.
!IBM ACIS (from Branscomb)
S. M. Elder
Director of Development
Academic Information Systems
1241 East Main Street
Stamford, CT 06904
(203) 359-7430
S. A. Carns
Director of Marketing
(as above)
∂10-Jan-84 1633 JMC
To: aaai-office@SUMEX-AIM
I'll try to finish it this week.
∂10-Jan-84 1703 JMC
To: DFH
still no word
∂10-Jan-84 1759 JMC need Washington Hotel reservation
To: DFH
Please make a reservation - prefer Sheraton. Try to call David Chudnovsky
and tell him name of hotel. If you don't get him, he'll try to call you.
∂10-Jan-84 1829 JMC McGregor
To: DFH
Please tell Virginia McGregor 415 422-7992 when I'll be available in
case she really does plan to have someone bring the form rather than
mail it as she first said she would do.
∂10-Jan-84 2236 JMC
To: DFH
newyor.1[let,jmc]
∂11-Jan-84 0026 JMC misuse of hacker
To: RPG
I have thought of a somewhat Machiavellian way to strike back.
The scenario begins with a letter from an authority figure (say me)
to a newspaper or magazine denouncing you. You are denounced for
marketing a computer game called Reporter that attacks the first
amendment. Naive teenagers playing the game get a false picture
of the American press, because they get points for such actions as
1. Asking a woman whose child has just been killed for her reactions.
2. Introducing the name "hacker" or of some other group as a shorthand
for a kind of criminal.
3. soliciting a bribe from a politician - giving it some euphemistic name
4. accusing any critic of attacking the First Amendment
5. stepping on the faces of colleagues
6. publicizing quack cures
7. encouraging rock throwers to do it again for the camera
8. having terrorists "claim credit"
9. people you agree with point out and others allege
10. Examiner headline "Boudin caught in daring Brinks caper"
11. conceal a murderer to protect a source
12. blab about what mistake crooks made that got them caught and how other
crooks can avoid it
13. sneer at astronauts and other heroes as subhuman
There are more.
My theory is that the journalistic world will easily believe that
millions of naive teenagers are being misled by your game. The letter
can be vague about whether the game actually exists and has been marketed
for the IBM pc, Atari, etc.
Of course, it someone feels like creating the game all the better.
The letter gets an indignant reply, which, however, leaves itself vulnerable
to a new attack for having the kind of ideology that Weizenbaum denounces.
If Weizenbaum could be sucked in, so much the better, but
the most entertaining controversy can be had if we operate on both sides.
This is rarely done, except occasionally by magazines themselves to
increase circulation, since effort is required and no-one is paid for
his letters to the editor.
Perhaps Don Woods would be willing to help actually create the game.
The initial letter should contain enough information to permit a reader
to write to you asking for the game. My letter would explain why you
shouldn't have retaliated for the misuse of hacker and would quote and
agree with the cartoonist who wrote you.
I would agree with him in such a sugary way as would make him sick.
The best target publication will require thought, and the obvious
strategy is to start at the top and work down. The odds that the
Stanford Daily would fall for it are excellent.
∂11-Jan-84 0206 JMC
To: ARK
jmc - Sears has them.
∂11-Jan-84 1016 JMC
To: CLT
∂11-Jan-84 1012 BH51@CMU-CS-A
Received: from CMU-CS-A by SU-AI with TCP/SMTP; 11 Jan 84 10:12:00 PST
Date: 11 Jan 84 1303 EST (Wednesday)
From: Beverly.Howell@CMU-CS-A
To: McCarthy@SU-AI
Message-Id: <11Jan84.130343.BH51@CMU-CS-A>
----- Unsent message follows -----
Date: 9 Jan 84 1101 EST (Monday)
From: Beverly.Howell@CMU-CS-A
To: John McCarthy@SU-AI
Subject: Paris Trip
CC: Raj.Reddy@CMU-CS-A
Message-Id: <09Jan84.110134.BH51@CMU-CS-A>
Raj asked me to let you know that the tickets for you and your wife were
$3,767.00 special coach fare. We deducted the price of a business class
ticket ($2,578.00) for you. That leaves you owing us $1,189.00 for your
wife. Please make the check payable to Carnegie-Mellon University and
send it to me: Beverly Howell, Robotics Institute, CMU, Pittsburgh 15213.
If there are any questions please get in touch with me or with Raj.
Bev Howell
Carolyn: Please pay this bill next time you write checks.
∂11-Jan-84 1058 JMC name
To: JMC
Seyrus F. Nourani
GTE Laboratories
40 Sylvan Road
Waltham, MA 02254
617 466-2882
generalized induction, forcing, non-monotonic logic
∂11-Jan-84 1116 JMC
To: DFH
for circumscription mailing list
∂11-Jan-84 1058 JMC name
Seyrus F. Nourani
GTE Laboratories
40 Sylvan Road
Waltham, MA 02254
617 466-2882
generalized induction, forcing, non-monotonic logic
∂11-Jan-84 1119 JMC slippage
To: kjb@SRI-AI
I will slip my lectures for a week because of an announcement not getting out.
∂11-Jan-84 1140 JMC
To: dkanerva@SRI-AI
Dianne:
Here is the pub source of an announcement of some lectures
I have arranged to give. Sorry I haven't time to strip out the
commands at the moment. It is for inclusion in the CSLI bulletin.
My secretary Diana Hall, DFH@SAIL can strip them out if
it isn't obvious what isn't part of the text. I'll be away till
Friday.
.cb Four Lectures on the Formalization of Common Sense Knowledge
.<<lectur[f83,jmc]>>
John McCarthy will give four lectures on the formalization of common
sense knowledge.
The lectures will be on Fridays at 3pm.
The first will be on Friday January 20 and will be held
in the conference room of the Center for Studies
in Linguistics and Information (CSLI) conference room in Ventura
Hall at Stanford.
1. The "situation calculus". Expression of the facts about the
effects of actions and other events in terms of a function
result(e,s) giving the new situation that arises when the
event e occurs in the situation s. The frame and qualification
problems. Advantages and disadvantages of various reifications.
2. The circumscription mode of non-monotonic reasoning.
Mathematical properties and problems of circumscription.
Applications of circumscription to formalizing common sense
facts. Application to the frame problem, the qualification
problem and to the STRIPS assumption.
3. Formalization of knowledge and belief. Modal and first order
formalisms. Formalisms in which possible worlds are explicit
objects. Concepts and propositions as objects in theories.
4. Philosophical conclusions arising from AI work. Approximate
theories, second order definitions of concepts, ascription of
mental qualities to machines.
The treatments given in the lectures are new, but the material
is related to the following papers.
%3McCarthy, John and P.J. Hayes (1969)%1: "Some Philosophical Problems from
the Standpoint of Artificial Intelligence", in D. Michie (ed), %2Machine
Intelligence 4%1, American Elsevier, New York, NY.
%3McCarthy, John (1980)%1:
"Circumscription - A Form of Non-Monotonic Reasoning", %2Artificial
Intelligence%1, Volume 13, Numbers 1,2, April.
.<<aim 334, circum.new[s79,jmc]>>
%3McCarthy, John (1977)%1:
"On The Model Theory of Knowledge" (with M. Sato, S. Igarashi, and
T. Hayashi), %2Proceedings of the Fifth International Joint Conference
on Artificial Intelligence%1, M.I.T., Cambridge, Mass.
%3McCarthy, John (1979)%1:
"First Order Theories of Individual Concepts and Propositions",
in Michie, Donald (ed.) %2Machine Intelligence 9%1, (University of
Edinburgh Press, Edinburgh).
.<<aim 325,concep[e76,jmc]>>
%3McCarthy, John (1979)%1:
"Ascribing Mental Qualities to Machines" in %2Philosophical Perspectives
in Artificial Intelligence%1, Ringle, Martin (ed.), Harvester Press, July 1979.
.<<aim 326, MENTAL[F76,JMC]>>
∂12-Jan-84 2140 JMC
To: pack@SU-SCORE
How about Friday at 3:30?
∂12-Jan-84 2142 JMC
To: ullman@SU-SCORE
I guess I'd better get off CS440 list.
∂12-Jan-84 2148 JMC Opera Production: Maria Padillo
To: CLT
∂12-Jan-84 0929 cheriton@diablo Opera Production: Maria Padillo
Received: from SU-NAVAJO by SU-AI with TCP/SMTP; 12 Jan 84 09:29:13 PST
Received: from Diablo by Navajo with TCP; Thu, 12 Jan 84 09:28:36 pst
Date: Thu, 12 Jan 84 09:26 PST
From: David Cheriton <cheriton@diablo>
Subject: Opera Production: Maria Padillo
To: su-bboards@Diablo
Pocket Opera in San Francisco is performing Maria Padillo by Donizetti
on Tues. Jan. 17th in the Herbst Theatre, SF (near Opera House).
(My wife Iris is singing the leading role.)
Tickets through me or Iris are $13 for orchestra and $9 for dress circle
(and I understand $2 more directlyt from the box office.) Please let me
know by tomorrow noon if you want to purchase tickets thru me. Ideally I
would like the money tomorrow too. (Sorry for the short notice.)
If interest warrants, I will also try to organize a car pool to the performance.
∂13-Jan-84 1451 JMC
To: reid@SU-GLACIER
Please be specific about Terman deficiencies.
∂13-Jan-84 1545 JMC
To: LEP
auto[f83,jmc] Notes
circum[f83,jmc] AI Aplications of Circumscription
circum.abs[f83,jmc] Circumscription abstract for Parikh at Brooklyn College
circum.not[f83,jmc] Notes for circum[f83,jmc]
common[f83,jmc] Common sense reasoning is not natural language reasoning
forwar[f83,jmc] Doing everything by forward chaining
know2[f83,jmc] Passing the salt
tower[f83,jmc] axioms for towers
unique[f83,jmc] The unique names assumption
∂13-Jan-84 1549 JMC lists
To: ullman@SU-SCORE
I want to remain on the SUPER list. Perhaps I'm on the other
in the guise jmc-lists@sail rather than jmc@sail, since I like
to keep mail sent to lists distinct from mail sent to me
personally.
∂13-Jan-84 1611 JMC
To: DFH
Lynn (422-9373) said it was sent to my home and wants to know when we find it.
∂15-Jan-84 0059 JMC password
To: ME
Is there some way I can make my auxiliary areas impossible to log in on, i.e.
with no usable password. I alway log in on 1,jmc and alias to the other
areas.
∂15-Jan-84 1838 JMC re: "Proofs" of Fermat's last theorem (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - It is a lot of work to find the first error in an alleged proof,
especially if the terminology is vague or if correct parts are done
sketchily. I have experimented with asking people to find the error
in Kempe's 1879 wrong proof of the four color theorem. Most people
put their effort into attacking a correct but sketchy part.
I once was sent a 20 page paper containing an alleged method for
trisecting an angle with ruler and compasses. The author was a friend
of a friend of my father's, and I didn't want to be rude but reading
20 pages of geometry looking for the mistake was not my idea of a
rewarding way to spend a weekend. Fortunately, while despondently
thumbing through the paper I came upon a formula allegedly giving
sin 1/3 x as a polynomial in sin x. This formula had to be wrong, and
it had to be wrong for x = 30 degrees, since sin 10 degrees isn't
rational. So I subsituted, found the discrepancy, and wrote back,
thankful to have escaped so easily. Surely it would take him a long
time to try again and produce another 20 page printed paper giving
the result. However, I got an immediate reply
asking me didn't I know that it was illegitimate to bring up particular
cases in verifying general statements. There's a bit more, but I'll
spare you.
∂15-Jan-84 2106 JMC
To: DFH
Please phone Stanford Postoffice to see if they have Livermore package.
∂16-Jan-84 0908 JMC re: Simple database system? (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - A crude system just puts all information in one file and relies on
string search in the editor to find the information. Present computers
are fast enough to make this crude approach practical.
jmc - I'll argue that a one level system is better than a tree structured
system unless the database is too large for linear search. One doesn't
want to spend one's time re-arranging the database unless re-arranging
is one's hobby. When an item is entered, one can put as many key words
in as one can think of. A linear search of a 60 block file takes .6 sec
of computer time.
∂16-Jan-84 0912 JMC
To: gotelli@SU-SCORE
As Erbring and I agreed, he will pay for it himself.
∂16-Jan-84 0913 JMC
To: DFH
Sorry Livermore gave me no number. They told me it was registered,
but a search turned up no claim check. I guess we wait till tomorrow.
∂16-Jan-84 0956 JMC
To: reges@SU-SCORE
CC: DFH@SU-AI
The SAIL accounts for CS258 are needed because of the character
set used for the EKL interactive theorem prover and the integration
of EKL with the editor. The use of EKL on LOTS last quarter was
rather unsatisfactory. The students will also use the Boyer-Moore
theorem prover (Maclisp version) which is being supported on SAIL
only at present. There are about 12 students and I'll guess $150
apiece.
∂16-Jan-84 1115 JMC
To: JK
I would like to suggest the following sentence for the EKL manual
acknowledgement section. It would go right after the sentence about
discussions. This chain of previous efforts strongly affected the
suggestions I made and also affected FOL. "The project benefited
indirectly from experience with previous interactive theorem provers, most
recently Weyhrauch's FOL, but also from Milner's LCF, Diffie's PCHECK and
still earlier programs and papers by W. Weiher, P. Abrahams and J.
McCarthy."
∂16-Jan-84 1227 JMC
To: reges@SU-SCORE
Thanks Stuart; Diana spells her name with one n.
∂16-Jan-84 1227 JMC manual
To: boyer@UTEXAS-20
Is the old manual from SRI the best thing for a class using the prover
as part of their work (short of your book)?
∂16-Jan-84 1247 JMC re: DDN MGT Bulletin # 18 (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - Now that Sriram has told us his system, we have a reduced search
space. Seriously, the security of a system should not depend on keeping
secret aspects of the system itself. I don't see any solution that isn't
going to be a lot of trouble, and I am inclined to balance observed nuisance
from vandals against the trouble of having many passwords and changing
them often. In the long run, the builders of operating systems must
accept the problem of protecting the system and other users
from vandals who break into the directories of non-wheel users.
∂16-Jan-84 1559 JMC
To: CLT
Boyer says that "the best we can offer at the moment is the
file CODE.DOC, of which Carolyn has a copy".
∂16-Jan-84 1603 JMC
To: Solomon@MIT-MULTICS
I'll be in town that week but won't have a lot of time. Late some
afternoon except for Friday would be ok. Why don't you call when
you get in town. (office: 497-4430) (home: 857-0672). Herbert Stoyan
was rooting around in my old files and found two more early M.I.T.
memos concerning time-sharing.
∂16-Jan-84 1834 JMC
To: rem@MIT-MC
I can't think of a reference, and I don't suppose I said precisely that.
In my theoretical work I emphasized what is sometimes called extensional
properties of programs, i.e. the input-output relation.
∂17-Jan-84 2223 JMC re: DDN MGT Bulletin # 18 (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - Let me propose an elaboration of Misra's idea. Suppose we have a
universal, i.e. used by everyone, hash function of two arguments.
A user uses this function to combine his single personal password with
the name of the installation to get the password he will use on that
installation. The hash function must be information losing, so that
the personal password cannot be recovered from the name of the installation
and the user's password for that installation. Thus I need remember only
one personal password. If installation passwords that change every month
(say) are wanted, then we use a three argument password that includes the
month, e.g. 1984jan, as a third argument. If people are obliged to
use installation specified passwords on each installation, they will surely
store them in files. It would be nice if the algorithm for the hashing
the password were nicely hand computable, but a special purpose one chip
calculator could be used if this were not so. Actually this algorithm
could be built into terminals or available as a program on users' home
computers.
∂18-Jan-84 1019 JMC
To: aaai-office@SUMEX-AIM
I have a class from 1:15 to 2:05 Mon, Wed, Fri. 2:30 or even 2:15 would
be feasible.
∂18-Jan-84 1700 JMC
To: bscott@SU-SCORE
I would ask him, although I would presume it would be CS229.
∂19-Jan-84 1522 JMC
To: DFH
01-24 Tuesd, Inference Board 1pm,L.A.
∂19-Jan-84 1540 JMC re: Beware: automobile vandalism on campus (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - This hasn't happened to me in 21 years at Stanford. Maybe someone
dislikes Linda, or maybe I should resst the temptation to buy a silver
RX-7.
∂19-Jan-84 2251 JMC computational linguistics at Columbia
To: traub@COLUMBIA-20
Alice ter Meulen, a recent Stanford philosophy PhD, who is being interviewed
by the Columbia Philosophy Department on February 3, would like to know if
there is any work or interest in computational linguistics in the Computer
Science Department. She is visiting here for the next week.
∂20-Jan-84 1447 JMC re: DDN MGT Bulletin # 18 (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - Experimenting with the program led me to guess the above algorithm
except that my experiments were consistent with the use of bigrams rather
than trigrams supplemented by a special hack to exclude to give repeated
letters like aaa credit for only one good transition rather than two.
This would be even simpler to implement than Nagle's algorithm.
∂20-Jan-84 1448 JMC
To: jbn@FORD-WDL1
jmc - Experimenting with the program led me to guess the above algorithm
except that my experiments were consistent with the use of bigrams rather
than trigrams supplemented by a special hack to exclude to give repeated
letters like aaa credit for only one good transition rather than two.
This would be even simpler to implement than Nagle's algorithm.
∂20-Jan-84 1645 JMC
To: traub@COLUMBIA-20
I got your message for Alice and transmitted it. She will phone you Monday.
∂21-Jan-84 0107 JMC
To: JK
I also have an improvement in LISP definitions that will allow doing
flat and the match(pat,exp,alist) required for my class. Here it is,
but it doesn't work yet. Notice the λ-expression in newsexpinductiondef.
;;; defini.lsp[w84,jmc] EKL definition principle
(decl par (type: |ground|))
(decl atc (type: |ground⊗ground→ground|))
(decl genc (type:|ground⊗ground⊗ground⊗(ground→ground)⊗(ground→ground)→ground|))
(decl flat (type: |ground⊗ground→ground|))
(axiom |∀atc genc.
(∃fun.(∀x y par.(atom x ⊃ fun(x,par) = atc(x,par))
∧ (fun(x.y,par) = genc(x,y,par,λpar.fun(x,par),λpar.fun(y,par)))))|)
(label newsexpinductiondef)
(define flat |∀x y u.atom x ⊃ flat(x,u) = x.u ∧
flat(x.y,u) = flat(x,flat(y,u))| (use newsexpinductiondef))
∂21-Jan-84 0114 JMC
To: y.yduj@LOTS-A
You have it backwards. The criterion is that there be at least two
hard triples, i.e. that don't occur in his dictionary.
∂21-Jan-84 0120 JMC
To: jbn@FORD-WDL1
If bigrams are used, one would want to exclude only bigrams that
have a significant frequency. Doing this would probably give an acceptable
number of passwords of acceptable unobviousness. At least my experiments
with your program suggested this. Actually I'm not sure one wants to exclude
all words in the dictionary if one wants to be sure that people memorize
the passwords rather than write them down.
∂21-Jan-84 1153 JMC
To: amsler@SRI-AI
Seems like more than is presently needed.
∂22-Jan-84 0038 JMC New definition schemes
To: JK
My scheme works now. The commands are in defini.lsp[w84,jmc]. For doing flat,
mine and yours do about the same. Yours is clearly more general in that
yours will do higher types. However, if the goal is an ordinary Lisp
definition, I think they are of equivalent power. If there is a counter-example
I'd like to see it.
∂22-Jan-84 0044 JMC definitions
To: JK
I see that I still have to provide mine with a more general parameter
list. On the other hand, I don't have to define the auxiliary higher
type function first. The parameter list will then have the usual
prejudice in favor of recursion on the first argument. Perhaps there's
a way of cleaning things up to get all the advantages, since the logic
is now clear. The problem of defining partial functions is still
with us, however.
