COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	ai[w87,jmc]		DARPA support of AI
C00007 ENDMK
C⊗;
ai[w87,jmc]		DARPA support of AI

Dear Saul:

	I am quite worried by my conversations with Bob Simpson about
our renewal.  To put it frankly, I'm uncertain that we will be able
to come up with  proposal that will get renewed DARPA support in
spite of the fact that our group has made major contributions to
AI and computer science over the years and that the group is now
the strongest it has ever been.

	Let me mention five past contributions.  First there is the
idea that common sense reasoning is the key to general AI and the
use of mathematical logic for formalizing it.  The first paper was
done in 1958.  It also started planning beginning with axioms describing
the effects of actions.

Second there is LISP.
This has been the leading AI language for 25 years.  Its development
started in 1958.
Third there is the situation calculus, which started in 1963 and whose
major publication dates from 1969.
This and its variants and alleged improvements by various research
groups has been the main formalism for planning sequences of actions
to achieve goals.  Fourth Raj Reddy's speech recognition PhD thesis
in 1966 was the first work on speech recognition based on the
digital computer.  Fifth there is the circumscription method of
non-monotonic reasoning.  The first publication was in 1977, and
it was continued with major papers of mine in 1980 and 1986 and
important papers by Lifschitz beginning in 1985.  The significance
of this work is indicated both by the numerous references to it
in the literature and by my receiving the first Research Excellence
Award by IJCAI in 1985 as well as the Turing award in 1971.

	My opinion is that the work on the common sense database
and on formalized contexts has excellent prospects of being as
important as the situation calculus and circumscription.

	However, none of this work was accomplished according to
the ``directed research'' model which I understand DARPA to be
requiring for our future work.

	If DARPA continues to support our work, it seems likely that
we will proceed about as we have in the past.  Some fraction of our
work will be unsuccessful in meeting its immediate goals, and other
parts will result in major contributions, mainly theoretical.
I doubt that it is feasible for our group to work with an industry
collaborator, and anyway I don't have contact with such a potential
collaborator.  The reason is that my method of work is to think about
a variety of problems at once, and write up results when I have them.
I have not been successful in predicting which of these efforts
will succeed in which time period.