COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	%harvar.abs[e89,jmc]	Abstract for lecture at Harvard, 1989 Oct 2
C00005 ENDMK
C⊗;
%harvar.abs[e89,jmc]	Abstract for lecture at Harvard, 1989 Oct 2

COMMON PROBLEMS OF ARTIFICIAL INTELLIGENCE AND PHILOSOPHY

by John McCarthy, Stanford University

	Intelligent computer programs need a general framework
into which particular facts about the world can fit---that's
philosophy.  AI thus shares problems with philosophical logic,
epistemology, the philosophy of mind and the philosophy of
language.  However, AI suggests a more concrete approach to these
problems.  For example, our formalizations of knowledge need to
be able to express what a four year old child knows about buying
and selling, and such things suggest new approaches.

	The focus of theoretical work in AI has been the
common sense informatic situation---in particular expressing
the facts and reasoning that allow a person or machine to
reason about the effects of contemplated actions under
conditions where what phenomena are relevant is not known
in advance.

	The lecture will concentrate on two topics---formalized
nonmonotonic reasoning (about 10 years old) and a proposal for a
programming language involving speech acts---inputs and outputs
identified in the language as assertions, questions, commands,
permissions and promises.  We get program correctness criteria
based on the idea that assertions should be truthful and promises
should be kept.  Austin's distinction between illocutionary and
perlocutionary speech acts develops into notions of illocutionary
and perlocutionary specifications for computer programs.