∂22-Jan-84 0117 JMC
To: JK
In my opinion, unrestricted Skolemization with its consequent axiom
of choice is a feature not a bug from any practical point of view.
∂22-Jan-84 1522 JMC
To: rms%oz@MIT-MC
Part of support of arbitrary character sets depends on hardware, but
part of it also depends on editors, etc. For example, an editor that
supports arbitrary character sets must allow characters to have
variable length internal representations. Also, although I have not
used UNIX, what I have read about its defects on BBOARD makes me
somewhat skeptical of anything based on its external characteristics.
∂22-Jan-84 1832 JMC
To: feigenbaum@SUMEX-AIM
Friday is possible but only 4:30 or later, because I'm lecturing
at IBM at 2pm. Monday is preferred. I plan to teach CS206 in
the Fall and CS226 in the Winter.
∂22-Jan-84 2232 JMC
To: feigenbaum@SUMEX-AIM
Rod is ROD@SAIL.
∂23-Jan-84 1154 JMC
To: jbn@FORD-WDL1
As it happens, I use obsolete telephone numbers. My present password
is the number of a 1952 girl friend, or at least the way I remember
that number. In my judgment this is secure enough. A rigid rule
enforced by the installation preventing the use of telephone numbers
may reduce security by encouraging the user to keep the password in
a file somewhere. I didn't know about the problem of running the
dictionary against the file of encrypted passwords, so I'll have to
think more.
∂23-Jan-84 1220 JMC
To: JK
Richard, Robin, Whitfield, William, Paul and John
∂23-Jan-84 1222 JMC
To: JK
I would like to leave as late as possible tomorrow from SF or SJO for 1pm arrival.
∂23-Jan-84 1228 JMC
To: DFH
∂23-Jan-84 1224 JK
∂23-Jan-84 1222 JMC
I would like to leave as late as possible tomorrow from SF or SJO for 1pm arrival.
-----------
Perhaps you meant this for someone else?
∂23-Jan-84 1414 JMC ekl demo
To: bosack@SU-SCORE
I need to demonstrate ekl on SAIL for a class of about 20. My idea
is to take a SAIL Datamedia to Terman and slave the monitors there
to the Datamedia. I'll need a phone to call SAIL and the ability
to slave. Do you know or can you find out if this will work. I
think we did it before with Ralph's help. I don't have a TA this
quarter, so I'll work directly with you. Wednesday or next Monday
would be good.
∂23-Jan-84 1929 JMC
To: CLT
Being now full of sushi, the answer is yes.
∂25-Jan-84 2344 JMC
To: faculty@SU-SCORE
It is my impression that with the general bureaucratization of our
procedures the faculty privilege of admitting one student has been
effectively abolished. I think it should be revived, although I have
not always been pleased with the student I got when I invoked it in
the distant past.
∂26-Jan-84 0209 JMC
To: DFH
Please enter recent files in letter[let,jmc].
∂26-Jan-84 0228 JMC
To: DFH
willia.1
∂26-Jan-84 0308 JMC seminar abstract
To: dkanerva@SRI-AI
This is for Barwise's seminar next Tuesday at 9:30.
John McCarthy
Program semantics from the applied point of view.
It is a mathematical disgrace that programs are debugged (even by
mathematicians) rather than proved correct. Moreover, it's a
practical disadvantage. A computer-checked proof that a program
meets its specifications can be finally debugged, while the debugging
of a program by test cases is never completed.
Progress has been made toward systems that can be practically used
for such proofs, and much of the known theory is relevant. The
lecture will survey the theories, the practical techniques, and
the existing systems. Some unsolved problems will be identified.
∂26-Jan-84 1600 JMC
To: aaai@SU-SCORE
1pm Feb 7 is ok for me.
∂27-Jan-84 0006 JMC re: Ph D Admissions (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - I used the "whiteball" about three times in twenty years at Stanford,
and I doubt that anyone else used it more often. It is rare that a professor
has a strong desire for a particular student, but when it happens, it should
be honored. It won't dominate the process.
∂27-Jan-84 1107 JMC my goof
To: golub@SU-SCORE
CC: reid@SU-SHASTA
It was my fault that the faculty discussion spilled over onto bboard.
Brian Reid's message was such that I thought I was on BBOARD and I used
my OBBD macro that sends messages to other bboards. Oh well, no harm
done, I believe.
∂27-Jan-84 2227 JMC
To: pourne@MIT-MC
Congratulations.
∂27-Jan-84 2232 JMC
To: faculty@SU-SCORE
jmc - I used the "whiteball" about three times in twenty years at Stanford,
and I doubt that anyone else used it more often. It is rare that a professor
has a strong desire for a particular student, but when it happens, it should
be honored. It won't dominate the process.
∂28-Jan-84 1122 JMC
To: halpern.ibm-sj@RAND-RELAY
If the seminar were held in the morning, I could come some of the
time.
∂28-Jan-84 1312 JMC
To: fagin.ibm-sj@RAND-RELAY
COMMENT | CALIBAN'S WILL by M. H. Newman. When Caliban's will was opened
it was found to contain the following clause:
"I leave ten of my books to each of Low, Y.Y. and 'Critic', who
are to choose in a certain order.
"No person who has seen me in a green tie is to choose before Low.
"If Y.Y. was not in Oxford in 1920 the first chooser never lent me
a green umbrella.
If Y.Y. or 'Critic' has second choice, 'Critic' comes before the
one who first fell in love."
Unfortunately, Low, Y.Y. and "Critic" could not remember any of
the relevant facts, but the family solicitor pointed out that, assuming
the problem to be properly constructed (i.e. assuming it to contain no
statement superfluous to its solution), the relevant data and order could
be inferred.
What was the prescribed order of choosing, and who lent Caliban
an umbrella? -- from Caliban's Problem Book - Dover 1961 |
∂29-Jan-84 2141 JMC re: Macintosh at Stanford (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - I don't believe that Stanford should have a microcomputer service
organization or should negotiate big buys. The service organization will
be an inefficient empire, and individuals, projects and departments are
better off finding the best buy available at the time a need develops.
Microcomputers are an extremely competitive industry, while SCIP, CIT
and ITS are very non-competitive. A new organization will be no
better. The microcomputer industry is equipped to service individual
buyers. I think the reduced prices of the big buys will prove an
illusory advantage, because price cutting will become general.
∂30-Jan-84 1024 JMC
To: feigenbaum@SUMEX-AIM
I will be in L.A. on Thursday, but I would like to get back on time
for the dinner for Ben Kuipers, but that depends on eventualities in
L.A. What time to you plan to begin?
∂30-Jan-84 1154 JMC
To: CLT
See biojmc[1,jmc].
∂30-Jan-84 1233 JMC
To: JK, DFH
∂30-Jan-84 1219 JK cv
RPG says he did on in TEX format for you.
Where is that?
Maybe Diana can find it.
∂30-Jan-84 1720 JMC star wars
To: su-bboards@SU-AI
n073 1629 30 Jan 84
BC-LAB 3takes
(ScienceTimes)
By WILLIAM J. BROAD
c.1984 N.Y. Times News Service
LIVERMORE, Calif. - Behind fences topped with barbed wire and doors
equipped with combination locks, dozens of young physicists and
engineers at the Lawrence Livermore National Laboratory work late
into the night, six and seven days a week, on classified projects
aimed at creating the next generation of nuclear weapons.
Their dream, they say, is to end the nuclear arms race.
In many ways they trace their lineage to another group of physicists
in the 1940s who dreamed of ending World War II. Yet they are
remarkably young for their level of achievement and for their
responsibilities.
Theirs is a milieu of blue jeans, soft drinks, an occasional
science-fiction novel - and seemingly endless, all-night bouts of
work in the lab. Offices are cluttered with books and plants. Some of
the young inventors are still in graduate school, working on their
doctorates. None have ever seen a nuclear explosion.
Their quest is to channel the energy of a nuclear detonation into
focused beams of intense radiation that travel thousands of miles at
the speed of light and destroy enemy missiles in flight, eliminating
the balance of terror that has kept an uneasy peace between the
superpowers for a third of a century. They believe that President
Reagan, in what has become known as his ''Star Wars'' speech, called
on them to speed development of such weapons as a way to help create
a defensive shield against attack from space. Their key designs
number a half dozen in all, although none except X-ray lasers and
microwave weapons have been mentioned outside the world of
government-imposed secrecy.
Critics say these complex systems based on a new generation of
nuclear arms will never work. They say an enemy could outsmart them
with countermeasures, such as ''hardening'' the skin of a missile or
simply overwhelming a defense with increased numbers of missiles,
decoys, and hard-to-detect cruise missiles. The critics often oppose
the secret nuclear projects as schemes meant to increase research
budgets and to blunt public pressure for a freeze on nuclear arsenals.
In a series of interviews, the youthful designers chided the critics
as being largely uninformed about their work at Livermore and about
the merits of defensive systems. They said lots of people worried
about the bomb, but they intended to do something about it.
''We can try to negotiate treaties and things like that,'' said
Lawrence C. West, 28 years old, who is pursuing a Ph.D. while he
works at the weapons lab. ''But one thing I can do personally,
without having to wait for arms control, is to develop the technology
to eliminate them myself, to eliminate offensive nuclear weapons.''
The designs of the young physicists have come to be known in
military circles as ''third generation'' nuclear weapons. The first
generation, built in the late 1940s and early 1950s, were large atom
bombs meant to be dropped from airplanes. The second came in the late
1950s and early 1960s with the advent of compact, high-yield hydrogen
bombs that could fit atop intercontinental missiles.
The third generation is altogether different in that the power of a
nuclear explosion is focused into tight beams of radiation that can
be directed at targets in space thousands of miles away.
Dr. Edward Teller, who helped invent the hydrogen bomb and was a
founder of the Livermore Laboratory, told Congress last year that he
knew of a half-dozen ''solid'' proposals for nuclear weapons that
could be used in a defensive shield, but could ''mention these topics
only in a superficial manner'' in open session.
''Here are the whole gamut of third-generation nuclear weapons,'' he
said. ''I am 75 years old and I am one of those of closed mind who
did not invent them. But I am blessed with some young friends who
come to my office and tell me there is something new under the sun. I
regularly throw them out saying, 'Nonsense!' But they have learned
something from me. They are stubborn. They come back with new
arguments, with new proofs, and even though slowly, I learn.''
At odds with the nuclear innovations of Teller's young friends is a
formidable array of critics, not a few of whom are veterans of the
earliest American attempts to unleash the hidden powers of the atom.
With surprising unanimity, these critics today lobby for a complete
ban on the construction and testing of all nuclear weapons. The false
promise of defensive systems, they assert, will only fuel the arms
race. Among other criticisms and questions are these:
-Rationalize how you will, bombs are ultimately meant to kill people.
When he first came to Livermore, West, the physicist pursuing his
Ph.D., had reservations about working on weapons, but eventually put
them aside.
''Nowadays I would be quite willing to go and do full-time weapons
work because I see the vast possibilities,'' he said. ''A tremendous
amount of creativity is needed, and there are very few scientists
willing to do it. Nuclear weapons can devastate the world. I
recognize that. But we are making anti-weapons. My primary interest
is not trying to find better ways to kill people, but better ways to
kill arms.''
He said, for instance, that X-ray lasers cannot be used against
cities but only against objects in space, such as speeding missiles,
because the weapon's rays will not penetrate the Earth's thick
atmosphere.
West was raised an Episcopalian and was a Boy Scout. About six years
ago he graduated at the top of his class from the California
Institute of Technology and joined Livermore. Since then he has
worked mainly on the theory and experimental design of a new
generation of supercomputers.
''This group was very exciting to me, right from the first day,''
West said. ''I could talk to most people here and have them
understand me very rapidly. I just loved it.''
-What about the pope's recent plea urging scientists to give up
their ''laboratories and factories of death?''
''I don't think I fall in that category, of working on weapons of
death,'' West said. ''We're working on weapons of life, ones that
will save people from the weapons of death.''
''It's a moral decision, and I believe in it very strongly,'' he
said. ''I can't understand why everybody in the world isn't working
on finding ways to eliminate nuclear war. Obviously, the decision to
build bombs has been there for 40 years, and we keep getting more of
them. Why not find technical solutions to a technical problem?''
-Do you ever worry that the technical solutions will fail and that
you will thus contribute to the end of the world?
''I just don't see how it could bring about a cataclysm,'' said
West. ''If you have a large system with lots of redundancy, it would
work.''
Third-generation ideas, which are being pursued by all three of the
government's nuclear weapons laboratories, first came to life at
Livermore, which is run by the University of California for the
federal Department of Energy. The facility, which has 7,200 full-time
employees, is about 40 miles southeast of San Francisco.
In a corner of the laboratory is a small cluster of buildings that
house O Group, a branch of the physics department. This is where West
says he works sometimes up to 30 hours at a stretch. This, too, is
where 50 other young scientists labor on advanced ideas in such areas
as astrophysics, supercomputer fabrication, spaceship propulsion, and
nuclear weapons design. Most of them have had educations heavy in
science and technology and fairly light in humanities. A top official
at Livermore characterized O Group as ''eccentric and extraordinarily
bright.'' The group is not the only place in the nation where people
plan third-generation nuclear weapons, but it is widely regarded as
the spark plug.
Here the average scientist is in his 20s, and few, if any, wear
wedding rings. No women are present except for secretaries. The
kitchen has a microwave oven, a hot plate, a refrigerator, and a
mountain of empty Coke bottles. Not a few of the young scientists
work straight though the night, when it is easier to monopolize the
laboratory's huge computers, some of the fastest in the world.
-Aren't special problems associated with a defense that relies on
nuclear weapons?
A veteran of the O Group is Dr. Roderick A. Hyde, 31, a graduate of
the Massachusetts Institute of Technology in astronautical
engineering who has pioneered plans for fusion drives for starships.
A senior member of O Group, he heads a section that analyzes the
technical feasibility of new ideas. One problem with a nuclear
defense, he said, is the short time available to respond to an
attack, especially because the president ostensibly controls the
release of all nuclear weapons. ''Obviously you worry about him being
shot or, even more effectively, merely kept alive but out of
communication, so there isn't a clear devolution of presidential
authority.''
-Aren't third-generation weapons really just so much speculation?
A luminary of the O Group is Dr. Peter L. Hagelstein, 29, who is
credited with major innovations. His recent Ph.D. thesis from MIT
focused on non-nuclear ways to create X-ray lasers for scientific
use. Thick with equations, the document breaksfrom
its esoteric pace at one point to list ''future applications''
suggested by science fiction books. One is ''Ringworld,'' by Larry
Niven, a tale in which a spaceship is attacked by beam weapons.
''We have been fired upon,'' cries a character in the book. ''We are
still being fired upon, probably by X-ray lasers. This ship is now in
a state of war. Were it not for our invulnerable hull, we would be
dead.''
Publicly, the best known technical triumph pioneered by Hagelstein
and O Group has been the nuclear-pumped X-ray laser, which first came
to life about four years ago as the result of a collaboration with
senior Livermore scientists such as Dr. George F. Chapline. The
weapon is being tested at the government's underground site in
Nevada. A small nuclear bomb at its core, it takes the power of a
nuclear explosion and channels it into laser rods that emit lethal
bursts of radiation. Its possible use as a weapon system is some
years off. But according to O Group scientists, clusters of X-ray
lasers will be able to put a very large dent into the entire force of
Soviet strategic missiles, currently some 1,500 strong.
The power of the X-ray laser concept resulted in the founding of a
separate group at Livermore known as R Program, a consortium of more
than 100 laboratory personnel from different groups who are
developing and testing the X-ray laser alone. The head of R Program
is Dr. Thomas Weaver, who at 34 is one of the oldest members of O
Group.
-Given the terrible risks and uncertainties, isn't working on arms
control better than constructing any kind of bomb?
''There's a simplistic view that says work on any weapon, defensive
or offensive, is intrinsically evil, and that we should lay down our
arms,'' Weaver said. ''The other position is that we are willing to
take prudent risks in order to maintain our freedom. I for one would
not argue that technological solutions alone are sufficient. I think
they need to be combined with arms control and discussions between
countries. But I think we have to be realistic. Without technical
advances to motivate the discussions, they're less likely to happen.''
-The Russians could overwhelm a defense with decoy missiles. And
even if only 2 percent of the Soviet Union's current total arsenal of
warheads broke through a defensive shield, the resulting force of
bombs would still number 200, enough to wreak havoc on this country's
major cities.
Visibly unmoved by any of the questions and criticisms is Dr. Lowell
L. Wood, 42, a Livermore physicist who founded and heads O Group. Apt
criticism is an aid in refining ideas, he said. More research was
clearly needed, he emphasized, but the potential power of the whole
spectrum of third-generation weapons was clear.
A large man with a full beard, Wood is the principal inspiration
behind the frenetic activity of the young scientists. In separate
interviews, each of them paid tribute to Wood's scientific insights,
which include advances in laser fusion and astrophysics. And they
praised his ability to build enthusiasm. ''He has lots of confidence
that no matter how young you are, whether you are fresh out of
college with a B.S. or whatever, that you can still make a
difference,'' said Weaver of R Program.
Wood is quick with replies to critics. He says, for instance, that
decoys cost about half as much as complete missiles, and that it is
economically and militarily feasible to shoot at all apparent
missiles, ignoring whether they are real or decoys.
And, even if some nuclear warheads got through a multilayer
defensive shield, he said, the Soviet Union could never be sure which
warheads could penetrate the defenses, and whether they would strike
cities or fall on missile silos in wheat fields. Such uncertainty,
Wood said, is enough to insure that the Russians would think longer
and harder before launching an attack than they need to at present.
-Couldn't the Soviet Union make X-ray lasers and use them
offensively as ''escorts'' to attacking missiles, knocking out our
defensive systems?
''Probably not,'' said Wood, ''since the defender always appears to
have the technical edge in these situations. Defensive
third-generation systems are compact and lightweight and thus have a
great capacity to carry protective armor, all the more so because
they have to fly relatively short distances. They would be able to
ride out an attack by offensive third-generation weapons and still be
able to fire at a fleet of offensive boosters. These, on the other
hand, would be quite vulnerable. Offensive missiles are very big and
can afford to carry little extra weight over the long distances they
must fly, so technically it's exceedingly difficult to armor them in
any significant way against attack.''
In any event, scientists at Livermore, young and old alike, say
research on a new generation of nuclear weaponry will continue, even
if it is never used for a defensive shield. They say it helps them
better understand the breakthroughs in this area they assume the
Russians are making as well.
The first generation of atomic physicists built the weapons that
initially shook the earth. The second generation refined them. And
now a third generation of weapons physicists has embarked on the
exploration of a new frontier.
''There're almost an infinite number of issues to be pursued,'' said
West. ''The number of new weapon designs is limited only by one's
creativity. Most of them have not been developed beyond the stage of
thinking one afternoon, 'Gee, I suppose you can do so and so.'
There're a tremendous number of ways one might defend the country.''
nyt-01-30-84 1939est
***************
∂30-Jan-84 1835 JMC re: signatures. A comment. (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - Another solution commonly recommended is to write your name
legibly as written on the check and write again your customary
signature. This is often recomended by banks.
∂30-Jan-84 1838 JMC
To: llw@S1-A, rah@S1-A
Broad's article on O-group is in tonight's NYT wire.
∂31-Jan-84 0045 JMC What if the sun goes out?
To: llw@S1-A, rah@S1-A
It occurs to me that you might find sun.out[1,jmc] at S1-A of interest
given the nuclear winter scenarios. I wrote it in 1976 and last revised
it in 1980. It probably has some inaccuracies that I would have tried
to correct had I wanted to try to publish it. The interesting question
is how long it would be before the atmosphere froze. Of course, the
nuclear winter wouldn't be as bad, although the sun.out scenario doesn't
suppose any devastation due to war.
