COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "let.pub[let,jmc]" source
C00006 ENDMK
C⊗;
.require "let.pub[let,jmc]" source;
∂CSL Dr. Wolfgang Bibel↓Institut fur Informatik↓Postfach 202420,D-8000
↓Munchen 2↓WEST GERMANY∞

Dear Wolfgang:

	Here are the promised comments on your Leningrad paper.  I looked
at it again at Herbert Stoyan's house.

	In general I agree with its emphasis on the importance of logic
in AI.  However, I don't agree with its dismissal of Tarskian semantics
which I think is also important for AI.  Suppose we are interested in
a theory of knowledge seeking programs.  Whether such a program obtains knowledge
and what knowledge it obtains depends on its interaction with the
world in which it operates.  A theory of such interactions
is required for AI, and one of its important features will be the
conditions under which the sentences that the knowledge seeker puts
in its database can be regarded as true about the world in which
it is imbedded.  Philosophically this is related to taking the
"third person point of view"; i.e. we regard epistemology and AI theory
as concerned with the relation between the knowledge seeker and
the world in which it is imbedded.  Having developed the theory for
arbitrary worlds, we are then interested in determining what is the
best fit to our own attempts to find information.  Some philosophers,
e.g. John Searle, insist wrongly on a "first person point of view"; i.e.
they start with Descartes's %2cogito ergo sum%1.

	My second problem with the paper is that I don't think it
will convince a skeptic of the value of logic in AI, because the
examples are trivial.  IS-A hierarchies are not enough, and
monadic predicates are a very special case.  The paper would be
strengthened by non-trivial examples, for example, axiom sets
describing the effects of actions - even if only in the blocks world.
In general I urge you and others to use logic to formalize non-trivial
common sense knowledge rather than just making propaganda about
its virtues.

.reg

P.S. I hope you enjoy Leningrad.  If you get a chance you might
inquire about Gregory Mints, who is a well known Leningrad logician
interested in applications to AI including non-monotonic reasoning.
He was at the last Leningrad conference, but since he applied to
emigrate he lost his job in the Institute of Mathematics
and is working as a programmer.  Tyuugu, who will probably be at
the conference, is in contact with him.  Mints is very much worth
talking to if you get a chance.