COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	∂03-Jul-82  0015	JMC  
C00004 00003	∂03-Jul-82  1529	JMC  
C00005 ENDMK
C⊗;
∂03-Jul-82  0015	JMC  
To:   CLT    
When on friday do you want to go to Carmel?

∂03-Jul-82  1529	JMC  
To:   equip at SU-SHASTA    
It was my understanding that the marginal price of the IBM 3380 disk
at 2.5 gigabytes was $40,000.  When we get some ARPA money, we must
survey the market and ask for proposals.  I don't understand the part
about moving fast.  I recognize that programming considerations may
dictate using less cost effective disks, but the difference here seems
rather large.

∂06-Jul-82  1455	JMC  
To:   equip at SU-SHASTA    
My memory of IBM price must have been incorrect.

∂07-Jul-82  1823	JMC  
To:   JMC    
balance=27.759,47

∂11-Jul-82  0255	JMC  
To:   marg at MIT-AI   
How is the Center of the World?

∂12-Jul-82  1626	JMC  
To:   bmoore at SRI-AI 
Axiomatizing of the blocks world using situations and circumscription

	One long term goal of AI is a general purpose database for
common sense facts about the effects of actions and other events.
Indeed a program with common sense must be able to apply general
knowledge to any problem that comes up and cannot expect some human
to give a presentation of the facts about (say) moving objects
specialized to the particular problem.

	The paper will present some recent progress towards this goal in
the area of the effects of actions that move objects, i.e. the blocks
world.  Using circumscription and considerable reification reduces
the frame and qualification problems greatly.  While more general
purpose than previous formalizations, there are still many limitations.

	Both the new methods and their limitations will be discussed.

∂12-Jul-82  1815	JMC  
To:   JMC    
EJS and REG Friday 1:30

∂12-Jul-82  1856	JMC  
To:   Colmerauer at MIT-MULTICS
CC:   csd.golub at SU-SCORE   
Dear Alain:

	The current proposal is approximately as follows:

	You would come to Stanford for the Winter Quarter 1983 and
teach one course of 3 hours per week in logic programming or other
topic of your choice.  Winter quarter classes start Jan 4 with
student registration Jan 2.  Classes end March 13 and examinations
are the following week.  Classes that meet on Monday, Wednesday and
Friday are usually 50 minutes and those meeting on Tuesday and
Thursday about 80 minutes.

	In order to make an official offer, Professor Gene Golub,
the Department Chairman, needs a curriculum vitae.  The intent is
that you would be paid half by the Department and half from one
of my research grants.

	You can send the c.v. by net mail to CSD.GOLUB@SCORE or
by ordinary mail to him at

Computer Science Department
Stanford, California 94305

	Best Regards,

∂13-Jul-82  1237	JMC  
To:   csd.bscott at SU-SCORE
Please call Dr. Chien at NSF 202 357-7345 concerning a report about my
unexpended funds on my grants that is required for processing the
grant for Ketonen's support.   Also  Chien tried to phone Ketonen and
was told by whoever answered the department phone that he didn't have
an extension.  It has been 4202 for at least 2 years.
I'm not entirely sure that Chien isn't confused about something.

∂13-Jul-82  2347	JMC  
To:   csd.genesereth at SU-SCORE 
Seminar by John McCarthy at 10:30, 1982 July 15 at SRI
Call Bob Moore at SRI if you need help getting in.

Axiomatizing of the blocks world using situations and circumscription

	One long term goal of AI is a general purpose database for
common sense facts about the effects of actions and other events.
Indeed a program with common sense must be able to apply general
knowledge to any problem that comes up and cannot expect some human
to give a presentation of the facts about (say) moving objects
specialized to the particular problem.

	The paper will present some recent progress towards this goal in
the area of the effects of actions that move objects, i.e. the blocks
world.  Using circumscription and considerable reification reduces
the frame and qualification problems greatly.  While more general
purpose than previous formalizations, there are still many limitations.

	Both the new methods and their limitations will be discussed.

∂13-Jul-82  2348	JMC  
To:   csd.golub at SU-SCORE 
I was supposing that you would decide that after you see the c.v.

∂14-Jul-82  0001	JMC  	Imlac    
To:   EJS    
I hope it is still convenient for you to take a look at my Imlac
this morning (Wednesday).  Please phone me at 857-0672.  The Imlac
prints are in the print tub in the room with the IIIs.  They are
the fourth from the computer room end.

∂15-Jul-82  1730	JMC  
To:   buchanan at SUMEX-AIM 
Yes, I plan to be in town Aug. 23-27.