Incidentally, they had me fill out a new PSQ and seemed to be in a hurry.
Someone actually brought it to Stanford from Livermore for me to sign. I
hope it means they'll actually start the investigation.
How is the thinking about parallel computation going?
∂31-Jan-84 0058 JMC re: Star Wars Message (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - Siegman's message is another attempt at proof by intellectual
intimidation. Kindly explain slowly and carefully your contention
that the Livermore people are insane. Explain how you can be sure.
The mere fact that this bullying line is taken by lots of authorities
will not suffice.
∂31-Jan-84 0059 JMC Star Wars Message
To: llw@S1-A
∂30-Jan-84 SIEGMAN@SU-SIERRA.ARPA 31-Jan-84 JMC Star Wars Message
Received: from SU-SIERRA by SU-AI with TCP/SMTP; 30 Jan 84 23:28:52 PST
Date: Mon 30 Jan 84 23:29:41-PST
From: Tony Siegman <SIEGMAN@SU-SIERRA.ARPA>
Subject: Star Wars Message
To: su-bboards@SU-SIERRA.ARPA
I'm acquainted with one or two of the scientists -- and Stanford
students -- mentioned in the chilling star wars news item just
reprinted here by John McCarthy.
God save us all from their insane technological hubris.
-------
jmc - Siegman's message is another attempt at proof by intellectual
intimidation. Kindly explain slowly and carefully your contention
that the Livermore people are insane. Explain how you can be sure.
The mere fact that this bullying line is taken by lots of authorities
will not suffice.
[JMC - Forwarded to OTHER-SU-BBOARDS from line 15]
∂31-Jan-84 0856 JMC
CC: siegman@SU-SIERRA
jmc - Siegman's message is another attempt at proof by intellectual
intimidation. Kindly explain slowly and carefully your contention
that the Livermore people are insane. Explain how you can be sure.
The mere fact that this bullying line is taken by lots of authorities
will not suffice.
∂31-Jan-84 1401 JMC re: star wars (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - Well, now we have some actual arguments - at least from Jeff and
Tom. You will note that my first remark about Tony Siegman's did not
discuss the issue but merely the terms of the debate - that the issue was
debatable. I expect to deal with the substance in the future, but I'm
still not satisfied with the terms of debate as assumed by Jeff and Tom.
My problem has to do with the question of burden of proof. Let us suppose
that Jeff correctly cites Kissinger as saying that MIRVs are a mistake,
and that he and others who make U.S. policy made that mistake. This
doesn't guarantee that critics of Administration defense policy are right
and that the star wars defense proposals are a mistake. Presumably the
supporters these weapons have proposals for testing or otherwise verifying
them, and these proposals will be subject to debate when deployment is
proposd. Perhaps this debate will be public, and perhaps it won't.
Presumably whether the debate will be public will itself be an issue.
In case perfection is not to be expected from either side.
One's attitude partly depends on whether one regards past
deterrent policy as a success or as a failure. On the one hand, it hasn't
removed the risk of nuclear war and made the public feel safe. On the
other hand, there hasn't been a war, so it can be said to have worked. I
take the latter view, because I can't see any way the goal of make
ourselves feel safe can be achieved. Incidentally, we have avoided war
under six administrations with at least six different strategic policies.
This suggests that the result is fortunately not so sensitive to the exact
policy as is claimed by the people who threaten us with death unless the
policy they propose is instantly adopted. This applies to both sides.
∂31-Jan-84 1630 JMC
To: DFH
Remind me again if Stark letter not done by tomorrow.
∂01-Feb-84 1702 JMC
To: SJG
1984 Feb 1
Child machine
Many people have proposed to build a child-machine. It knows little
or nothing about the world but is capable of learning everything
humans know from its experience. Presumably it can understand
language or can learn to do so. This is a plausible idea, but attempts
to realize it have not been successful, usually not even resulting
in published papers. In spite of the fact that the approach has been
so far unsuccessful, trying again is worthwhile provided one recognizes
the difficulties that have been encountered and has ideas for overcoming
them. Here is my idea of what the difficulties are.
1. Representation. Consider the very first thing the child-machine
is to learn. Whatever is learned, whether it be a fact or the appropriate
response to a class of stimuli, it must be represented somehow in the
memory of the machine. Therefore, the designer must provide for this
representation. Here he faces a dilemma. If he wants some sophisticated
fact to be representable, it seems that he must build in quite a few
concepts. But this is the problem the proposers of the child-machine
wants to avoid. The extreme avoiders of this difficulty have a tendency
to build stimulus-response machines. They have two problems. First,
they are usually forced to admit only fixed stimuli and this makes the
machine quite far from a real world child who never sees exactly the
same stimulus twice and must begin with a complex stimulus classifier.
Second, it seems that little that children learn takes a simple
stimulus-response form. Children mainly learn facts which they subsequently
use more-or-less intelligently, i.e. with the aid of inference involving
other facts.
Well, maybe I only know that one difficulty.
∂01-Feb-84 1739 JMC PROLOG Digest V2 #6
To: LEP
∂01-Feb-84 0232 RESTIVO@SU-SCORE.ARPA PROLOG Digest V2 #6
Received: from SU-SCORE by SU-AI with TCP/SMTP; 1 Feb 84 02:32:43 PST
Date: Tuesday, January 31, 1984 5:24AM
From: Chuck Restivo (The Moderator) <PROLOG-REQUEST@SU-SCORE.ARPA>
Reply-to: PROLOG@SU-SCORE.ARPA
US-Mail: P.O. Box 4584 Stanford University, Stanford CA 94305
Phone: (415) 326-5550
Subject: PROLOG Digest V2 #6
To: PROLOG@SU-SCORE.ARPA
∂01-Feb-84 1745 JMC
To: ZM
The file is mtc.syl[inf,csd].
∂01-Feb-84 1906 JMC UT hassles
To: bledsoe@UTEXAS-20
∂01-Feb-84 1844 AAAI-OFFICE@SUMEX-AIM.ARPA UT hassles
Received: from SUMEX-AIM by SU-AI with TCP/SMTP; 1 Feb 84 18:44:45 PST
Date: Wed 1 Feb 84 18:48:43-PST
From: AAAI <AAAI-OFFICE@SUMEX-AIM.ARPA>
Subject: UT hassles
To: JMC@SU-AI.ARPA
cc: aaai-office@SUMEX-AIM.ARPA
Telephone: (415) 328-3123
Postal-Address: 445 Burgess Drive, Menlo Park, CA 94025
John,
This morning I received a call from Clifford Norton,director of
Performing Arts Center. He said no one from UT informed him of the
size of our conference or the nature of its operation, and did we
know that as an outside organizwtion we are required to sign a
contract.
This contract stipulates that an audit will be performed after
the conference and any surplus funds will be returned to UT.
He is sending me a copy of the contract and the profit/loss
statement which, I assume, is the audit form. I asked him
if we could apply a certain percentage of our operating expenses
to the p/l statement and he said yes. I also told him our UT
contact, Dean Heller, never informed me that we'd have to
sign this contract, and this fact was a real suprise to us at this
late date. IF we knew this fact a year ago, we would have chosen another site.
After this shock, I called Thompson Conference Ctr (location for
the tutorials) to see if that policy applied to the operation of the
tutorial program - no, it doesn't. Thank god! However, if we use any
rooms in the Perfoming Arts Center for tutorials, those revenues are
included in the overall p/l statement. So, I advised we don't have any
tutorials in the PAC. Now that I don't trust UT support staff to
convey accurate information or at least some basic facts, I've begun
to call hotels to see if we can hold the tutorials in their meeting
rooms. No answer yet.
I talked to Richard Fikes, and it's not clear how much
money we might have to return to the hordes at UT yet.
Some unknowns (ie UT services -electicians,technicians) are
difficult to determine. We could possibly return from $10-50k
or more (and this was the year, I had developed ways to save
money for the preparation and operation of the conference!)
It's to our advantage now to spend more so that we don't
have to give to UT.
I talked to Don and Ron, and both want to move the entire show
out of Texas. My only reservation is that it may very difficult
to find another site at this late date.
Feeling disgusted,
Claudia
-------
∂01-Feb-84 1911 JMC
To: aaai-office@SUMEX-AIM
Claudia,
I phoned Woody who will see what the top administration of
the University can do. You should expect a call from him, because as
long as he is going to the top, he might as well know the complete
set of problems including those we had earlier decided to live with.
The attitude of the top administration toward computer science is such
that I'm sure they'll try to solve it.
John
∂01-Feb-84 1911 JMC
CC: bledsoe@UTEXAS-20
Claudia,
I phoned Woody who will see what the top administration of
the University can do. You should expect a call from him, because as
long as he is going to the top, he might as well know the complete
set of problems including those we had earlier decided to live with.
The attitude of the top administration toward computer science is such
that I'm sure they'll try to solve it.
John
∂03-Feb-84 1011 JMC
To: dkanerva@SRI-AI
My last lecture co-incided with a philosophy colloquium, and I want to
avoid this conflict. I would be grateful if you would announce my next
lecture, the one on non-monotonic reasoning and circumscription for the
next Friday that does not conflict with this colloquium. Please repeat
the fact that it is one of four and repeat the abstract for this lecture.
Thanks for reminding me.
∂03-Feb-84 1059 JMC
To: dkanerva@SRI-AI
Your proposed dates will be fine. Please make sure about the room.
∂03-Feb-84 1503 JMC cmu system
To: reid@SU-SHASTA
Do you know:
(1) How one resolves an ambiguity about logins at the cmu 20s?
I tried finger scott%cmu-cs-c and it told me that there were two and
gave their names, but it didn't say how their logins were distinguished.
In this it is unlike all the other fingers I have encountered.
(2) Do you know to whom I might grumble there?
∂03-Feb-84 1630 JMC Stark papers
To: fagin.ibm-sj@RAND-RELAY
His address: Richard Stark, Math Dept.,U. of South Florida, Tampa, FL 33620
papers:
1. Logics of Knowledge and their decision procedures. - to apprear in
Theoretical Computer Science.
2. A parallel process logic (billed in a note as "A totally different
approach to knowledge". I didn't understand it. This is an informal
draft and you should probably ask him for it.
3. "A Logic of Knowledge" in Zeitschr f. Math. Logik und Grundlagen d. Math.
Bd. 27, S. 371-374 (1981).
Probably it would be simplest for you to get them from him directly.
∂04-Feb-84 0637 JMC
To: reid@SU-GLACIER
thanks
∂04-Feb-84 1833 JMC starwars
To: su-bboards@SU-AI
jmc - Here are some opinions on the main issue and some side issues.
The main issue is whether the current efforts to develop a defense
against ballistic missiles, in particular those pursued by the
Livermore O-group mentioned in the NYT article, are a good idea.
My opinion is that they are for the following reasons:
1. Mutually assured defense is better than mutually assured destruction.
Any chance that it can be achieved should be pursued, even long
shots. The current budget of $250 million for this work is cheap
at ten times the price.
2. Most of the arguments that have been raised against it are in the
nature of doubts. Some answers have been offered, but the arguments will
be real only at the time deployment is proposed. The argument that
research makes deployment inevitable is simply false, as history shows.
As the article says and Reagan said in his speech, many possible systems
are being studied. My opinion is that partial defenses are better than
none at all.
3. The Soviet Union has specific characteristics that make it dangerous to
the world in general and the U.S. in particular in a way that no other
country is. The Soviet Union is really different from the U.S. in this
respect. Its main significant politics occurs within the Politburo, and
each change of leadership carries with it the danger of a nuclear
Napoleon. Its military industry gets an increasing share of the national
income, and there is no known counterbalance. Its effective secrecy
compared to our system of openness and ineffective secrecy means that
there is a substantial probability of our getting unpleasant major
surprises. Its practice of invading any country that threatens to get
away from its orbit means that Grenada is the only country so far to
escape from communism.
Fortunately, its present leaders share to a substantial though
unknown degree our perception that nuclear war would be disastrous
to both sides. Therefore, treaties are possible.
4. Some anti-defense activists express, especially when explicitly
pressed, the view that the U.S. is the main villain in the world
today. Naturally they oppose any efforts to make the U.S. militarily
stronger. Most, however, agree that the U.S. needs defense; they
merely oppose specific measures on grounds that they are unnecessary
or promote an arms race. Some will express support for increasing
conventional strength. Others support defense in general but
name no specific defense measure they support leading to certain
doubts about their sincerity among us reactionaries. Moreover,
this is mixed in electoral politics. A politician will not usually
defend the policies of electoral opponents attack even when his own
policy on the issue would not be very different. This leads to
ideologues always feeling betrayed by the people they help elect
by attacking their opponents.
In the controversies over defense, especially at Stanford, it is
very hard to disentangle the different motivations. It seems to
me that those who take an anti-U.S. position are the most single
minded and lead the others and are naturally unmotivated to
bring up matters on which they differ from their allies.
This is sometimes interpreted as the activists duping the others,
and sometimes that characterization is correct.
Side issues:
1. "Insane technological hubris". This appears to have been
dropped once challenged.
2. Where it is safe to walk the streets. Both the Soviet Union
and South Africa are relatively safe. Unfortunately, this is
not much of an index of civil liberties.
3. I wish JMM would elaborate the thought behind his
"JMM- Thank God the Soviets have nuclear weapons. I feel much better knowing
the Americans are not wreaking havoc around the world!
On a more serious note, I think the Vietnamese, El Salvadorans, Afghans, Poles,
Nicaraguans etc would tend to differ from this utopic vision."
The first paragraph was a reply to someone's "Thank God the U.S. has
nuclear weapons". However, the second paragraph leaves me puzzled.
Why does he think that the Vietnamese (hundreds of thousands of whom
have drowned fleeing communism), the El Salvadoreans (engaged in a
civil war with atrocities on both sides), the Afghans (invaded by
the Soviet Union which shot the communist prime minister and replaced
him with one they brought from Czechoslovakia), the Poles (whose
government destroyed their free trade union movement at the
instigation of Moscow under the threat of invasion), and Nicaraguans
(involved in a civil war) should have special reasons to regret that
the U.S. has nuclear weapons if that is the Utopia being referred to?
A separate answer for each country with identification of the etc
would be most illuminating.
∂05-Feb-84 0232 JMC
To: CLT
reconnect your phone.
∂05-Feb-84 1531 JMC Shostak paper
To: JK
Have you seen Robert Shostak's "Deciding Combinations of Theories",
JACM, 1984 Jan?
∂06-Feb-84 1115 JMC
To: llw@S1-A
When would it be convenient for me to phone you?
∂06-Feb-84 1500 JMC
To: avg@SU-HNV
If relative military strength were the sole consideration, this might
be so, but we aren't at war with them, and the game isn't zero sum.
Therefore, trade and some two-way sharing of science and technology
is beneficial to us as well as to both the Soviet people and the Soviet
Government. I'm not sending this to bboard now, because I intend to
reply at length to JMM after adequate thought.
∂06-Feb-84 2041 JMC
To: RPG
Please explain more about what you want and why I (of all people) should
take it up with Brachman. I haven't interacted directly with him about
AAAI conference. Why isn't for the Lisp conference anyway, of which Boyer
is boss?
∂07-Feb-84 1021 JMC
To: aaai-office@SUMEX-AIM
Claudia, there is a distinction between UT bending their rules, which is
their business, and our doing something illegal, which we shouldn't, e.g.
certifying something grossly false.
∂07-Feb-84 1617 JMC
To: JMC
dennett cognitive wheels
∂07-Feb-84 1658 JMC
To: SJG, LGC
Is "Analyzing Incomplete Information" either of yours?
∂08-Feb-84 0004 JMC
To: ARK
newmac[s83,jmc] Re meeting about new time-sharing machine
∂08-Feb-84 1914 JMC
To: DFH
If Kay Mills calls and I'm not here, you can read her common[w84,jmc].
∂08-Feb-84 2315 JMC ai reading list
To: lamping@SU-SCORE, schreiber@SU-SCORE
I note that the reading list in the syllabus in today's message includes
the AI Handbook, while Knuth's memo recommends the second edition of
Winston. I previously suggested Elaine Rich's book, but I haven't looked
at Winston's yet. I certainly prefer Rich to the AI Handbook and would
probably prefer Winston to the Handbook. Whether Winston is preferable
to Rich, I currently don't know, but I will try to form an opinion. If
other people who have read both have an opinion and a quick decision has
to be made, I'll go along with it, but I certainly prefer to replace the
the Handbook by a *single* text of reasonable length.
∂09-Feb-84 1215 JMC
To: lamping@SU-SCORE
One quarter's notice has been considered adequate in the past.
∂09-Feb-84 1515 JMC Computer Science Archive
To: su-bboards@SU-AI
Disk areas on SAIL and SCORE are being set up for the permanent storage
of theses, reports and papers. The object is to keep them as an on-line
library. The details are not yet worked out, but don't erase your reports
when you have finished printing them. Len Bosack will work out details while
the permanent system is being set up.
∂09-Feb-84 1641 JMC re: South Africa (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - In what sense do we support South Africa? We trade with them
as we do with the Soviet Union. We don't apply trade restrictions to
them, because we don't expect them to attack us. MRC doesn't remember
enough smuggling incidents if he thinks very many of them go through
South Africa.
My opinion is that South Africa will change when it is no longer possible
for a white brick layer to have two black household servants. This can
occur in one or both of two ways. First the neighboring countries can
become sufficiently prosperous so that their workers are not motivated
to take jobs in South Africa, and South African blacks are tempted to
move out - that's preferable. The other way is that it may become to
dangerous for the bricklayer to trust black servants in his house.
The result, whether anyone likes it or not and probably no-one at all
will like it, is most likely a form of Apartheid. Only dividing the
country 87 percent for whites and 13 percent for blacks will prove
infeasible as leading to too much war, and something like 50-50 or
even 75-25 will prove more viable. However, the bricklayer will
probably continue to vote for the Nationalists as long as he can
maintain his comfortable way of life. Come to think of it, since
social change is so slow, robots servants may become available world wide
in time to affect the situation.
∂09-Feb-84 1647 JMC ai reading list
To: schreiber@SU-SCORE
CC: lamping@SU-SCORE
How promptly must the faculty approve in order to make the change for the
May qual. In my opinion, and traditionally, a quarter's notice suffices.
∂09-Feb-84 1720 JMC Re: ai reading list
To: schreiber@SU-SCORE, lamping@SU-SCORE
∂09-Feb-84 1657 SCHREIBER@SU-SCORE.ARPA Re: ai reading list
Received: from SU-SCORE by SU-AI with TCP/SMTP; 9 Feb 84 16:57:04 PST
Date: Thu 9 Feb 84 16:56:17-PST
From: Robert Schreiber <SCHREIBER@SU-SCORE.ARPA>
Subject: Re: ai reading list
To: JMC@SU-AI.ARPA
In-Reply-To: Message from "John McCarthy <JMC@SU-AI>" of Thu 9 Feb 84 16:47:00-PST
John,
It's my opinion that the faculty's action is supposed to occur in the
spring and take effect with the following January comp. A student
who has read the AI Handbook this past fall would naturally be
displeased if she were obliged to now read all of Winston or Rich.
Rob
-------
Students rarely read that far in advance. Let's decide what's desirable,
and then find out if anyone objects. Anyway it probably wouldn't object
to taking the questions from the intersection of the AI Handbook and the
others.
∂09-Feb-84 1802 JMC
To: DFH
Did I bill Inference for expenses of last trip?
∂09-Feb-84 1848 JMC promise
To: shoham@YALE
I have a note that suggests I promised to send you something, but I
can't remember what.
∂09-Feb-84 2325 JMC the complaint
To: golub@SU-SCORE
I have finally examined it in detail. My opinion is that there are
a few practices he should be asked to change, along the lines mentioned
in our meeting, in order to better conform to Stanford rules and
practice. However, in general the complain seems ill-founded. I
will also mention that he is correct in saying that one of the
complainers wouldn't have passed quals without his personal
statement to me that he wanted that student to pass.
