COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002		This is a request for [continuation] renewal of NSF Grant xxx
C00007 ENDMK
C⊗;
	This is a request for [continuation] renewal of NSF Grant xxx
supporting Basic Research in Artificial Intelligence.  The renewal will
continue for the second year, the activities proposed for three years in
our proposal of zzz.  The amount requested is yyy in accordance with the
budget submitted in the previous proposal.

Proposed Research

	During the academic year 1979-80, Professor John McCarthy
will be at the Center for Advanced Study in the Behavioral Sciences (CASBS)
at Stanford.  There he will take part as chairman in a study on
artificial intelligence and philosophy as part of a group of seven
computer scientists, philosophers and psychologists.  The study is
His work at CASBS will be precisely along the lines of the proposal
we are proposing to continue, and the grant will mainly
finance his participation.  The activities of the other participants,
though of course in accordance with their own scientific interests,
will contribute to solving the philosophical problems of formalizing
action and knowledge, which we have argued are fundamental for further
progress in artificial intelligence.  Appendix A describes the study
further.  Its objectives are not more fully prescribed than is expressed
in that appendix.

	McCarthy's personal research in that period will probably center
on non-monotonic reasoning and the further formalization of concepts.

	Non-monotonic reasoning, as described in the paper
%2Circumscription Induction - A Way of Jumping to Conclusions%1, is
increasingly seen as a requirement for AI systems.  Drew McDermott of Yale
recently wrote and Jon Doyle of M.I.T.  a paper on non-monotonic reasoning
proposing a different approach from McCarthy's.  An impromptu
mini-conference on non-monotonic reasoning was held at Stanford in
November 1978 and attracted 50 participants who heard papers by John
McCarthy, Drew McDermott, Terry Winograd and Richard Weyhrauch.  The
proceedings of the conference will be a special issue of the journal
%2Artificial Intelligence%1.

	The work on on non-monotonic reasoning will involve comparing
circumscription with the McDermott approach and discussing how to
do non-monotonic reasoning in computational systems.  This may not
be easy, since none of the systems with the exception of pre-arranged
sets of default assumptions has any obvious implementation.

	2. Applying circumscription to proofs of non-knowledge
seems promising.

	McCarthy's approach to formalizing concepts has been further
pursued by McCarthy and also by Dr. Lew Creary.