∂16-Jul-82  0009	JMC  
To:   wilkins at SRI-AI
The McCarthy prize problem

	John McCarthy offers $100 of his own money for the best satisfactory
solution of the following problem.  Solutions must be in writing
and delivered to McCarthy or Nils Nilsson by Friday, July 23, 1982,
and the solvers must be prepared to defend their solutions orally
at SRI on that date.  Unless there is a satisfactory solution, no
prize will be awarded.

	The problem is to give axioms from which the following
piece of chess reasoning can be carried out.  The point isn't
chess, and solutions can use suitable chess lemmas as axioms.
On the other hand, it must be clear that the solution follows
entirely formally from the axioms.  Solutions will be compared,
if there is more than one meeting the formal requirements, according
to how well they express the informal reasoning presented below.


A chess example of reasoning about concurrent actions.

	The chess position of Figure 1 is mentioned in (Berliner 197x).
All he does with it is to remark that unless the lookahead sees very far
ahead, present day programs will keep the white king behind the white
pawns in order to preserve control of the center and hence won't win.
However, the following argument that white can win shows that
chess reasoning involves reasoning about concurrent events even though
the actual moves in chess are sequential.

	We note that black must stay within the eight marked squares or else
white will promote his pawn, and this is assumed to win for white.
White can win by moving his king along the path shown in the
figure and black cannot prevent his reaching b5 without letting
the pawn promote.  When it is white's move with his king on b5, if
the black king is not on c7, then the white king can immediately
advance to c6.  If the black king is on c7, then white moves his king
to a6 and can subsequently move to b6 and then to c6 without further
interference.  When it is white's move with his king on c6, if the
black king is not on e7, then white can capture the black pawn on
d6 - which is assumed to win.  If the black king is on e7, then white
move his king to c7 and can capture the pawn on the next move.

	From the AI point of view, the most interesting part of this argument
is that we do not need to examine in detail what black can do while
white is advancing his king.  All we need to reason about is whether
black stays within the eight squares that prevent the pawn from
queening.  We achieve an enourmous reduction in search space, because
we don't have to do a tree search on the black moves.  However, this
reduction is accomplished by reasoning about concurrent action, namely
"while white is advancing his king, black must stay within the eight
squares and therefore cannot interfere with the advance".  The last
two steps of reasoning, each of which makes a case analysis according to whether
the black king is or is not on a specific square, are also worth noting.

Figure 1 has the white king on d3, the black king on d8, white pawns
on c4, d5, e6, f5 and g4, and black pawns on c5, d6, f6 and g5.

	It seems to me that this problem provides the opportunity for
an advance beyond the simple situation calculus.  The problem is to
give a set of first order axioms that allow a proof that white can
win.  Of course, it is permissible to abstract away the specifically
chess aspects of the problem.

∂16-Jul-82  1315	JMC  
To:   MDD, DCL    
I have one last caveat concerning award to Wang.  Does he want it?
If he regards his work in theorem proving as unimportant, having
him give a lecture on it may be an embarassment both to him and
to the audience.  Martin, do you know is attitude?  If not, would
you sound him out before we go public.

∂16-Jul-82  1416	JMC  	postponement  
To:   bmoore at SRI-AI 
Unless others have a lot to say, I would like to postpone the
session on order of circumscription scheduled for 4pm to a time
next week to be specified on Monday.  I have no scheduling problem
and could come by, but I haven't thought about the problem enough.

∂19-Jul-82  1727	JMC  
To:   bmoore at SRI-AI 
Any chance of moving McD back to 3?  I have a conflict with 1:30.

∂19-Jul-82  1751	JMC  
To:   bmoore at SRI-KL 
	10:30 for McD would be fine.

∂19-Jul-82  1844	JMC  
To:   untulis at SRI-AI
Have you Bob Moore's home phone?  A seminar time change is dangling.

∂19-Jul-82  1850	JMC  
To:   untulis at SRI-AI
Many thanks.  I didn't know which Robert Moore he was.

∂19-Jul-82  2156	JMC  
To:   bmoore at SRI-AI 
Thanks Bob, I'll call in the morning.

∂20-Jul-82  1506	JMC  
To:   konolige at SRI-AI    
What time Thursday your talk?

∂20-Jul-82  1508	JMC  
To:   MDD
CC:   DCL   
The present form of the Wang citation is fine with me.

∂21-Jul-82  0041	JMC  
To:   MDD, DCL    
Let's accomodate Woody.

∂21-Jul-82  1257	JMC  
To:   MDD    
I also agree.