∂10-Feb-84 0029 JMC re: star wars (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - JMM is correct that I view Poland and Afghanistan differently
from El Salvador, Nicaragua and Vietnam. Let us begin with Vietnam.
At the time of our sending troops to Vietnam, our government claimed
that South Vietnam was defending itself against a Russian supported
invasion from North Vietnam, and we were helping the South Vietnamese
defend themselves. The opponents of our intervention, of which I
was one, denied this and claimed that we were intervening against
a South Vietnamese National Liberation Front which united all opponents
of a corrupt and brutal regime. Between 1968 and 1971 I changed my
mind, and I think that subsequent events have shown that the U.S.
Government was right in its assessment of the situation and that we
opponents were wrong. (1) The National Liberation Front entirely
disappeared immediately after the North Vietnamese victory. [An
interesting sidelight on Stanford attitudes is shown by a dissident
Vietnamese journalist jailed by South Vietnam and made an ASSU
visiting professor at Stanford. After the North Vietnamese victory,
it happened that the North Vietnamese put the same journalist in
jail. The ASSU dropped their making him a visiting professor.]
(2) Most Vietnamese were willing to give their North Vietnamese
conquerors the benefit of the doubt, especially as the American
media were influential in determining what Vietnamese thought
of their own country. The mass attempts to escape from Vietnam
didn't begin until two years later.
Nicaragua: The attempts to overthrow the Sandinista regime are
initiated by its Nicaraguan opponents to whom the U.S. is giving
some help. Some of them have a Somoza regime background and some
have a Sandinista background. The opponents of this help should
say what rules they think the Russians are following and what
rules they think we should follow in view of this. Because of
their fear of our help to the contras, the Sandinistas are making
gestures in the direction of free elections. If they really
follow through with it and hold the elections, then we should
recognize the results. However, we cannot abandon the contras
who depend on our aid in response to mere gestures.
El Salvador: The government there is no better than most Latin
American governments, but it did hold free elections a year or
two ago. Unfortunately for our plans, the right wingers one.
The right wing death squads seem very bad, and we are exerting
pressure for their suppression, but left wing murder has also
been going on for years, so the left cannot be granted moral
superiority.
Chile: I don't believe we did instigate or help the coup that
overthrew Allende. There was talk in the CIA, etc., about preventing
his taking office, but apparently nothing was done. Nothing should
have been done. If we did help the coup, we shouldn't have. Allende's
term ran till 1976, and the Chileans should have waited. Pinochet
offered the excuse that Allende was about to create an illegal
militia, i.e. not authorized by parliament, and I don't know how
to evaluate this.
I have given a detailed discussion of the countries mentioned
by JMM and DEW, while they offered only some random charges.
Here are some more general considerations. President Reagan referred
to the Soviet Union as an "evil empire" and later as "the focus of
evil in the world". In my opinion, the first characterization was
correct in both of its words, but the second was mistaken. There is
not a focus of evil in the world. Communism is an evil that would
exist without the Soviet Union and which would probably be even
stronger than it is in most of the world were it not for the unpleasant
examples the actual communist countries provide. After all, it wasn't
the Russians who instigated the slaughter of 2 million or so Cambodians,
and the Russians weren't responsible for the cultural revolution or
the great leap forward. There are theoretical arguments why communist
power should always lead toward a corrupt, brutal, tyrannical,
aggressive and feudal regime, but they are not very convincing. The
actual experience of communist rule of about 20 countries has been much
more convincing. [Each of the above five adjectives is intentional,
and if anyone doubts any of the five, I'm prepared to defend their
applicability].
Since World War II, U.S. administrations have pursued a variety of
policies. Some have been militantly anti-communist, and others have
tried sweetness and light. There have been two unqualified successes in
negotiation with the Russians, the Austrian peace treaty and the treating
banning atmospheric nuclear explosions. In general, however, sweetness
and light haven't worked. Nixon unilaterally dismantled U.S. biological
warfare capability, and the Russians seem to have used it in
Afghanistan. Ford tried saying that the Poles were free. Carter tried
hugging Brezhnev and giving money to the Sandinistas. There have
been numerous detentes and mini-detentes. When a leader of
a capitalist country does something or says something the Russians
like, do you know how they praise him? He is called "realistic".
Of course, most people are glad to be considered realistic and the
subject regards it as praise. What he may fail to notice is that
this gives him absolutely no credit for any good will he may have
shown. If he subsequently displeases the Russians, he has become
a mad dog.
U.S. international morality isn't perfect, but it's somewhat
better than it used to be. It is also better than that of almost
any other country, especially the Western Europeans, who want us to defend
them, but will rarely do anything much to help anyone else defend
their freedom. They have more population and more industry than
the Soviet bloc, and yet they require the permanent presence of
300,000 American soldiers.
Well, that was a long flame, and I hope it's my last for
a while.
∂10-Feb-84 1030 JMC
To: DFH
There is no consulting fee, because the arrangement has changed.
There is just round trip, $6 parking, $9 taxi, and I believe $128
air fare.
My phone doesn't work for outgoing calls either, so I call when I come
in.
∂10-Feb-84 1206 JMC AAAI chapter in Northwest
To: aaai-office@SUMEX-AIM
Miro Benda of Boeing Computer Services came to see me about forming
a chapter of AAAI up there. He'll probably call you. It seems to
me that we shouldn't have any objection. He wants no money and will
give us no money. All we have to do is agree not to sue him. We
don't have to make any effort to form other chapters but presumably
it means that we will tolerate them. I want to circularize the
executive committee about this. What is the name of the file at
Sumex to which I send the message?
∂10-Feb-84 1259 JMC
To: CLT, DFH
Home phone is now fixed.
∂12-Feb-84 1150 JMC expert systems for weather prediction
To: laws@SRI-AI
Before people make expert systems following the rules of the Farmer's
Almanac or folk maxims, they should attempt to determine whether these
methods make predictions better than random. My impression is that the
weather prediction field was strong on statistical comparisons of various
methods in the 50s or earlier. My further impression is that none of
the folk methods and few of the systematic methods showed good results.
None of the folk methods can take into account the general fact that
weather moves from West to East, because they don't use information
from anywhere but the place whose weather is being predicted. If expert
systems are to be useful, they must take into account distant observations
and probably should be built on top of extensive numerical computations.
∂12-Feb-84 1742 JMC postponement
To: JK
The class will show up in 352 tomorrow. If I can get the room and
demo for Wednesday, I'll do so. Otherwise, we'd better proceed.
Two lectures might also be a possibility.
∂13-Feb-84 1742 JMC
To: RPG
2 FCZ755 ARPA/MDA903-80-C-0102 56,8l4 5-3l-80
main ARPA
∂14-Feb-84 0038 JMC
To: HST
It seems to me that I was a part of the U.S. Algol 58 committe, but I was
not one of those who went to Europe, and I didn't take an active part.
However, it is also possible that I didn't become associated with Algol
until after the 58 meeting. I'm pretty sure that at the time I sent
the memo to Perlis and Turanski, I must have been a member of the Algol
committee. I'll look at the memo and see if I can recall something
else. At present I don't remember doing differentiation without lambda.
Note that it isn't necessary to use lambda to define maplist - only to
use it.
∂14-Feb-84 0042 JMC
To: ME
AP is garbage now. Why does this happen, and how does it get fixed?
∂14-Feb-84 0116 JMC Wolf, Goat and Grain
To: phw%oz@MIT-MC
It seems to me that figure 2-1 on page 23 of your new edition and the
accompanying text are misleading. In the figure, it is indeed obvious
how to solve the problem, but this is because the states have already
been arranged so that the problem is solved by moving to the right.
It is also obvious in the figure that goin around the cycles is
pointless, because it is clear that they are cycles. However, suppose
one only knows the states. Then it might be reasonable to represent
each state by a quadruplet (fa fo go gr) where each term is 0 or 1
according to which bank the corresponding object is. It is further
convenient to arrange the states in two parallel columns according
to the position of the farmer since that always changes. If the
states are otherwise arranged in lexicographic order, then it is
much less obvious how to proceed. In particular the cycles are
not obvious to the eye.
Thus the diagramatic representation you have chosen does not merely
represent the correct choice of states. It also represents an
ordering of the states convenient for the solution.
Regards,
John
∂14-Feb-84 1100 JMC riddle
To: su-bboards@SU-AI
What did the goat say about the recent texts in AI?
∂14-Feb-84 1525 JMC an idea
To: lenat@SUMEX-AIM
It occurs to me that that examples you give of getting new concepts
by various kinds of generalization apply to programs and not to facts.
Suppose for example that one has axioms for the integers in a suitable
form. If one throws away multiplication, order and the commutative
law of multiplication one has the concept of group. This operation
is very like the modifications of program in AM but applied to facts
rather than programs. My knowledge of AM and Eurisko is very sketchy,
so forgive me if you have already treated the problem. However, if you
haven't, you may get a large increase in power by doing so. Of course,
once one has formed the concept of group, the question arises of whether
there are any groups other than those obtained by forgetting some
properties of the integers. Your rather exhaustive methods may be
suitable for answering the question.
Regards,
∂14-Feb-84 1542 JMC prescriptive, etc.
To: holtzman@SUMEX-AIM
I have thought about the matter before, and my opinion is the same
as yours. Some philosopher, probably nineteenth century, put it
this way, "You can't derive an ought from an is". However, as soon
as you admit one ought into your axioms, then with the help of facts,
you may be able to derive many more oughts. The people who thought
they were derving oughts from ises always let one ought, e.g. survival,
sneak into their systems. The relevant technical result is presumably
much simpler. Given an arbitrary predicate, OUGHT in the present
case, we cannot derive a conclusion in which this predicate occurs
postively unless it occurs positively in some premiss. Intuitively,
it would seem that there are stronger theorems. Under suitable conditions,
we can allow rules that have P in both premiss and conclusion but
still be unable to derive any P(a) unless there is some P(x)
unconjoined with not P(y) somewhere in the premisses.
∂14-Feb-84 1543 JMC
To: greep@SU-DSN
Nope.
∂14-Feb-84 1548 JMC people
To: SJG
Joe Halpern,
find halpern in phon[1,jmc]
Lew Creary, lgc
Lofti Zadeh, UC Berkeley, fuzzy set man
Stan Rosenschein (prospector) and Bob Moore at SRI
∂14-Feb-84 1817 JMC
To: SJG
Jon Barwise, CSLI, pronounced Sicily
∂14-Feb-84 2318 JMC
To: lundstrom@SU-SCORE
Steve,
I'm interested and will call you tomorrow, probably in mid-morning.
Dick has no doubt told you about our work, originally mine and now
almost entirely his, on Lisp for multi-processors. Perhaps I'll have
more ideas in that area. I'm also interested in combining AI and
numerical computing. Edward Teller asked me, a year or two ago, to look
at a proposal by Chandrasekharan of Ohio State on an AI system for
helping reactor operators, and I concluded that the AI proposed could work well
only in conjunction with a numerical simulation of the entire nuclear
and steam system. My understanding is that other people have come to
similar conclusions. I don't have specific ideas of my own and would
be happy to listen to yours.
John McCarthy
∂15-Feb-84 0936 JMC electronic library
To: simonds@CMU-CS-C, reddy@CMU-CS-C
How did the funding for the project go? What are the prospects
for getting the principals together in late March as planned?
Should I still plan to come?
∂15-Feb-84 1026 JMC riddle
To: su-bboards@SU-AI
Winston tastes good like an AI book should.
∂15-Feb-84 1036 JMC
To: lenat@SU-SCORE
I don't understand your latest message very well, and this shows that
I understand your whole project less well than I thought. Can we get
together again, since I take it that you put off your trip?
∂15-Feb-84 1054 JMC Oxford Readings
To: perry@SRI-AI
Is the idea of including some version of "Ascribing Mental Qualities
to Machines" still alive?
∂15-Feb-84 1204 JMC
To: lamping@SU-SCORE
Do you have an opinion on how Rich compares with Winston?
∂15-Feb-84 1213 JMC AI readings
To: schreiber@SU-SCORE
CC: lamping@SU-SCORE
John Lamping and I have agreed that Elaine Rich's book should eventually
replace the Handbook for the AI part of the Comprehensive and that the
present exam should be based on on the intersection. Unless you
disagree, we propose to announce that the exam will be based on
the intersection and send an appropriate notice to bboard.
∂15-Feb-84 1253 JMC
To: JDM@SU-AI
It is in the index to my collection, but I can't find it. Therefore,
I'd like to copy yours (Diana will do it), but maybe only if you
can't find a better copy somewhere.
∂15-Feb-84 1414 Mailer failed mail returned
To: JMC
In processing the following command:
MAIL/su
The following message was unsent because of a command error:
------- Begin undelivered message: -------
∂15-Feb-84 1414 JMC 15-Feb-84 1303 PETERS@SRI-AI.ARPA Areas B1 and D4 meetings
Received: from SRI-AI by SU-AI with TCP/SMTP; 15 Feb 84 13:03:46 PST
Date: Wed 15 Feb 84 13:04:08-PST
From: Stanley Peters <PETERS@SRI-AI.ARPA>
Subject: Areas B1 and D4 meetings
To: csli-b1@SRI-AI.ARPA
Jerry Hobbs has asked that we postpone his talk from today. And
several other folks have belatedly sent me their schedules,
listing conflicts. In light of the new information, it looks
like Monday afternoons at 3:15 would be a better time than
Wednesday afternoons. Can people please give me their reactions
to this idea?
Since Jerry has to give a different talk next Tuesday, he prefers
to talk in our meetings after that. So I propose we meet Monday
Feb. 27th at 3:15 to hear him. If anyone wants to talk this
coming Monday, the 20th, they should let me know so I can
announce the meeting.
And, again, if the Monday time won't work, please let me know
right away. Thanks.
-------
peters%sri-ai
name on list
Please change my name on the csli-b1 automatic mailing list from
JMC@SU-AI to JMC-LISTS@SU-AI. I like to get mail sent to lists
in this way.
------- End undelivered message -------
∂15-Feb-84 1415 JMC
To: peters@SRI-AI
peters%sri-ai
name on list
Please change my name on the csli-b1 automatic mailing list from
JMC@SU-AI to JMC-LISTS@SU-AI. I like to get mail sent to lists
in this way.
∂15-Feb-84 1428 JMC name on list
To: JMC
Please change my name on the csli-b1 automatic mailing list from
JMC@SU-AI to JMC-LISTS@SU-AI. I like to get mail sent to lists
in this way.
∂15-Feb-84 1428 JMC name on list
To: JMC
Please change my name on the csli-b1 automatic mailing list from
JMC@SU-AI to JMC-LISTS@SU-AI. I like to get mail sent to lists
in this way.
∂15-Feb-84 2302 JMC Serial or parallel
To: ailist@SRI-AI
It seems to me that introspection can tell us that the brain
does many things serially. For example, a student with 5 problems
on an examination cannot set 5 processes working on them. Indeed
I can't see that introspection indicates that anything is done
in parallel, although it does indicate that many things are done
subconsciously. This is non-trivial, because one could imagine
a mind that could set several processes going subconsciously and
then look at them from time to time to see what progress they
were making.
On the other hand, anatomy suggests and physiological
experiments confirm that the brain does many things in parallel.
These things include low level vision processing and probably
also low level auditory processing and also reflexes. For example,
the blink reflex seems to proceed without thought, although it
can be observed and in parallel with whatever else is going on.
Indeed one might regard the blink reflex and some well learned
habits as counter-examples to my assertion that one can't set
parallel processes going and then observe them.
All else seems to be conjecture. I'll conjecture that
a division of neural activity into serial and parallel activities
developed very early in evolution. For example, a bee's eye is
a parallel device, but the bee carries out long chains of serial
activities in foraging. My more adventurous conjecture is that
primate level intelligence involves applying parallel pattern
recognition processes evolve in connection with vision to records
of the serial activities of the organism. The parallel processes
of recognition are themselves subconscious, but the results have
to take part in the serial activity. Finally, seriality seems
to be required for coherence. An animal that seeks food by
locomotion works properly only if it can go in one direction
at a time, whereas a sea anemone can wave all its tentacles at
once and needs only very primitive seriality that can spread
in a wave of activity.
Perhaps someone who knows more physiology can offer more
information about the division of animal activity into serial
and parallel kinds.
∂15-Feb-84 2307 JMC re: riddle (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - I don't know what the sheep said, but I know what the mosquito said.
∂15-Feb-84 2322 JMC Industry lecturers
To: ZM
Halpern has submitted a course description, and I have agreed to
his proposal and given the catalog description to Kathy Berg.
Please ask Rom to submit a resume, a course description suitable
for the catalog, and, if he likes, a longer description. I need
these within a week or it will miss the catalog. So far Halpern
is the only applicant, although I have a feeler from Clarence
Ellis of Xerox, but no description for the catalog. It would
be nice to have three, one for each quarter.
∂15-Feb-84 2326 JMC Industry lecturers
To: faculty@SU-SCORE
So far I have only one completed application for the industry
lectureship (Joe Halpern of IBM), and it would be nice to have
three, one for each quarter. If you know someone in industry
who is thinking of applying or should apply, they should send
me a vita, a course description suitable for the Stanford catalog
and, if they like, a longer description. We need it sometime next
week so it can meet the catalog deadline.
∂16-Feb-84 0021 JMC ellis
To: wiederhold@SUMEX-AIM
Ellis expressed interest but hasn't yet got around to submitting a
catalog description.
∂16-Feb-84 1044 JMC name on lists
To: HALPERN.SJRLVM1.IBM-SJ@RAND-RELAY
Please replace my name (JMC@SU-AI) on your seminar notification list
by JMC-LISTS@SU-AI, but continue to use JMC@SU-AI for messages not sent
to lists. Does your new and longer installation name also apply to
Moshe Vardi and Ron Fagin?
∂16-Feb-84 1047 JMC industry lecturer
To: cheriton@SU-HNV
It does, but I'd need the catalog description, vitae and optional longer
description promptly.
∂16-Feb-84 1348 JMC
To: DFH
biojmc fixed. font B added as zero30 used for accents such as umlaut.
∂16-Feb-84 1413 JMC course
To: rom.tsca@SRI-TSCA
CC: ZM@SU-AI
Thanks for the course description. I may have more applications than slots,
so I'll let you know next week or perhaps the week after.
∂16-Feb-84 1444 Mailer failed mail returned
To: JMC@SU-AI
In processing the following command:
MAIL/su
The following message was unsent because of a command error:
------- Begin undelivered message: -------
∂16-Feb-84 1444 JMC .ARPA
To: HOLTZMAN@SUMEX-AIM
ought and is
I don't agree with MEU - either as describing actual human behavior,
as a norm for human behavior, or what should be built into computer
programs. I think my objection is different from those usually given,
and it applies to any unitary ought rule, i.e. a rule deriving all
oughts from a single principle.
People appear to have a wide variety of rules, learned at different
times, determining their goals and what ought or ought not be done
to accomplish them. This can and does lead to conflicts between what
the different rules advocate. Since one of our dislikes is internal
conflict, one human goal is to discover and adopt a general rule that
will encompass all the specific rules and enable us to avoid internal
conflict and also to provide a justification for others. However, I
don't believe this can be done except for in limited models of reality
parametrized in advance. My reasons for believing this are still somewhat
vague and are related to a whole collection of ideas about non-monotonic
reasoning. However, here are a few remarks.
1. When someone formulates a general rule, it is typically criticized
by bringing up examples that the author of the rule hasn't thought of.
Often concepts that the author thought were clear turn out to be
ambiguous in the circumstances postulated in the examples.
