COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	notes[e83,jmc]		miscellaneous notes
C00010 ENDMK
C⊗;
notes[e83,jmc]		miscellaneous notes

1983 July 2
	Consider an  is-a  hierarchy or other collection of facts
represented as a connection graph.  Can we use the "Tower of Hanoi"
trick to assign co-ordinates in some n-space to the nodes of the
graph that can be used to decide how to get from one vertex to another?

Carolyn suggests a 206 problem of assigning co-ordinates to
points of a graph to make paths easy to find - in the style of
my Tower of Hanoi problem.

1983 July 3
	The most important heuristic is hill climbing.  Heuristic
adequacy often consists of inventing spaces to hill climb in.
Is GPS different from hill climbing?

	A beginner at chess brings certain heuristics with him.
1. Possessions are power.  He is ready to capture and regrets
being captured without being told.
2. The geometry of the board.  He knows that to get somewhere one
moves towards it.  E.g. if he is asked to move his knight or king
to a square on the far side of the board, he doesn't have to study
how to do it.
3. Design before construction.  Asked to mate the opposing king,
many players will build a mating position before thinking about
how to achieve it.  I suppose that many won't, but all will be
capable of receiving the suggestion to do so.
4. Perhaps it would be a good problem to write a program to
write a program to play chess.  The program would be given
only heuristics that could be argued as being part of common
sense; it would have to develop the chess heuristics.
5. Alpha-beta raises an interesting question.  Chess beginners
don't have to be told to use alpha-beta, or some approximation
thereof.  Why not?  Presumably it is a consequence of some
more comprehensive common sense heuristic.  Well, it's used in
games against nature.  We enter a situation with an initial
alpha and quickly reject alternatives immediately seen to
yield bad results.  Only if alternatives are seen to be worse
do we return to the initially rejected action.
6. Perhaps it would also be worthwhile to study beginners learning
chess and observe the transfer of common sense heuristics into
chess.  It would be interesting also to compare the learning of
chess by very intelligent people with that of less intelligent.
Children would be good subjects.

"The System for Business Automation(SBA)" by
S.P. de Jong from Proceedings of the 1980 IFIP Congress, IFIP, Tokyo, 1980.

The circumscription paper needs a remark that we don't always circumscribe.
Moreover, it is possible to circumscribe the predicate  closed-world,  so
as to make the default assumption that there is something new to
be found.

A useful domain for CBCL is advertisement.  We sell shoes, ships and
sealing wax.

Can we formulate axioms that give outcomes amounting to the weighing of
 conflicting evidence using circumscription?

%3Hewitt, Carl and Peter de Jong (1982)%1: %2Open Systems%1, M.I.T.
AIM 691. December 1982.
Propaganda piece on actors vs. first order logic.

1983 July 7

Thought while running: The following seems to be missing from the
papers on planning reprinted in v. 3 of the AI Handbook.  In the
discussion of the blocks world, the programs cannot learn or be
told the fact that putting a block above another that is not in
final position and which must be lower in the tower than the block
just placed is an interference.  We want a blocks program to which
this can be an addition - not a change.  As such the program can
learn for itself or it can be told.  First we need merely to have
it as a sentence in a logical language.

Some notation:

must(move(block,location),s).

This means that the block must be moved eventually - not necesarily
right away in  s  itself.  Perhaps it is better to include a goal
as a parameter, so we have

requires(goal,move(block, location),s).

Maybe for some purposes  requires  is too strong and we want merely
suggests.  Also for this purpose we need a more general notion
of location, since we are considering moves where the precise
location has not been determined.

interferes(change1,change2)

In the present situation making  change1  intereferes with making  change2
in the sense that after  change1,  change2  will require more steps.

1983 July 11
Corky's example

g(0,0,*) = 0
g(1,*,0) = 1
g(*,1,1) = 2

cannot be definedusing conditional  xpressions and parallel or.

1983 July 19
Conversation with Jim Myers of Library
Full text in sciences in chemistry.
Full text of a number of newspapers.
Doug Ferguson will phone.
Stanford on line catalog
Look up mode.
ECDMO, pass=socrates
C.I.T. dial up number

EXTENS.AX[S76,JMC] 15-Jun-76	Extensional forms
EXTENS[F77,JMC] 07-Oct-77	"#. Programs as objects and extensional forms."
NOTES[E78,JMC]/2p 30-Jul-78	iterativity is extensional in the functional
EXTENS[S76,JMC] 30-Jun-76	EXTENSIONAL FORMS
				ABSTRACT FORMS - abstract synt + ext forms

Send net mail to CMU for Centre Mondial.
Michelle Gaillard is Servan-Schreiber's chief of staff.
Janet Asbury is Raj Reddy's secretary.
ASBURY@CMU-CS-A
REDDY@CMU-CS-A