COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	ailog[s88,jmc] AI in logic for press
C00005 ENDMK
C⊗;
ailog[s88,jmc] AI in logic for press

	McCarthy's main work in artificial intelligence has been
theoretical.  Since 1958 he has developed ways of representing
common sense knowledge and reasoning in the languages of
mathematical logic.  The idea is that an intelligent computer
program should express much of what it knows about the world in
general and also about the particular situation in mathematical
logic.  It should also express its goals, and decide what to do
by reasoning that a certain course of action had the best chance
of achieving its goals.  Originally the only known way of doing
the reasoning in a precise way was logical deduction, but in the
late 1970s McCarthy and others developed systems of so-called
``nonmonotonic reasoning'' to supplement logical deduction.
McCarthy's system of nonmonotonic reasoning is called
circumscription, and it has been quite popular both in forms
developed by McCarthy and in variants developed by other AI
researchers.

	Nonmonotonic reasoning allows jumping to conclusions on
the basis of limited information that may have to be retracted
when more information is available.  For example, when someone
asks a handyman to build a cage for her bird he plans to put a top
on it, but if the bird turns out to be a penguin he changes his
mind.  It wasn't easy to understand the reasoning required for
this well enough to make computers do it.