2. That our concepts have limited applicability is the normal intellectual
situation of humans and it will be normal for machines also. We can
only formulate rules governing small (in some sense) models of
certain aspects of the world. These rules may be unambiguous
in the cases encountered in existence of a particular human or machine,
but there is no reason to suppose that they won't be ambiguous
in cases that might arise.
3. Let me propose a semi-mathematical problem. Suppose that we have
a collection of a dozen or so ought rules. Suppose that we have
a certain parametrized set of circumstances, and suppose that
the prescriptions of the set of rules are unambiguous and intuitively satisfactory
in these circumstances. Suppose now that we propose to expand
the collection of circumstances considered. It may happen that
the rules become ambiguous or that the results are intuitively
unsatisfactory. Is there necessarily a more comprehensive rule
that will give even approximately the same results in the limited
circumstances, which can be compactly formulated and which is
intuitively satisfactory. My opinion is that this is not guaranteed.
Therefore, when we invent rules of behavior for robots, we will have
provide a collection covering the circumstances we can anticipate,
and we can expect conflicts when certain unanticipated circumstances
occur.
4. But surely, you may argue, the robot will exhibit some behavior
in the unanticipated circumstances, and isn't this the general rule.
Yes, but the robot's behavior may be intuitively unsatisfactory.
In particular it may depend on adventitious aspects of the situation,
such as which horn of the dilemma was encountered first.
I don't know whether any of this is relevant to the problems you are
trying to solve, so I'll stop.
------- End undelivered message -------
∂16-Feb-84 1444 JMC ought and is
To: HOLTZMAN@SUMEX-AIM
I don't agree with MEU - either as describing actual human behavior,
as a norm for human behavior, or what should be built into computer
programs. I think my objection is different from those usually given,
and it applies to any unitary ought rule, i.e. a rule deriving all
oughts from a single principle.
People appear to have a wide variety of rules, learned at different
times, determining their goals and what ought or ought not be done
to accomplish them. This can and does lead to conflicts between what
the different rules advocate. Since one of our dislikes is internal
conflict, one human goal is to discover and adopt a general rule that
will encompass all the specific rules and enable us to avoid internal
conflict and also to provide a justification for others. However, I
don't believe this can be done except for in limited models of reality
parametrized in advance. My reasons for believing this are still somewhat
vague and are related to a whole collection of ideas about non-monotonic
reasoning. However, here are a few remarks.
1. When someone formulates a general rule, it is typically criticized
by bringing up examples that the author of the rule hasn't thought of.
Often concepts that the author thought were clear turn out to be
ambiguous in the circumstances postulated in the examples.
2. That our concepts have limited applicability is the normal intellectual
situation of humans and it will be normal for machines also. We can
only formulate rules governing small (in some sense) models of
certain aspects of the world. These rules may be unambiguous
in the cases encountered in existence of a particular human or machine,
but there is no reason to suppose that they won't be ambiguous
in cases that might arise.
3. Let me propose a semi-mathematical problem. Suppose that we have
a collection of a dozen or so ought rules. Suppose that we have
a certain parametrized set of circumstances, and suppose that
the prescriptions of the set of rules are unambiguous and intuitively satisfactory
in these circumstances. Suppose now that we propose to expand
the collection of circumstances considered. It may happen that
the rules become ambiguous or that the results are intuitively
unsatisfactory. Is there necessarily a more comprehensive rule
that will give even approximately the same results in the limited
circumstances, which can be compactly formulated and which is
intuitively satisfactory. My opinion is that this is not guaranteed.
Therefore, when we invent rules of behavior for robots, we will have
provide a collection covering the circumstances we can anticipate,
and we can expect conflicts when certain unanticipated circumstances
occur.
4. But surely, you may argue, the robot will exhibit some behavior
in the unanticipated circumstances, and isn't this the general rule.
Yes, but the robot's behavior may be intuitively unsatisfactory.
In particular it may depend on adventitious aspects of the situation,
such as which horn of the dilemma was encountered first.
I don't know whether any of this is relevant to the problems you are
trying to solve, so I'll stop.
∂16-Feb-84 1805 JMC test
To: ellis.pa@PARC-MAXC
I received your message with catalog entry. I now have more applications
than positions, so we'll have to decide. I'll let you know next week
or at latest the week after.
∂16-Feb-84 2049 JMC meeting
To: lenat@SU-SCORE
The only good time is Wednesday evening, because I will be in L.A.
Tuesday and Wednesday during the day. So how about Wednesday
evening at 8:45 in my office? The TI people got in touch, and I
agreed to visit, but they didn't mention that it would be in
connection with the forum. If they regard it as such, then I'll
accept your offer. I'll have no other source of expenses unless
they pay.
∂17-Feb-84 1034 JMC agreement
To: ATP.Barbara@UTEXAS-20, aaai-office@SUMEX-AIM
This is for Claudia Mazzetti if she's still at Texas. She'll get it
here too.
The agreement looks ok. Congratulations to you and all who co-operated.
The phrase "return it to me for your records" should read "return it to
me for our records".
∂17-Feb-84 1443 JMC
To: DFH
Pls. send copy psych. today article to John Perry.
∂18-Feb-84 0040 JMC abnormality
To: pack@SU-SCORE
It seems to me that with your axioms, minimizing ab won't lead to the
conclusion that Tweety, known only to be a bird, flies. We will
just get
ab aspect2 Tweety or ab aspect1 Tweety.
I still haven't gotten around to reading your Prolog program, but I
hope to do so this weekend. We should get together next week anyway,
though I'll be gone Tuesday and Wednesday. Thursday afternoon
after 3 might be ok if convenient for you.
∂18-Feb-84 0042 JMC
To: pack@SU-SCORE
Just got to your second message.
∂18-Feb-84 1237 JMC is-a
To: pack@SU-SCORE
I haven't checked whether your proposed is-a formalization suffers
from the same problem as your bird formalization. However, it
doesn't meet my goal in another way, and that is your use of an
ancestor function; if it were an ancestor relation, that would be
ok, but you use it as a parent function. For example, you write
ancestor(bird) = animal, but you might have written
ancestor(bird) = vertebrate and ancestor(vertebrate) = animal.
Our human inheritance algorithm, whatever it may be, does not
require a fixed parentage chain. It can deal with the situation
in which there are an unknown or even undefined number of intermediates
between bird and animal. This is the goal I tried to state.
Returning briefly to your first message, let me state the goal that
axiom systems should be devised so that circumscribing ab gives
a definite result in those cases where our informal reasoning would
give a definite result.
Please persist in your efforts to formulate axioms. A few more tries
and you'll get some I won't be able to find flaws in. At least they
won't be the same flaws that I found in some of my own previous efforts.
If there is anything unclear about these comments, I'll be glad to discuss
them either in person or by phone. My home phone is 857-0672, and I'm usually
free to talk except at dinner time.
∂19-Feb-84 2355 JMC
To: CLT
You might have said good-bye or even just said you were going.
∂20-Feb-84 0001 JMC algol history
To: HST
I would imagine that I was the person Morse sent. At least it seems to
me that no-one else went to the meetings. I'm pretty sure that I went
to at least one meeting before the delegation went. I don't know
whether Perlis is reachable; I've certainly never received any net
mail from him, and he is not in the 1980 ARPAnet directory. I'll
check the more recent directory in the office.
∂21-Feb-84 0019 JMC Festschrift
To: rah@S1-A
I'm visiting Jerry Pournelle tomorrow night to work on Shackleton
paper. If you have any information about whether an asteroid trip
would be feasible, please mail me a message. I'll also try to
phone you tomorrow night.
∂22-Feb-84 1842 JMC
To: DFH
Please send Jerry Pournelle Short Waits to his L.A. address.
∂23-Feb-84 0006 JMC
To: HST
There was no such thing as a programming language specialist, and I
was the faculty member with the most programming language interests,
although one of the programmers had worked in the Fortran group at
IBM. I believe I was already making propaganda for the idea of
an algebraic list processing language.
∂23-Feb-84 0013 JMC
To: DFH
Can you find some filed brochures from the Tatung company?
∂23-Feb-84 0100 JMC Shackleton and thanks
To: pourne@MIT-MC
Thanks for the dinner and conversation. I want to discuss it
with Lowell, who is sick, before making a commitment to give
a talk about Shackleton, but I still think it's the right thing
to do.
∂23-Feb-84 0103 JMC Shackleton paper
To: llw@S1-A, rah@S1-A
Jerry Pournelle wants me (or us) to give a talk on it at the L-5
sponsored Space Industrialization Conference in SF. I think it's
a good idea, but I want to discuss it with you first.
∂23-Feb-84 0949 JMC letter to UT
To: aaai-office@SUMEX-AIM
Claudia,
You might as well sign it off. Are the arrangements substantially all
we want or did we substantially compromise our ability to run the
conference the way want and to use it to finance our activities?
John
∂23-Feb-84 1105 JMC industry lecturers
To: rodrigue@LBL-CSAM
Only the heading of your message arrived.
∂23-Feb-84 1907 JMC
To: pack@SU-SCORE
hierar[w84,jmc] inheritance
Discussion with Leslie Pack.
It looks like the problems with variable hierarchy vanish with
enough reification. Here are some axioms developed in a 1984 Feb 23
discussion with Leslie Pack.
ordinarily(Penguin, Not Fly)
[Ec'.class is-a c' & c' is-a c & ordinarily(c',not p)] → ab aspect1(c,class,p)
ordinarily(c,p) & class is-a cc & ~ab aspect1(c,class,p) → ordinarily(class,p)
mem(x,c) & ordinarily(c,p) & ~ab aspect3(x,c,p) → ap(p,x)
mem(Joe,Adelie)
Adelie is-a Penguin
mem(x,c) & c is-a c' → mem(x,c')
c'' is-a c' & c' is-a c → c'' is-a c
mem(x,c') & c' is-a c & ordinarily(c',not p) & ab aspect3(x,c,p)
Questions:
1. Can these be used directly as a Prolog program?
2. Does this set of axioms really work? Some checking is required.
∂23-Feb-84 2153 JMC Corrected version
To: pack@SU-SCORE
mem(x,c') ∧ c' is-a c ∧ ordinarily(c',not p) → ab aspect3(x,c,p)
∂23-Feb-84 2155 JMC Corrected corrected version
To: pack@SU-SCORE
mem(x,c') & c' is-a c & ordinarily(c',not p) → ab aspect3(x,c,p)
∂23-Feb-84 2311 JMC Lisp history
To: HST
It wasn't I who had worked in the Fortran group; it was Sheldon
Best at the M.I.T. Computation Center. I mentioned him only in
connection with the question of whether Morse had any alternatives
to me as far as the Algol Committee was concerned. It still seems
that he didn't, and it may even be that the letter to Morse was
a formality, because I believe I had previously talked about various
ideas at a U. Michigan summer school that Carr ran.
∂23-Feb-84 2312 JMC
To: HST
The point being that perhaps I was already informally invited.
∂23-Feb-84 2350 JMC Dyson
To: SJG
My expectations about the Dyson article (part 1 of 4) were substantially
disconfirmed. Except for his praise of Helen Caldicott, I find little
to disagree with, and I think the article will bring many liberals around
to a more reasonable point of view.
∂24-Feb-84 0926 JMC
To: bosack@SU-SCORE
3600
∂24-Feb-84 2007 JMC
To: golub@SU-SCORE
It turns out the deadline is Wednesday.
∂24-Feb-84 2319 JMC isajmc
To: pack@SU-SCORE
I copied it into my area, because I don't know how to make Prolog
except any but one identifier file names, and tried it. It works
although ordinarily(adelie,X) generates stack overflow. It may
be interesting to try to make a precise version of Reiter's idea
that circumscription using ab converts directly into Prolog.
There will surely have to be some restrictions, which should be
made precise, and it would also be worthwhile to give some rules
for the conversion. This might be a good thing for you to
consider, although it may turn out to be a difficult problem.
Perhaps it is a lack of imagination, but I'm not sure what I
can accomplish by running the program.
∂26-Feb-84 0026 JMC phone numbers and circumscription and visit
To: uw-beaver!ubc-vision!reiter@LBL-CSAM
What are your current phone numbers?
Would you like to visit here for a couple days in April or
May to visit our non-monotonic reasoning seminar? We'll
pay expenses and a small honorarium. I can more or less
do unique names with circumscription, but there are still
some puzzles. Please acknowledge quickly so I'll know the
mail path still works.
∂26-Feb-84 0029 JMC for your mailing lists
To: CLT
While I think of it, there is a bboard message from MRC about a change
in the form of net addresses. The basic idea is that . must soon
be changed to %.
∂26-Feb-84 1126 JMC re: riddle (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - By popular demand it is revealed that the mosquito said,
"Winston tastes good like an AI author should."
∂27-Feb-84 0004 JMC chat
To: llw@S1-A
Sure, let's get together. If you're free for dinner we can do that.
Otherwise, or also, let's talk afterwards. I'll see you there in any
case.
∂27-Feb-84 1258 JMC
To: DFH
todoro.9 but it can wait till you return.
∂27-Feb-84 1446 JMC Pat Hayes Dinner.
To: hobbs@SRI-AI
I would like to come.
∂27-Feb-84 1632 Mailer failed mail returned
To: JMC
In processing the following command:
MAIL/su
The following message was unsent because of a command error:
------- Begin undelivered message: -------
∂27-Feb-84 1632 JMC course
We would like you to give your office information systems course as
part of the industrial professor program. Winter quarter is preferred.
Unless you have some other pay arrangement concerning what I hear is
a sabbatical, we would pay $3000 or 1/16 of your annual salary, which
is less. Please reply promptly, because the information is going into
the catalog.
------- End undelivered message -------
∂27-Feb-84 1632 JMC course
To: ellis.pa@PARC-MAXC
We would like you to give your office information systems course as
part of the industrial professor program. Winter quarter is preferred.
Unless you have some other pay arrangement concerning what I hear is
a sabbatical, we would pay $3000 or 1/16 of your annual salary, which
is less. Please reply promptly, because the information is going into
the catalog.
∂27-Feb-84 1635 JMC course
To: waldinger@SRI-AI
We would like you to give your program synthesis course as
part of the industrial professor program. Autumn quarter is preferred.
We would pay $3000 or 1/16 of your annual salary, which
is less. Please reply promptly, because the information is going into
the catalog.
∂28-Feb-84 0054 JMC
To: dkanerva@SRI-AI
Don't forget to put my Friday lecture in csli bulletin, please.
∂28-Feb-84 1305 JMC
To: CLT
After talking to Mike, I decided to postpone my next visit to France.
∂28-Feb-84 1305 JMC next visit
To: simonds@CMU-CS-C
As you may remember, I was planning to visit the Centre Mondial again
at the end of March. However, it would be more convenient for me
personally to postpone the trip to about the end of April, although
I could come in March. I have discussed this with Raj and with Mike,
and they see no reason why I shouldn't postpone. Raj will check further
with Jean-Jacques. Therefore, unless you see some reason why a March
trip is especially desirable, I'll postpone. The most likely reason
for coming on schedule would be if postponement would cause the project
to lose momentum, e.g. by the appearance of waning enthusiasm. Please
react by Thursday, because I would have to buy tickets by then to get
a good price.
∂28-Feb-84 1331 JMC
To: ME
ap is broken
∂28-Feb-84 1333 JMC
To: ME
I don't like to bother you redundantly. Do you usually know about AP?
∂28-Feb-84 1536 JMC
To: YOM
KK[S79,JMC] Ma's Mr. S and Mr. P
KK.PRF[S79,JMC] Ma's FOL proof of Mr. S and Mr. P
∂28-Feb-84 1557 JMC
To: SJG
In 5 minutes it will be time.
∂28-Feb-84 1609 JMC
To: YOM
MRHUG[S76,JMC] 13-Dec-76 AN EXAMPLE FOR NATURAL LANGUAGE UNDERSTANDING
∂28-Feb-84 1838 JMC shackleton
To: rah@S1-A
As I think I mentioned in a previous message, Pournelle wants a talk
on Shackleton at the San Francisco Space Industrialization Conference.
I have discussed it with Lowell, and he is agreeable to going ahead
with this. Are you also?
∂28-Feb-84 2157 JMC re: Talk: Interactive Computing & Work Stations (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - IBM has in fact lagged the users of DEC computer in the facilities
it provides for on-line work. In 1979 most people kept their terminals
in Yorktown shut off when they weren't actually using them, probably
because they made an unpleasant noise. As far as I know, the interactive
features of VM are still inferior to other systems, because IBM terminals
still provide a rather bad line editor in the terminal, and the computer
doesn't see the characters until ENTER is pressed. The reason for this
flame is that IBM people generally consider that what IBM does is the
state of the art and what other people do doesn't count until IBM does
it too. Unfortunately, the trade press often unwittingly accepts this
point of view. It would be interesting to know if the speaker on
Friday will express this point of view. I hope someone will ask some
sharp questions. Here are some samples.
Questions:
1. Has IBM repented half-duplex and the ENTER key?
2. When is IBM going to declare VM obsolete?
3. How about something really modern like running the
computer out of an editor.
4. Does IBM understand that an operating system that is
unsatisfactory by itself is still unsatisfactory when improved by
macros (they call them execs) that are individually tailored by
an expert for each user and are not described in any manual?
∂29-Feb-84 0244 JMC proof of impossibility
To: uw-beaver!ubc-vision!reiter@LBL-CSAM
I would very much like to see the proof of impossibility your student has
produced. Even more important to me is a precise statement of what he has
proved or claims to have proved. As I remarked, my attempt does something
but not everything one would want. I'll send you another message in
a few days.
∂29-Feb-84 1133 JMC re: Talk: Interactive Computing & Work Stations (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI
jmc - It isn't primarily a question of D.E.C. vs. IBM, because many of the
major operating advantages of the present D.E.C. systems were developed at
Stanford, M.I.T., Bell Labs and BBN. Indeed there are features in CP/CMS
that others should copy. However, the fact remains that IBM shows now signs
of making a break with the bad features of CP/CMS, and the abstract expresses
the common IBM belief that they occupy the Middle Kingdom and the rest of
the world is inhabited by barbarians.
Actually, the operating system research area is in bad shape everywhere
in the matter of interaction with the user and in efficiency. The problem
seems to me to be that enormous effort has been going into doing for small
machines and distributed systems what has already been done for time
shared machines. Rather than figure out how to make a genuine improvement,
people often merely re-implement the old features in a new environment.
Admittedly I should write a paper rather than a BBOARD flame.
∂29-Feb-84 1133 JMC
To: MA
May I have bboard for 10 seconds?
∂29-Feb-84 1147 JMC
To: ME@SU-AI, bosack@SU-SCORE
What's the prognosis on the KA? I have lots of pub output.
∂29-Feb-84 1509 JMC course
To: ellis.pa@PARC-MAXC
I am assuming that, as I proposed, that you agreed to teach your course
in the Fall quarter, but I forget whether you actually agreed to it.
∂29-Feb-84 2043 JMC course times
To: berg@SU-SCORE
CC: halpern.ibm-sj@CSNET-RELAY, waldinger@SRI-AI,
ellis.pa@PARC-MAXC
Please see if you can get into the catalog still the fact that Ellis will
teach in the Fall, Halpern in Winter and Waldinger in Spring.
∂29-Feb-84 2046 JMC office
To: ellis.pa@PARC-MAXC
I don't know whether we have provided office space for this year's
industrial lecturers. I understand you will be on sabbatical here,
and I suggest you talk to whoever arranged that.
∂29-Feb-84 2249 JMC welcome
To: DAVISM.ACF1@NYU
Virginia is welcome for the 11th and 12th.
(home: 857-0672),(office: 497-4430) (Carolyn office: 497-3334)
∂29-Feb-84 2252 JMC Lazowska
To: cheriton@SU-HNV
It looks like it would be a good course, but the three slots are
now filled. Perhaps the faculty could decide that yet one more
such course would be a good idea.
