COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	@make(letterhead,Phone"497-4430",Who "John McCarthy",Logo, Old, Department CSD)
C00007 ENDMK
C⊗;
@make(letterhead,Phone"497-4430",Who "John McCarthy",Logo, Old, Department CSD)
@style(indent 8)
@begin(address)
Professor Bernard Meltzer
Via Uzzarini 483,
Rossola, ZOCCA (MO)
ITALY
@end(address)
@greeting(Dear Bernard:)
@begin(body)
    	Many thanks for your letter of February 9.  It is nice to
hear from old friends and nice to get some reaction to one's ideas.

	I don't understand your proposal to "provide all terms
and relations with an argument place for the name of the person or 
group of persons or robot or group of robots, in whose knowledge
structure (assumed identical in all its members, in the case of
a group) they occur".  Is it proposed to write

	telephone(Mike,pat)

for Pat's idea of Mike's telephone number or perhaps

	telephone(Mike(pat), pat)

or

	telephone(Mike).Pat.

Taking your idea literally of adding arguments to terms is difficult,
because terms don't have arguments; functions or relations do.  Consider
a compound term

	telephone(brother(mike));

do "telephone", "brother" and "mike" all get extra arguments or just some of
them or is the extra argument somehow attached to the expression as
a whole.  Each of these can be done and solves some problems but leaves
others.

	The general idea of regarding each robot as having a world model
and indexing terms with this model has been used in many progrms, especially
the Micro-planner programs of the early seventies.  It works as long as
the knowledge state ascribed to the robot has a certain structure.  Namely,
the robot must have either no idea of the color of a block or a complete
idea.  It cannot express the statement that the robot knows or believes
that the block is either green or blue.  Bob Moore, in his 1975 M.I.T.
Master's thesis (since published by Garland), emphasizes the problem
that disjunctive belief gives for ascribing world models to persons.

	I believe (but am not sure) that the Schank and Abelson
formalisms also are unable to represent beliefs and knowledge of
disjunction in a satisfactory way.
@newpage

 	Let me weakly defend my usage in the paper.  I say @p[mike]
is Mike himself as a translation between my notation and conventional
English.  Since I didn't put quotes around the word, the usual convention
is that I am using it not talking about it.

	Your radical proposal that "language never describes reality,
it only expresses thoughts about reality" is unclear to me, because
I regard a sentence describing reality as also expressing a thought
about reality.

	It has been my experience (and I believe it is also the
experience of philosophers) that general reasoning without examples
often cannot escape confusion.  Therefore, I suggest that you try
to write out some examples in some detail of how you would propose
that a robot represent what it knows or believes about a particular
situation and how new beliefs about a situation follow logically
from earlier ones.
@end(body)
Best regards,




John McCarthy
Professor of Computer Science