∂29-Feb-84 2304 JMC halpern course description
To: berg@SU-SCORE
Can you computer mail me the Halpern course description? I want to
put it in a message to the Faculty and BBOARD. Thanks.
∂01-Mar-84 0957 JMC
To: YOM
It is written.
∂01-Mar-84 1000 JMC
To: berg@SU-SCORE
Therefore, can we make Halpern's 400B to correspond to Winter?
∂01-Mar-84 1152 JMC for your information
To: halpern.ibm-sj@CSNET-RELAY
From: Kathy Berg <BERG@SU-SCORE.ARPA>
Subject: CS400
To: JMC@SU-AI.ARPA
Stanford-Phone: (415) 497-4776
I had to change some of the wording Prof. Halpern used in his course
description. The editorial office does not permit one to use
phrases like "We will discuss....", so I amended these sentences.
Here is the description as it will read in Courses and Degrees.
400B Reasoning about Knowledge. Formal Systems for modeling aspects
of reasoning about knowledge, such as modal logic, nonmonotonic logic and
relevance logic will be considered. Discussions will address to what
extent these approaches can be used to deal with such problems as
reasoning in the presence of inconsistency, belief revision, and
knowledge representation. Familiarity with mathematical reasoning and
first-order logic will be assumed.
Spring quarter (Halpern) by arrangement.
Hope this is satisfactory. Also, as it now stands, Ellis will lecture
autumn, Halpern in winter, and Waldinger in spring.
Kathy
-------
∂01-Mar-84 1155 JMC industry lecturers
To: faculty@SU-SCORE, su-bboards@SU-AI
Here are the Industrial Lecture Courses for 1984-85.
They are numbered CS400A, B and C. Each course will be
given by the named computer scientist from industry.
Each year there is a new group of industrial lecturers,
and the courses are not expected to be repeated.
Clarence (Skip) Ellis (Xerox PARC)
Office Information Systems Design.
Technology, techniques, and design paradigms of electronic office
information systems. The objective is to present a coherent and cohesive
foundation for the understanding and analysis of office systems and
their implementation. Topics include: basic components and media such as
word processors, workstations, PBXs, and local area networks; office
firmware such as RasterOps, virtual keyboards, phone handlers, and
window managers; office system elements such as document editors, mail
systems, calendaring systems, and distributed servers. The course will
describe and discuss issues of user interfaces, user programming, office
modeling, and the social / organizational structures within which the
technology must exist. Prerequisites: computer organization (e.g.
cs111,cs112), computer software (e.g. cs142,cs146).
Fall 84 only.
Joe Halpern (IBM San Jose)
400B Reasoning about Knowledge. Formal Systems for modeling aspects
of reasoning about knowledge, such as modal logic, nonmonotonic logic and
relevance logic will be considered. Discussions will address to what
extent these approaches can be used to deal with such problems as
reasoning in the presence of inconsistency, belief revision, and
knowledge representation. Familiarity with mathematical reasoning and
first-order logic will be assumed.
Winter quarter (Halpern) by arrangement.
Richard Waldinger (SRI International)
Seminar in Program Synthesis:
Recent research on the systematic derivation of programs
to meet given specifications, with an emphasis on deductive
approaches. Related topics in theorem proving, logic prog-
gramming, planning, and program transformation. Individual
projects and some student presentations.
Prerequisites: CS157 A/B or equivalent.
Spring 85 only.
∂01-Mar-84 1157 JMC
To: berg@SU-SCORE
Thanks, Kathy.
∂01-Mar-84 2307 JMC
To: restivo@SU-SCORE
The 1982 Who's Who lists his office as Suite 400, 1800 K St. NW,
Washington, D.C. 20006
∂01-Mar-84 2317 JMC
To: craig@SU-SCORE
jmc - The file MAILIN[SEN,JMC] is an ancient mailing list. The first page
contains a PUB introduction and the rest gives the list in the correct
format. The output when printed on the xgp is in a format for
xeroxing on a sheet of mailing labels. These sheets were on sale
at the Stanford Bookstore and consisted of three or four rows of
labels on an 8.5 by 11 waxed backing. PUB is available only on
SAIL, I think.
∂01-Mar-84 2318 JMC
To: FWH
May I have bboard for 10sec?
∂01-Mar-84 2322 JMC
To: HST
The discussion of circumscription included in my 1977 IJCAI paper
(not using the term, I think) was obsoleted by my 1980 paper.
Is this what he wants; it was called "Epistemological problems
of AI".
∂02-Mar-84 1443 JMC
To: pereira@SRI-AI
Would there be any interest at SRI in helping pay travel for a visit
by Veronica Dahl, probably around the end of March? She would be
coming from Kentucky.
∂02-Mar-84 1728 JMC Veronica Dahl
To: pereira@SRI-AI
1/3 to 1/2 of the airfare and a couple nights lodging plus whatever
honorarium you customarily give if she gives a talk at SRI.
∂02-Mar-84 1829 JMC blocks axioms using circumscription
To: grosof@SU-SCORE
The pub source file is circum[f83,jmc]. It is a draft and may have loose
ends in various places. In particular it is complacent about having
solved the unique names problem. The part about moving and painting
blocks is the last section. It is more or less readable in source form,
but executing pub circum[f83,jmc] and printing the file on the xgp
will produce something prettier.
∂02-Mar-84 2242 JMC
To: restivo@SU-SCORE
jmc - Here is unexpected additional address info.
a240 1530 02 Mar 84
AM-Reagan-Kissinger,50
President Names Kissinger to Head Spy Advisory Board
WASHINGTON (AP) - President Reagan announced Friday he will appoint
former Secretary of State Henry A. Kissinger to the President's
Foreign Intelligence Advisory Board.
Kissinger, chairman of Kissinger Associates Inc., recently chaired
Reagan's commission on Central America.
He was secretary of state from 1973 to 1977.
ap-ny-03-02 1828EDT
***************
∂03-Mar-84 1342 JMC
To: restivo@SU-SCORE
It wasn't clear from your message whether you have this.
The 1982 Who's Who lists his office as Suite 400, 1800 K St. NW,
Washington, D.C. 20006
∂03-Mar-84 1659 JMC cross-listing
To: TOB@SU-AI, lenat@SUMEX-AIM, genesereth@SU-SCORE,
bgb@SUMEX-AIM, feigenbaum@SUMEX-AIM
I agree with Ed.
∂03-Mar-84 1701 JMC Rod Brooks
To: bgb@SUMEX-AIM, feigenbaum@SUMEX-AIM
should be included in discussions involving AI faculty
∂04-Mar-84 1316 JMC Dave Poole
To: llw@S1-A
I heard a rumor that Foonly was going out of business and that he was
looking for a job. If there is any match between what he would like to
do and the S-1 Project, I would recommend him. Since he would be leaving
the business world for more steady income and less business hassles, he
might also be a stabilizing influence.
∂04-Mar-84 1802 JMC dover output
To: CLT
I have spooled unh[1,val]. Would you pick it up sometime before I
pick you up or you come home? I'm not in an immediate hurry to eat
and would prefer around 8pm, but if you have some other preference,
I can go along. I hope the class was fun.
∂04-Mar-84 1914 JMC pattern matching
To: CLT
PATTER.LSP[W84,JMC] has a maclisp matcher for the kind of patterns
you were discussing in class. It gains some simplicity by using
difference lists.
∂04-Mar-84 1922 JMC logout
To: VAL
When you left today, you didn't log out properly. I observe that the
job is now listed as disowned, and I don't know what you did that caused
that, but I suspect we are still paying for connect time.
∂04-Mar-84 1926 JMC
To: llw@S1-A, rah@S1-A
I see that RAH is now listed as unknown on SAIL.
∂04-Mar-84 1927 JMC unknowns
To: pmf@S1-A
pmf and jbr and rah are now unknown to sail. tm is known, although he
hasn't logged in since June.
∂05-Mar-84 1845 JMC
To: pereira@SRI-AI
She has promised to phone me in a week about dates.
∂05-Mar-84 2322 JMC answers to queries
To: HST
The paper isn't yet ready for publication. I'll put him on the list
for it when it is ready. I don't know whether ML is still operating -
it's a KA-10, and I understand they are being phased out by M.I.T. -
but Szolovits is also known as PSZ@MIT-MC.
∂05-Mar-84 2323 JMC paper
To: DFH
Please put Wolfgang Bibel on the list to receive my next circumscription
paper.
∂06-Mar-84 1016 JMC SIGLUNCH ANNOUNCEMENT -- Friday, March 9, 1984
To: SJG
∂06-Mar-84 0908 MULLEN@SUMEX-AIM.ARPA SIGLUNCH ANNOUNCEMENT -- Friday, March 9, 1984
Received: from SUMEX-AIM.ARPA by SU-AI.ARPA with TCP; 6 Mar 84 09:08:04 PST
Date: Tue 6 Mar 84 09:05:05-PST
From: Juanita Mullen <MULLEN@SUMEX-AIM.ARPA>
Subject: SIGLUNCH ANNOUNCEMENT -- Friday, March 9, 1984
To: siglunch@SUMEX-AIM.ARPA
(NOTE: We are trying out a different location this week to see
if this hall is more suitable for our purposes)
------------------------------------------------------------------
SIGLUNCH
Friday, March 9, 1984
LOCATION: Braun Lecture Hall (smaller), ground floor of Seeley Mudd
Chemistry Building (approx. 30 yards west of Gazebo)
12:05
SPEAKER: Ben Grosof
Stanford University, HPP
TOPIC: AN INEQUALITY PARADIGM FOR PROBABILISTIC KNOWLEDGE
Issues in Reasoning with Probabilistic Statements
ABSTRACT:
BACKGROUND: Reasoning with probabilistic knowledge and evidence is
a key aspect of many AI systems. MYCIN and PROSPECTOR were pioneer
efforts but were limited and unsatisfactory in several ways. Recent
methods address many problems. The Maximum Entropy principle
(sometimes called Least Information) provides a new approach to
probabilities. The Dempster-Shafer theory of evidence provides a new
approach to confirmation and disconfirmation.
THE TALK: We begin by relating probabilistic statements to logic. We
then review the motivations and shortcomings of the MYCIN and
PROSPECTOR approaches. Maximum Entropy and Dempster-Shafer are
presented, and recent work using them is surveyed. (This is your big
chance to get up to date!) We generalize both to a paradigm of
inequality constraints on probabilities. This paradigm unifies the
heretofore divergent representations of probability and evidential
confirmation in a formally satisfactory way. Least commitment is
natural. The interval representation for probabilities includes in
effect a meta-level which allows explicit treatment of ignorance and
partial information, confidence and precision, and (in)dependence
assumptions. Using bounds facilitates reasoning ABOUT probabilities
and evidence. We extend the Dempster-Shafer theory significantly and
make an argument for its potential, both representationally and
computationally. Finally we list some open problems in reasoning with
probabilities.
-------
∂06-Mar-84 1029 JMC
To: DFH
Welcome back. My France trip is postponed till the end of April.
∂06-Mar-84 1259 JMC appelt proposal
To: aaai-office@SUMEX-AIM
Let's give him the $5,000.
∂06-Mar-84 2027 Mailer failed mail returned
To: JMC
In processing the following command:
MAIL/su
The following message was unsent because of a command error:
------- Begin undelivered message: -------
∂06-Mar-84 2027 JMC book
Sorry, I don't have it.
------- End undelivered message -------
∂06-Mar-84 2028 JMC book
To: de2smith@SUMEX-AIM.ARPA
Sorry, I don't have it.
∂07-Mar-84 1754 JMC
To: pack@SU-SCORE.ARPA
Has the Department agreed to consider your late application?
∂07-Mar-84 1759 JMC
To: pack@SU-SCORE.ARPA
OK, I'll get mine in. Can you get a good recommendation from some philospher?
∂07-Mar-84 2025 JMC
To: DFH
I presume I can afford to hire Leslie Pack as Research Assitant for Spring?
∂07-Mar-84 2140 JMC
To: reid@SU-GLACIER.ARPA
I hereby whiteball Leslie Pack to 2nd round.
∂08-Mar-84 0035 JMC
To: RPG
See denelc[w84,jmc].
∂08-Mar-84 1119 JMC
To: DFH
Tell Galskoy yes, thanks. 7-0163.
∂08-Mar-84 1125 JMC abstract
To: ticsl@BBNA.ARPA
title: Some Expert Systems Require Common Sense
abstract: Human common sense involves certain knowledge and ability
that is only beginning to be identified by artificial intelligence
research. Only part of it has been formalized enough to be usable
by computer programs, and even less is involved in current expert
systems. Many current expert systems are quite impressive and even
useful without it. However, many important tasks require common sense
knowledge and reasoning ability. This lecture describes some of this
knowledge and ability and discusses what kinds of tasks require it.
∂08-Mar-84 1435 JMC
To: DFH
Please type a note for me to send to May at the Wire dept. of Bank of America
asking them to mail me the check from Sweden.
Wire dept: May, 530 Lytton, 94301
∂08-Mar-84 1745 JMC
To: DFH
That was presumably Veronica Dahl about whom Kennedy called.
∂08-Mar-84 1824 JMC
To: SGF
I may be needing the terminal in the near future.
∂08-Mar-84 2203 JMC (on TTY20) e
To: BUG-e
The message "cannot write out text in unformatted file" is unaccompanied by
any hint of how to format the file. Just naming a suitable command in the
message would be enough.
∂08-Mar-84 2214 JMC (on TTY20) e
To: BUG-e
This relates to my previous gripe. I am unable to reproduce the phenomenon
about which I complained. First, I don't know what caused the file to lose
its directory and become a one page file. After that I edited it and
created another page with Xmark. Then I left and later re-entered it,
but somehow it didn't ask me whether it could reformat it or whether
I would settle for readonly mode. Once in I didn't know how to preserve
the changes. Even though the phenomenon is rare, the message should still
say how to recover.
∂09-Mar-84 0248 JMC unique names
To: VAL
Let's discuss it when you have a chance. I have a clue which is related
to one of your ideas. It seems that we can do it with circumscription
provided we never introduce the objects at all. We only use names.
We have eqden(n1,n2), but we don't introduce a denotation function
or even a denotation predicate. We can have a function name-den(n),
which secretly is name(denotation(n)), and with its aid we can
have sentences like name-den('a) = 'a, and name-den('b) = 'a when
the denotations of 'a and 'b are the same. These will satisfy
the axiom
(n1)(n2)(eqden(n1,n2) iff name-den(n1) = name-den(n2)).
The abnormality sentence is then just
¬ab aspect1(n1,n2) ⊃ ¬eqden(n1,n2).
∂09-Mar-84 1308 JMC
To: CLT
03-26 Monday, 4pm, Kennedy reception for Mitterrand
∂09-Mar-84 1723 JMC mailing list
To: smh@SRI-KL.ARPA
I'm an occasional consultant and don't have a regular SRI mailing
address. I suppose the message was forwarded to me, although I'm
not sure why. I'm also connected with CSLI, although I don't think
I have an SRI account from that. My login at SRI is MCCARTHY, but
I prefer mail addressed to JMC@SU-AI.
∂09-Mar-84 2244 JMC whiteball
To: reid@SU-GLACIER.ARPA, feigenbaum@SUMEX-AIM.ARPA,
golub@SU-SCORE.ARPA
I agree with Ed for essentially the reasons he gave. I also believe
that the prerogative should be used rarely. I have used it once or
twice in more than twenty years.
∂09-Mar-84 2315 JMC travel
To: RPG
CC: DFH
The contract still isn't in proper shape. I suggest you put the case
by MAIL to Betty Scott, and ask her what she can do, following it up
by a phone call on Monday. However, if worst comes to worst, I'll
back up the trip with unrestricted money so that if the charge is
disallowed, that will pay. Apparently, the incompetence in NAVELEX
continues, so we are proceeding on some sort of month by month basis
with only expenses deemed essential allowed.
∂09-Mar-84 2317 JMC assistantship offer
To: pack@SU-SCORE.ARPA
This is to offer you an assistantship for Spring Quarter. I believe
the pay, etc. are standard. The work will involve mainly helping
with formalization of common sense knowledge but will also involve
helping me get out a new paper on circumscription including low
level items like proof reading and compiling references. This won't
amount to a lot, however.
∂10-Mar-84 0004 JMC beliefs of thermostats
To: laws@SRI-AI.ARPA
Thanks for your message. I have two comments.
1. First, you missed a transition that arose in the discussion. Ascribing
beliefs about what temperatures made what humans comfortable was in
connection with a hypothetical more elaborate temperature control
mechanism that was designed to take such facts into account. For
the very simplest thermostat, I proposed only the beliefs, "the room is
too cold", "the room is too hot", and "the room is ok".
Perhaps I made the transition without adequate notice.
2. Second, there is the question of whether the thermostat believing
"the room is too cold" requires that the thermostat know how sentences
are constructed from parts, i.e. that it somehow generate or understand
the sentence. I don't want to assume any such thing. "The room is
too cold" is how we express the belief. Ascribing the belief is
supported by the fact that the thermostat does what will achieve
the goal we also ascribe to it, i.e. making the temperature ok.
Also, please remove me as a subscriber to the AI list digest, since
I'll change to reading the public copy.
∂11-Mar-84 0053 JMC tentacles
To: Richard.Wallace@CMU-RI-ROVER.ARPA
None were built after the orm, and the orm was never connected
to the computer. The orm ought to be around somewhere, possibly
still in the old Lab building, possibly in the 5th floor storeroom
of the Computer Science Department. Les Earnest (LES@SU-AI) might
know and so might Victor Scheinman who built it (I think).
∂11-Mar-84 1449 JMC
To: CLT
I like Schubert, so perhaps I'll go anyway.
∂12-Mar-84 1213 JMC
To: SJG
"Oh, to be in England when April's here" or something like that. How
about this afternoon at three?
∂12-Mar-84 1431 JMC
To: Solomon@MIT-MULTICS.ARPA
I'm sorry to here about Ithiel. He accomplished a lot.
∂12-Mar-84 1434 JMC
To: JG@SU-AI.ARPA
Maybe it's "Oh to be in England now that April's there."
∂12-Mar-84 1629 JMC circumscription sometimes works
To: reiter%ubc@CSNET-RELAY.ARPA
Your proof about circumscribing equality looks correct to me, though
I want to ponder it further. However, circumscribing EX X.P(X) does
lead to E! X.P(X) as the following EKL proof shows. EKL is Jussi
Ketonen's interactive theorem prover. The only part that may require
explanation is the use of DEFINE in step 3. This rule of EKL allows
one, given a proof of existence of an object satisfying given conditions,
allows one to define a new symbol as representing such an object. It's
the way we do existential instantiation in EKL.
Incidentally, I have been challenged to give examples of the practical or
AI utility of circumscription in the case where it leads to a disjunction.
Do you have good examples from databases? My interlocutor grumbles about
the example of the employer in Vancouver and the wife in Toronto on the
grounds that he personally wouldn't want to draw any conclusion at all
about the person's city of residence and is backed up by his psychologist.
1. (AXIOM |EX X.P(X)|)
2. (AXIOM
|ALL P1.(EX X.P1(X))&(ALL X.P1(X) IMP P(X)) IMP (ALL X.P(X) IMP P1(X))|)
3. (DEFINE A |P(A)| (USE 1))
4. (UE ((P1.|LAMBDA X.X=A|)) 2 (OPEN A))
(ALL X.X=A IMP P(X)) IMP (ALL X.P(X) IMP X=A)
5. (DERIVE |ALL X.X=A IMP P(X)| (3) NIL)
ALL X.X=A IMP P(X)
6. (TRW |ALL X.P(X) IMP X=A| ((USE 4) (USE 5)))
ALL X.P(X) IMP X=A
7.
∂12-Mar-84 2354 JMC re: Western Culture in the Computer Age (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - We too could have our very own classicist as a visiting fellow
who could then impress gullible journalists.
∂13-Mar-84 1058 JMC chapters
To: aaai-office@SUMEX-AIM.ARPA
Claudia,
The draft on local chapters is ok. I suggest you send it out
over both our names. In general, I think that nothing bad is likely
to happen, and if something does, we can correct it, since chapter
charters should be revokable.
John
∂13-Mar-84 1153 JMC
To: aaai-office@SUMEX-AIM.ARPA
I'll make another try at a message. Thanks.
∂13-Mar-84 1252 JMC
To: grosof@SUMEX-AIM.ARPA
Unfortunately, I will not be able to attend Thursday. A definite date
somehow didn't get in my calendar, and I will be in Dallas talking at
Texas Instruments on Thursday. Some other time would be possible.
∂13-Mar-84 1446 JMC
To: PHY
I'll be in Dallas Thursday. Next Monday would be ok.
∂13-Mar-84 1448 JMC
To: PHY
Tuesday the 20th is also ok.
∂13-Mar-84 1653 JMC final
To: haym@SU-SCORE.ARPA
I'm finally convinced that your COMPACTIFY is correct, and now I'll
see what that does to the grade. Sorry I took so long and had to be
reminded so many times, but looking at that exam again was not something
I wanted to do.
∂14-Mar-84 1106 JMC lunch
To: TOB
Lunch Friday is possible. I suggest you make a reservation
for noon, since I have a 1:15 class.
∂14-Mar-84 1431 JMC @angio
To: MDD
Incorporating @angio in a message doesn't incorporate that file
in the message. I guessed you were trying to incorporate angio[1,mdd]
and read that file, so I'm informed, thank you. However, if you want
to send that message to someone else, the correct thing to do is either
mail <some friend> @angio
executed from the system or else use the E editor to copy from angio
into the file you want to send as a message. Sorry it didn't turn
out better and glad it didn't turn out worse.
∂15-Mar-84 2147 JMC
To: bosack@SU-SCORE.ARPA
OK, I think you should go ahead.
∂15-Mar-84 2308 JMC investigation
To: feigenbaum@SUMEX-AIM.ARPA
Sure talk to him. I brought in on myself by applying for a clearance.
In fact, I think I gave you as a reference.
∂15-Mar-84 2354 JMC LISP history
To: HST
The New York Herald Tribune indeed doesn't exist any more. A newspaper's
files of old stories is called its morgue. I would suggest a library, e.g.
the New York Public Library. I suppose there must be an index of newspaper
holdings of libraries that would say what libraries keep files of the
New York Herald Tribune.
∂16-Mar-84 0009 JMC
To: HST
I probably have many polaroid pictures from that era, but I don't
remember taking any of the LISP group. I did take photos of the
Algol meetings that Peter Naur eventually copied. I'll look around.
What sort of picture are you looking for?
∂16-Mar-84 1015 JMC
To: ME
ap was down since yesterday afternoon but revived by tty exist 16
∂16-Mar-84 1050 JMC
To: ME
I suggest that the NS greeting mention tty exist till the NS fans know about it.
∂17-Mar-84 0008 JMC
To: CLT
It is damn rude to hang up on me.
∂17-Mar-84 1027 JMC history of the computer
To: Solomon@MIT-MULTICS.ARPA
Do not call before 10am Pacific Time. My phone numbers are
(work: 415 497-4430) (home: 857-0672). However, at the moment
I have no useful ideas that you won't have thought of. It isn't
something I can get personally involved in. Well maybe I have
a few ideas.
∂17-Mar-84 1156 JMC
To: ARK
Why not?
∂17-Mar-84 1827 JMC example
To: reiter%ubc.csnet@CSNET-RELAY.ARPA
Thanks. I think that example will do. The result you already have
looks pretty strong.
∂17-Mar-84 1829 JMC disjunctions
To: SJG
∂17-Mar-84 1822 reiter%ubc.csnet@csnet-relay.arpa disjunctions
Received: from CSNET-RELAY.ARPA by SU-AI.ARPA with TCP; 17 Mar 84 18:22:12 PST
Received: by csnet-relay via ubcpob; 17 Mar 84 21:11 EST
Date: Sat, 17 Mar 84 18:03:31 pst
From: Ray Reiter <reiter%ubc.csnet@csnet-relay.arpa>
Message-Id: <8403180203.AA13743@ubc-vision.UUCP>
Received: by ubc-vision.UUCP (4.12/3.14)
id AA13743; Sat, 17 Mar 84 18:03:31 pst
To: JMC%su-ai.arpa@csnet-relay.arpa
Subject: disjunctions
Dear John
Thanks for your proof on circumscribing an existential. I now see what I
overlooked.
Re a convincing example where circumscription yields a disjunction. How about
the following:
In a taxonomy we usually have a number of necessary conditions on a class, e.g.
DOG(x) --> MAMMAL(x)
CAT(x) --> MAMMAL(x)
etc.
If we have reason to believe that our taxonomy exhausts the class of mammals,
we can circumscribe the predicate MAMMAL to obtain
MAMMAL(x) --> DOG(x) V CAT(x) V...
Is this the kind of example you were seeking?
Regards, Ray.
PS The search continues for a proof that circumscription cannot yield new
inequalities. The old "proof" failed for theories without minimal models. We
conjecture that universally quantified Horn theories always have minimal
models and therefore for such theories, circumscription yields nothing new
about inequalities.
"reiter%ubc.csnet"@csnet-relay.arpa
example
Thanks. I think that example will do. The result you already have
looks pretty strong.
∂17-Mar-84 2235 JMC
To: SJG
Please MAIL me your schedule of when you will be here and when away.
∂17-Mar-84 2318 JMC Poole
To: llw@S1-A.ARPA
Poole called me twice in the last week with what seems to me like
a last ditch effort to sell his shiny new F1 system. Were I in the
market for a TOPS-20, it might well be a good buy, but I don't think
he'll succeed.
Poole see Foonly, (beeper: 363-3038, after answer dial my number)
Foonly 969-7815
(home: 941-7563)
120 Monroe Drive, Palo Alto 94306
∂18-Mar-84 1005 JMC
To: ARK
It seems to me that your memo is capable of being misunderstood and
needs clarification. You should be explicit that the projects in question
involve improving CSD-CF for all users and are not the students' own
research and that the payoff for the student is computer time that he
can use for his own research.
∂18-Mar-84 2341 JMC
To: pack@SU-SCORE.ARPA
I'll see what I can do about an office.
∂19-Mar-84 1103 JMC
To: zauderer@SU-SCORE.ARPA
CC: lamping@SU-SCORE.ARPA
How about Thursday, Mar 29 at 2pm? April 3 is not possible.
∂19-Mar-84 1113 JMC
To: lamping@SU-SCORE.ARPA
Thursday afternoon would be better. How about 1:30?
∂19-Mar-84 1159 JMC
To: lamping@SU-SCORE.ARPA
How about Friday at 4pm?
∂19-Mar-84 1422 JMC cheng
To: yao@SU-SCORE.ARPA
I could see her at 3pm, but probably I'm not the right person for her.
There are other people in AI, e.g. Feigenbaum, who have had more
contact with the Stanford psychologists than I have. Also there
is considerable interaction through CSLI, and the right person to
inquire of there is probably John Perry.
Perry, John JRP@sri-ai 28 7-1275 28 327-0649 (HM)
∂19-Mar-84 1606 JMC
To: grosof@SU-SCORE.ARPA
"reiter%ubc"@csnet-Relay
∂19-Mar-84 1614 JMC
To: grosof@SUMEX-AIM.ARPA
Reiter,Ray (office: 604 228-4142) (home: 604 685-3825)
reiter%ubc.csnet@csnet-relay.arpa
"reiter%ubc"@csnet-Relay
The Dept. of Computer Science, UBC, Vancouver, BC, V6T 1W5.
"uw-beaver!ubc-vision!reiter"@LBL-CSAM
∂19-Mar-84 1616 JMC
To: grosof@SU-SCORE.ARPA
"reiter%ubc"@csnet-Relay
generalizing the unique names problem
The "unique names problem" is the following. Some constant symbols,
a, b, etc. are constrained by an axiom A. We want to find
a general way of constructing a formula UN(A), very likely
of second order logic, such that in the models of UN(A) maximize
inequality of a, b, etc. It now seems unlikely that this
can be accomplished by circumscribing some formula.
Maybe "unique names" isn't an informative name for the problem.
Perhaps we should call it the "distinctness problem" or the
"problem of expressing a presumption of distinctness".
It may be worthwhile to think about a generalization of the problem.
There are some related mathematical concepts. In geometry
one speaks about a figure being in general position provided the
co-ordinates of distinguished points satisfy no algebraic relations
not implied by its definition. Thus a right triangle in general
position is not isosceles but also no side has twice
the length of another. Indeed the sides satisfy
no algebraic equations other than x↑2 + y↑2 = z↑2. In algebraic
geometry one speaks of a generic point of an algebraic variety.
Again it is one that satisfies no relations other than those implied
by the fact that it is a point of the variety. If the field is the
complex numbers and the variety is irreducible, i.e. isn't the union
of two or more varieties, then generic points exist. If the field
is, for example, the rationals they need not exist, but can be shown
to exist in a suitable transcendental extension field. For example,
(2t/(1+t↑2), (1 - t↑2)/(1 + t↑2)) is a generic point of x↑2 + y↑2 = 1.
In the theory of forcing, one speaks of a generic set. This is a set
of natural numbers such neither it nor its complement contain any
infinite recursively enumerable sets, i.e. it has no special properties
other than finite ones. This example probably isn't on the right
track for us, because we are primarily interested in finite systems.
I am not an expert on any of the above-mentioned mathematical
concepts, but I think they suggest something to us. Here is an
direct generalization of unique names. We have an axiom A.
We are interested in finding a wff GCU(A) (for generalized circumscription,
unique) which distinguish those models in which as few terms as
possible are assigned equal value. There will be two cases. In the
first case, analogous to an irreducible variety or to the existence
of a unique minimal model with circumscription, GCU(A) will insure
that two terms are inequal unless this is implied by A. In the
more general case, e.g. when A is x=y or w= z, there will be
a disjunction (possibly infinite). The generalization consists
in allowing arbitrary ground terms and not just constants.
It is by no means clear that GCU(A) exists as a
higher order formula systematically derivable from A, let alone
being obtainable by circumscription. Maybe
it sometimes exists. Maybe the idea of a generic point of a
variety can be taken over directly.
∂20-Mar-84 1201 JMC
To: pjb@S1-A.ARPA
I plan to be at the Lowell, Rod and Mike party Saturday.
∂20-Mar-84 1712 JMC Laura Breeden
To: JMC
She is the CSnet contact. Her address is breeden@CSNET-CIC
∂20-Mar-84 1821 JMC re: Strategic Computing in the New York Review of Books (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - It's not the New York Review of Books that's on the wrong track,
it's the author Emma Rothschild personally. On the basis of her previous
publications, I doubt the sincerity of her concern for the competitiveness
of the American private sector.
∂20-Mar-84 2235 JMC re: Strategic Computing (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - It isn't as paradoxical as all that. NSF spends a billion and
NIH several. As for DoD, it's obscure (in some sense deliberately)
what fraction of its larger research budget goes for basic research.
However, NIH is specialized to medicine and biology, and NSF spreads
its research out rather evenly. Therefore, DoD remains the best
source for large projects not fitting under NIH, NASA or DoEnergy
or so well supported by the scientific establishment that NSF can
be got to support it. AI and computer science benefitted especially
from ARPA in the early days, because J.C.R. Licklider persuaded the
DoD that it computer science was basic to command and control rather
than being a minor branch of mathematics or a tool for physicists,
which was the way the scientific establishment looked at it.
∂21-Mar-84 1949 JMC Veronica Dahl
To: chappell@SRI-AI.ARPA, pereira@SRI-AI.ARPA
Elsie, that's Veronica not Victoria. I can pay 1/3 of the expenses.
I suggest that one of the organizations pay (not mine) and tell my
secretary Diana Hall (DFH@SU-AI) what our share is. Do you have titles
and abstracts yet of talks she can give.
∂22-Mar-84 0059 JMC
To: jrp@SRI-AI.ARPA
CC: stucky@SRI-AI.ARPA
I'll phone you about Kuo, but I fear he is taking quite some time getting
used to this environment, so I haven't got a lot out of him either. I
suggest a long talk with him.
∂22-Mar-84 1654 JMC equality
To: grosof@SUMEX-AIM.ARPA
∂12-Mar-84 0026 reiter%ubc.csnet@csnet-relay.arpa equality
Received: from CSNET-RELAY.ARPA by SU-AI.ARPA with TCP; 12 Mar 84 00:26:46 PST
Received: by csnet-relay via ubcpob; 12 Mar 84 3:09 EST
Date: Sun, 11 Mar 84 19:46:27 pst
From: Ray Reiter <reiter%ubc.csnet@csnet-relay.arpa>
Message-Id: <8403120346.AA15038@ubc-vision.UUCP>
Received: by ubc-vision.UUCP (4.12/3.14)
id AA15038; Sun, 11 Mar 84 19:46:27 pst
To: JMC%su-ai.arpa@csnet-relay.arpa
Subject: equality
Dear John,
Etherington's proof seems to have a fatal flaw.
Here is my simple proof that circumscribing the equality predicate in a theory
T(=) yields nothing new, i.e. that
T(=) proves Circumscription schema for =
Assume T(=) contains the usual axioms for equality, in particular (x)x=x, and
axioms sanctioning the substitution of equal terms for each other.
We want to show
T(=) proves [T(phi) & [(xy)phi(x,y) --> x=y]] --> [(xy)x=y --> phi(x,y)]
i.e. show
T(=), T(phi), [(xy)phi(x,y) --> x=y], a=b proves phi(a,b)
Now T(=) contains (x)x=x, so T(phi) contains (x)phi(x,x)
So with (x)phi(x,x), a=b and the principle of substitution we can prove
phi(a,b). QED
I also think that circumscription may have other limitations. For example,
intuitively, circumscribing P in (Ex)P(x) should yield (E!x)P(x). I have
tried very hard to get this but can't. Do you know how?
Best wishes, Ray.
∂22-Mar-84 1958 JMC re: fifth generation (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - Yes, as FIFTH[E83,JMC]@SAIL. It is also in the lounge and readable.
∂22-Mar-84 2333 JMC
To: CLT
You will need your driver's license for ID Monday.
∂23-Mar-84 1538 JMC
To: grosof@SUMEX-AIM.ARPA
Lifschitz, Vladimir 915 581-4261
Computer Science Dept.
University of Texas at El Paso
El Paso, TX 79968
408 733-2194
(home: 857-1628)
∂23-Mar-84 1836 JMC
To: dkanerva@SRI-AI.ARPA
Please send me CSLI Technical reports 2 and 4 by Barwise and Konolige.
∂23-Mar-84 1839 JMC
To: reges@SU-SCORE.ARPA
jmc - I don't think Stuart's message provides any excuse for PEDIT asking
me to repeat my address and telephone numbers. The Department has that
information. I will not do it again unless the information actually
changes.
∂23-Mar-84 1841 JMC re: Relative merits of US & USSR foreign policy (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - I wouldn't agree that U.S. actions, or even what the leftists accuse
the U.S. of doing, are comparable with the Russians did in Czechoslovakia
Afghanistan, Hungary, the Baltic countries or Iran (1943-46). The U.S.
has supported governments we like. This is comparable to Russian actions
in Cuba or Vietnam, where the Russians supported governments they liked
with arms and economic aid. American actions supporting the governments
of El Salvador and Greece were comparable. The U.S. via the CIA has
encouraged occasional revolutions and coups, e.g. in Iran and Guatemala.
This is comparable to Russian
support of the Sandinistas and perhaps of the original military coup
in Afghanistan. There is no evidence that the U.S. was responsible in
any substantial way for the coup in Chile, though there had been discussion
of helping prevent Allende from taking power, but a decision to do
nothing. There is no evidence of any special support of the Philippine
Government today other than refraining from trying to overthrow it and
allowing normal commercial relations with it. We also have our bases
by agreement with it. None of the governments you mention are comparable
to Hitler. I haven't read Mother Jones on Australia, but I'll bet it
is some kind of disinformation.
Hungary: They overthrew the government they recognized. When the
Prime Minster (Nagy) and the head of the armed forces took refuge
in the Yugoslav Embassy, they persuaded them to emerge by promsing
they wouldn't be harmed and then executed them. Andropov himself
was the Soviet Ambassador who organized this.
Czechoslovakia: The Russians invaded and overthrew a government
they recognized. They took the leaders of the government to Moscow
and held them prisoner till they agreed to changes. The Hungarians
had angered the Russians by announcing withdrawal from the Warsaw
Pact. The Czechs mistakenly thought they could avoid invasion
by staying in.
Afghanistan: They landed at the airport, arrested and killed the
prime minister of a government they recognized and installed as
prime minister a man they brought from Czechoslovakia. They are waging
a war of scorched earth with poison gas against the inhabitants.
Neither our interventions in the Dominican Republic nor Grenada are
remotely comparable. I don't remember precisely what triggered the
Dominican invasion, but it wasn't a communist or nationalist threat,
something more like ordinary gangsterism, not anti-U.S. The Grenadian
invasion was at the request of the Governor General and the other
Carribean states and turned out to be supported by 91 percent of
the Grenadian population. We'll be out as soon as the Grenadians
let us.
It takes a real distortion of reality to regard these as comparable.
∂24-Mar-84 1107 JMC re: Attention people with SCORE accounts (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - In the particular case, the amount of data is small enough, so that
the conversion should have been done by hand. This would be far less
work than writing and debugging a program to do it. My guess is that
it amounts to about one or two clerk days, and the department has an
excess of clerical and administrative personnel.
∂25-Mar-84 0132 JMC re: Summer Job For American Indian Student (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - So much for equal opportunity.
∂25-Mar-84 0134 JMC
To: ME
May I have bboard for 10 sec?
∂25-Mar-84 2102 JMC re: CPSR (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - I note there is nothing about the social responsibilities of
computer professionals for harm of their own making. For example,
computer professionals built systems in which the information was
stored on magnetic tapes in a way that was only accessible by
tape-to-tape update runs. As a result customers and clients of
companies and government agencies using such system often find
it difficult or impossible to get wrong information corrected.
A second example is that the writers of interactive computer
programs often regard their users as idiots who must be carefully
restricted in what they can do so they won't make a mistake. This
fascist ideology often results in programs that are rejected
in academic environments. However, in ordinary office environments
the workers often have no choice but to submit to the whims of
the arrogant programmer bureaucrats.
It seems to me that CPSR has demonstrated no specifically computer
related interests. Putting "computer" into the name has so far
served merely as a gimmick for getting access to propaganda
opportunities from which their activities would otherwise be
excluded as irrelevant to the function.
I don't wish to claim that this is important, except in so far
as it preempts consideration of the real social responsibilities
of computer professionals - to develop standards for the computer
systems to be polite to people.
A final example is given by Stuart Reges demanding that everyone
provide again information that the Department already has rather
than being willing to re-edit the information he already has.
It is possible to make computer files in such a way that the
information is guaranteed to be reformattable automatically,
but this again is nothing that CPSR would imagine relevant
to the social responsibilities of computer professionals.
∂25-Mar-84 2106 JMC
To: larson@SU-SHASTA.ARPA
jmc - I note there is nothing about the social responsibilities of
computer professionals for harm of their own making. For example,
computer professionals built systems in which the information was
stored on magnetic tapes in a way that was only accessible by
tape-to-tape update runs. As a result customers and clients of
companies and government agencies using such system often find
it difficult or impossible to get wrong information corrected.
A second example is that the writers of interactive computer
programs often regard their users as idiots who must be carefully
restricted in what they can do so they won't make a mistake. This
fascist ideology often results in programs that are rejected
in academic environments. However, in ordinary office environments
the workers often have no choice but to submit to the whims of
the arrogant programmer bureaucrats.
It seems to me that CPSR has demonstrated no specifically computer
related interests. Putting "computer" into the name has so far
served merely as a gimmick for getting access to propaganda
opportunities from which their activities would otherwise be
excluded as irrelevant to the function.
I don't wish to claim that this is important, except in so far
as it preempts consideration of the real social responsibilities
of computer professionals - to develop standards for the computer
systems to be polite to people.
A final example is given by Stuart Reges demanding that everyone
provide again information that the Department already has rather
than being willing to re-edit the information he already has.
It is possible to make computer files in such a way that the
information is guaranteed to be reformattable automatically,
but this again is nothing that CPSR would imagine relevant
to the social responsibilities of computer professionals.
[JMC - Forwarded to OTHER-SU-BBOARDS from line 146]
∂25-Mar-84 2256 JMC
To: feigenbaum@SUMEX-AIM.ARPA
I saw it. Thanks.
∂26-Mar-84 0228 JMC Forcing and non-monotonic logic
To: JEF@SU-AI.ARPA
Do you have an extended abstract by Marek on this subject? It refers
to your book, so I thought he might have sent it to you also. Anyway
there is one notation I don't understand that might possibly be
conventional in database theory. He gives as an example of an
integrity constraint [X →→ Y] where the brackets are curly. Is
this a standard notation?
∂26-Mar-84 0235 JMC More on Marek paper.
To: JEF@SU-AI.ARPA
I suspect I'd better give a bit more detail. The database is listed
as H = [p(x,y,z), p(x,y1,z), p(x,y,z1)], and the constraint is
listed as Sigma = [X →→ Z]. The previous [X →→ Y] was a mistake
of mine.
∂26-Mar-84 0342 JMC
To: JEF@SU-AI.ARPA
Disregard previous messages. Arthur Keller was on duty.
∂26-Mar-84 1135 JMC IBM
To: HST
Mr. E. T. Hahn of IBM Germany will call you trying to understand better
what you told me about the use of IBM computers.
∂26-Mar-84 1348 JMC
To: PHY
May I have 2:30 tomorrow with Tsichritzis?
∂26-Mar-84 1755 JMC
To: DFH
Could you pick up 3 copies of docdil.xgp from xgp and mail them to Susie?
∂26-Mar-84 1932 JMC
To: SGF
I'll pick it up Wednesday at 11 or a bit earlier.
∂26-Mar-84 1941 JMC determinism and choice
To: phil-sci%oz@MIT-ML.ARPA
I am most in sympathy with the point of view expressed by DAM. It is
expressed more elaborately in McCarthy and Hayes, "Some Philosophical
Problems from the Standpoint of Artificial Intelligence", Machine
Intelligence 4. The essence is that what a person or machine
CAN DO is determined by its position in the interaction structure
rather than by its internal structure. The latter determines what
it will do. This general idea is variously decorated in the paper.
∂26-Mar-84 2259 JMC address
To: aaai-office@SUMEX-AIM.ARPA
Could you look up the address of John McDermott, the inventor of R1,
and MAIL it to HST@SU-AI?
∂26-Mar-84 2307 JMC Tsichritzis
To: PHY
I forgot another appointment at 12 tomorrow. However, I'm going to
his seminar and I saw him both at lunch and at Knuth's, which should
suffice, since I got to ask all the questions I could think of.
∂27-Mar-84 1342 JMC
To: boyer@UTEXAS-20.ARPA
Yes, and I would like both.
∂27-Mar-84 1344 JMC
To: pack@SU-SCORE.ARPA
I'll see what I can do about a SAIL terminal for your desk. Also we
can lend you a SAIL terminal for home if you want.
∂27-Mar-84 1702 JMC
To: DFH
∂27-Mar-84 1657 DFH Dahl expenses
Should your share be from your unrestricted or NSF?
Make it NSF.
∂27-Mar-84 2241 JMC
To: ailist-request@SRI-AI.ARPA
As you wish. If you wish to copy it, the file is fifth[e83,jmc].
∂28-Mar-84 0002 JMC
To: DFH
Please inform Paolucci that the 10th will be ok.
∂28-Mar-84 0123 JMC determinism and choice
To: phil-sci%oz@MIT-MC.ARPA
I am most in sympathy with the point of view expressed by DAM. It is
expressed more elaborately in McCarthy and Hayes, "Some Philosophical
Problems from the Standpoint of Artificial Intelligence", Machine
Intelligence 4. The essence is that what a person or machine
CAN DO is determined by its position in the interaction structure
rather than by its internal structure. The latter determines what
it will do. This general idea is variously decorated in the paper.
∂28-Mar-84 0850 JMC
To: SGF
Ignore previous message.
∂28-Mar-84 0849 JMC
To: SGF
∂25-Mar-84 0132 JMC re: Summer Job For American Indian Student (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - So much for equal opportunity.
∂28-Mar-84 1039 JMC
To: JEF@SU-AI.ARPA
I had a question about database notation, specifically what does X →→ Y
mean, but ARK answered it for me. The Marek paper purports to provide a
proper semantics for the McDermott-Doyle non-monotonic logic using
model-theoretic forcing, but this seems to me a bit like using a
sledge hammer to kill a fly, since forcing really gets its power in
infinite domains. Marek takes his examples from database theory. If
you're interested, I can supply a copy of the Marek note. You need
a computer like SAIL that takes regular disk dumps. I haven't
permanently lost a file since 1969.
∂28-Mar-84 1112 JMC Feferman terminal
To: ME
I took it from him and have it at home (in addition to my own). Carolyn
will use it for a while.
∂28-Mar-84 1147 JMC DAM message
To: perlis%umcp-cs.csnet@CSNET-RELAY.ARPA
∂26-Mar-84 1509 @MIT-MC:DAM@MIT-OZ Determinism and Choice
Received: from MIT-MC.ARPA by SU-AI.ARPA with TCP; 26 Mar 84 15:08:49 PST
Date: Mon, 26 Mar 1984 17:49 EST
Message-ID: <DAM.12002512402.BABYL@MIT-OZ>
From: DAM%MIT-OZ@MIT-MC.ARPA
To: phil-sci%MIT-OZ@MIT-MC.ARPA
cc: MINSKY%MIT-OZ@MIT-MC.ARPA
Subject: Determinism and Choice
I would like to present a theory of "free will" which I have
not seen before. Rather than discuss "free will" I would like to
discuss the related notion of "choice", and in particular statements
of the form "I COULD do X", or "I COULD HAVE done X instead of Y".
Consider a deterministic reasoning machine which is engaged in a game
of chess. The machine is considering various legal moves (it knows
the rules of chess). Furthermore suppose that it expresses the fact
that move M is legal via the statement:
S1: "I could make move M."
This statement is clearly TRUE in the sense that M is a legal move.
Furthermore it would be a mistake (it would be mal-adaptive) for the
machine to say that S1 is probabably false because "I am determined
and I in fact probably won't make move M". The point is that move M
is IN FACT an option which must be considered. Thus,
for the machine to function effectively it MUST be aware of its
choices. The machine does in fact "have choices" (which it should
consider and be aware of!) even though it is determined.
The situation (I claim) is much the same for people. We have
choices (physically allowed options) which we must consider in order
to do effective planning. The fact that we "have choice" in no way
contradicts the possibility that we are deterministic machines.
Determistic machines must also be aware of their "choices" in order
for them to make effective decisions.
Thus one interprets "free will" as "having choice" then we can
be BOTH determined AND have free will. If one interprets "free will"
as "not being determined" then the notion of free will becomes
irrelevent to our day-to-day existence. I think that people believe
in "free will" simply becasue they know (correctly) that they must
decide amoung certain choices.
David Mc
∂28-Mar-84 1155 JMC determinism and choice
To: phil-sci%oz@MIT-MC.ARPA, perlis%umcp-cs.csnet@CSNET-RELAY.ARPA
McC and Hayes suggests treating the phenomenon mentioned by Perlis as follows.
Regard the brain, (or perhaps better the mind, since the form of physical
implementation is not relevant) as composed of interacting parts. What one
part can do depends on its position in the causal structure, and this takes
into account the other parts. For example, the folk psychology decomposition
of the mind into Will and Intellect works pretty well for
Perlis's example of jumping provided we add the Body as a part. Namely,
the Body can jump 4 feet, because it has outputs that would do it. However,
if the Will tells the Intellect to do it, the Intellect won't emit the
required order to the Body, so the Will can't. This may not be quite right,
but we will have to provide our robot programs with some kind of
notion of CAN based on interaction structure so that it will explore its
options and decide what to do in some reasonable way.
∂28-Mar-84 1432 JMC
To: golub@SU-SCORE.ARPA
Welcome back. Please put promoting Dick Gabriel to Senior Research
Associate on the agenda. However, it would seem that a PI should be
able just to ask for objections for such a promotion, since it is
possible to appoint to that rank from the outside by such a procedure.
I much prefer resolving the anomaly that way rather than by making
appointments go through a faculty meeting.
∂28-Mar-84 1803 JMC datadisk
To: ME@SU-AI.ARPA
CC: bosack@SU-SCORE.ARPA, pack@SU-SCORE.ARPA
Leslie Pack, just moving from the masters to the PhD program is joining
my group. Her desk is in 450. Is it feasible to put a Datadisc on it,
and do you have one?
∂28-Mar-84 1804 JMC Your paper
To: SJG
I have read it.
1. I don't see that the method is efficient, except in the simple
cases in which all presently proposed methods are efficient.
2. You don't provide a proposal for how to get the certainty of
an event in case a rule is de-activated with a certainty other
than 1. In that case the rule can be considered as de-activated
by a circumscription or default, etc. Neither providing such a
rule nor pointing out that you haven't is not good.
I have a few other comments.
∂29-Mar-84 1024 JMC reference
To: perlis%umcp-cs.csnet@CSNET-RELAY.ARPA
CC: dam%oz@MIT-MC.ARPA
Can you give me a reference to philosophers who consider determinism
essential for free will? I shall elaborate this in a phil-sci message,
but it seems to me that you and I and DAM are in essential agreement
on the philosophical problem involved in free will. The others are
raising irrelevancies. It is interesting to ask whether this is a
question of information like a scientific error. Would a better
explanation help or are philosophical disagreements psychologically
more deep rooted?
While the main philosophical problem may be resolved, there remain
many AI questions about what facts a robot should have to enable
it to reason about what it and other actors can do. Perhaps we
can devote a few messages to some of these.
∂29-Mar-84 1414 JMC 1980 lisp conference proceedings
To: boyer@UTEXAS-20.ARPA
They're out of print again. John Allen and Ruth Davis, who own the LISP Company
don't want to reprint it again. They would welcome someone else taking over.
How about getting whoever is printing the 1984 Proceedings to take over keeping
the 1980 Proceedings in print. Their address is
The LISP Company, John Allen, prop.
The LISP Conference
P.O. Box 487
Redwood Estates, CA 95044
Proceedings $22.50 U.S., Canada, Mexico; $25 elsewhere
out of print 1984
(Ruth Davis: 408 984-4358), (lisp company: 354-3668), (home: 353-3857)
∂29-Mar-84 1606 JMC
To: JK
1. (DEFINE A
|A(AB,FLIES)≡
(∀X.¬AB(ASPECT1(X))⊃¬FLIES(X))∧(∀X.BIRD(X)⊃AB(ASPECT1(X)))∧
(∀X.BIRD(X)∧¬AB(ASPECT2(X))⊃FLIES(X))∧(∀X.OSTRICH(X)⊃AB(ASPECT2(X)))∧
(∀X.OSTRICH(X)∧¬AB(ASPECT3(X))⊃¬FLIES(X))| NIL)
;labels: UNEQUAL SIMPINFO
2. (AXIOM
|(∀X Y.ASPECT1(X)≠ASPECT2(Y))∧(∀X Y.ASPECT1(X)≠ASPECT3(Y))∧
(∀X Y.ASPECT2(X)≠ASPECT3(Y))|)
3. (DEFINE A1
|A1(AB,FLIES)≡
A(AB,FLIES)∧
(∀AB1 FLIES1.A(AB1,FLIES1)∧(∀Z.AB1(Z)⊃AB(Z))⊃(∀Z.AB(Z)≡AB1(Z)))|
NIL)
4. (ASSUME |A1(AB,FLIES)|)
deps: (4)
5. (DEFINE FLIES2 |∀X.FLIES2(X)≡BIRD(X)∧¬OSTRICH(X)| NIL)
6. (DEFINE AB2
|∀Z.AB2(Z)≡(∃X.BIRD(X)∧Z=ASPECT1(X))∨(∃X.OSTRICH(X)∧Z=ASPECT2(X))|
NIL)
7. (RW 4 (USE 3 MODE: EXACT))
A(AB,FLIES)∧(∀AB1 FLIES1.A(AB1,FLIES1)∧(∀Z.AB1(Z)⊃AB(Z))⊃(∀Z.AB(Z)≡AB1(Z)))
deps: (4)
(show 7)
(trw bird#7#2 (use 7 mode: exact))
8.
; failed to derive
∀AB1 FLIES1.A(AB1,FLIES1)∧(∀Z.AB1(Z)⊃AB(Z))⊃(∀Z.AB(Z)≡AB1(Z))
8.
(derive bird#7#2 (7))
∂29-Mar-84 1617 JMC
To: JK
In case you want it, the file is bird.lsp[w84,jmc].
∂29-Mar-84 1713 JMC
To: CLT
Lest I forget to tell you, Monday I'm attending a spouseless Hoover dinner.
∂29-Mar-84 2314 JMC serious bug
To: JK
Page 2 of bird.lsp[w84,jmc] runs. The last step is wrongly accepted.
∂30-Mar-84 1004 JMC
To: DFH
Please borrow this for me from the philosophy library.
The anthology, "Free Will and Determinism" edited by Bernard Berofsky
(Harper and Row, 1966) contains several such papers.
∂30-Mar-84 1844 JMC
To: JK
I have more dumb questions. Phone home or office if not too late.
∂30-Mar-84 2136 JMC re: ultimate dictionary (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - The ultimate in one important sense is the Oxford English Dictionary,
commonly abbreviated OED. It took more than 80 years to compile and
came out in 26 volumes. The compact edition in two volumes requires
a magnifying glass (supplied) to read. It costs about $75 but is usually
available as a Book-of-the-Month premium for joining for about $20.
The OED has a supplement containing more recent words. As an
etymological dictionary, it is unrivalled.
∂30-Mar-84 2159 JMC
To: JK, JJW
The problem was mine - a parenthesis error.
∂30-Mar-84 2329 JMC stuck again
To: JK
bird.lsp is the file. p.2 contains the commands, p.3 contains
the result of running the commands up to step 20 which
gets PDL overflow. Why this step should give problems isn't
at all clear. The steps beyond ;foo are debris left over
from a previous attempt that got beyond step 20 but had a
wrong 20. Namely, the ¬ on ab2 was omitted.
∂31-Mar-84 0826 JMC
To: reiter@RUTGERS.ARPA
I hope I can have till Monday to mail my paper for nonmncon.
∂31-Mar-84 1252 JMC der
To: JK
I did in fact, go around that problem. Stay tuned and perhaps I'll have
another problem.
∂31-Mar-84 1522 JMC re: ultimate dictionary (from SAIL's BBOARD)
To: OTHER-SU-BBOARDS@SU-AI.ARPA
jmc - Printer's Inc. had the OED last time I looked.
∂31-Mar-84 1618 JMC
To: JK
2. (AXIOM
|(∀X Y.¬(ASPECT1(X)=ASPECT2(Y)))∧(∀X Y.¬(ASPECT1(X)=ASPECT3(Y)))∧
(∀X Y.¬(ASPECT2(X)=ASPECT3(Y)))|) (label xx)
(trw |ASPECT1(X)=ASPECT2(X2)|)
∂31-Mar-84 1845 JMC another bug
To: JK
Here seems to be yet another bug - in it or me. I want to get rid
of the conditions like ASPECT1 X = ASPECT1 X1 using facts like
∀x y. aspect1 x = aspect1 y ⊃ x=y. I put these facts in axiom 2,
and step 27 should do it, but even rewriting 27 with 29 which is
the assumption of these facts doesn't do it. Incidentally it takes
a quite a few seconds not doing it.
;labels: UNEQUAL SIMPINFO
2. (AXIOM
|(∀X Y.¬ASPECT1(X)=ASPECT2(Y))∧(∀X Y.¬ASPECT1(X)=ASPECT3(Y))∧
(∀X Y.¬ASPECT2(X)=ASPECT3(Y))∧(∀X Y.ASPECT1(X)=ASPECT1(Y)⊃X=Y)∧
(∀X Y.ASPECT2(X)=ASPECT2(Y)⊃X=Y)∧(∀X Y.ASPECT3(X)=ASPECT3(Y)⊃X=Y)|)
27. (RW 26 ((USE 25 MODE: EXACT) (OPEN AB2)))
(∀X.¬(∃X1.BIRD(X1)∧ASPECT1(X)=ASPECT1(X1))⊃¬FLIES(X))∧
(∀X.BIRD(X)∧¬(∃X2.OSTRICH(X2)∧ASPECT2(X)=ASPECT2(X2))⊃FLIES(X))
deps: (4)
28. (RW 27 (USE SIMPINFO))
(∀X.¬(∃X1.BIRD(X1)∧ASPECT1(X)=ASPECT1(X1))⊃¬FLIES(X))∧
(∀X.BIRD(X)∧¬(∃X2.OSTRICH(X2)∧ASPECT2(X)=ASPECT2(X2))⊃FLIES(X))
deps: (4)
29. (ASSUME
|(∀X Y.ASPECT1(X)=ASPECT1(Y)⊃X=Y)∧(∀X Y.ASPECT2(X)=ASPECT2(Y)⊃X=Y)∧
(∀X Y.ASPECT3(X)=ASPECT3(Y)⊃X=Y)|)
deps: (29)
30. (RW 27 (USE 29 MODE: EXACT))
(∀X.¬(∃X1.BIRD(X1)∧ASPECT1(X)=ASPECT1(X1))⊃¬FLIES(X))∧
(∀X.BIRD(X)∧¬(∃X2.OSTRICH(X2)∧ASPECT2(X)=ASPECT2(X2))⊃FLIES(X))
deps: (4 29)
31.
∂31-Mar-84 2050 JMC unbug
To: JK
I discovered that if I replaced ⊃ by ≡ in the second parts of my axiom 2,
the rewrites work. I thought they were advertised to work for ⊃. I now
have almost the result I want. I only lack showing that ostriches don't
fly even if they are birds. Most likely this is my fault rather than
that of EKL. Thanks for your help.
∂31-Mar-84 2158 JMC rebug
To: JK
As far as I can see step 27 loses information. See page 4
of bird.lsp[w84,jmc]. Namely, it loses the informtion that
ostriches can't fly. We have ∀x.ostrich x ∧ ¬ab aspect3 x ⊃ ¬flies x,
and ab2, which is substituted for ab is always false on
any aspect3 x.
∂01-Apr-84 0840 JMC bird done
To: JK
The last fix let me finish it. We have bird.lsp[w84,jmc] and bird.prf[w84,jmc].
The Boise has printed two copies and whichever of us gets in first can
pick them up. After a while, I may want to see if the number of steps
can be reduced without loss of clarity, but I won't try too hard till your
new DERIVE is available. Even as it is, I'm fairly impressed with DERIVE
in that it saves thought as well as saving typing. It would take me a little
while to figure out how to do some of the things that DERIVE did using small
steps.
∂01-Apr-84 1033 JMC Bossu-Siegel paper
To: Colmerauer@MIT-MULTICS.ARPA
What is the publication status of their paper. We are preparing a
bibliography on non-monotonic reasoning, and I'd like to know the
best way of referring to it. Also if you know of other European
work, I'd like to have references.