COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	     Summary of the US/Japan Seminar: Cooperation Through Competition
C00011 ENDMK
C⊗;
     Summary of the US/Japan Seminar: Cooperation Through Competition
			     June 18-21, 1985
			      Tsukuba, Japan
			       sponsored by
	      The Japan Society for the Promotion of Science
	      The United States National Science Foundation,
				   and
		   The United States Information Agency

Introduction

This seminar, scheduled in conjunction with Expo 85, was dedicated to 
Dr. John R. Pasta, who contributed so much to the development of computer
science, to the understanding of social issues, and to cooperation between
the United States and Japan. The organization and planning of the seminar
was a joint effort of Professor Bernard A. Galler and the late Professor
Tohru Moto-oka. This summary is dedicated to Professor Moto-oka, who 
contributed so much to the seminar, but who was not able to participate after
all.

The seminar was taped, and the summary which follows was taken from the
transcription of the tapes. The decision was made to extract excerpts from 
the actual presentations to form the summary. What is missing, besides the
many examples that each speaker used, are the very useful discussions that
occurred after each formal presentation. It is hoped that the full transcript
will eventually appear as a separate publication, and that the entire discussion
will be a contribution to the fields of computer science and knowledge systems.
It is especially hoped that the seminar will be a contribution toward increased
cooperation between our two countries.

John McCarthy:

The core of "common sense" includes what every human being who is not
feeble minded knows about reasoning about situations and the facts 
concerning situations. By this I mean something static. If this is true 
in a situation, then that is true in the same situation. Events and how
they change situations in time, including causal relations and actions,
and facts about goals and their achievement, and space and motion
through space...

Here are some issues concerning common sense. The first is, is it necessary for
expert systems? I'm going to make some arguments that it is indeed necessary
for some kinds of expert systems, including expert systems that are needed
to achieve that goals that various people have set themselves for the next ten 
years or even a shorter time than that. Then the next question is, can it
be done with just a very large number of production rules used in the way 
in which production rules are used today? I'm going to argue that the
answer to that question is "No." Some things that I believe are going to be
required are theoretical knowledge, and the ability to use theoretical 
knowledge...The next thing that I believe is going to be required is 
something I call elaboration tolerance...We need systems that can be 
elaborated by addition rather than by rewriting from scratch...Basically speaking, 
theoretical knowledge is knowledge which is not used in the form of direct
execution (or anything like direct execution) of the logical statements
involved... Common sense reasoning is what you do with common sense knowledge.

John McCarthy (for Nils Nilsson, who was not able to participate):

It is true in the United States that mathematical logicians have recently
begun to realize that there's a lot for them in the advances in the use of 
logic that computer sciencce and artificial intelligence have instigated. 
In particular, non-monotonic reasoning (formalized) is an invention of the
AI field in all of the forms that have been developed since 1975.  Mathematical
logicians are finding this a worthwhile field, so I believe we will get more
work out of the logicians than we have in the past.  The other area of
application of logic in computer science is the formalization of properties of
programs and proving the properties of programs... I think that computer 
scientists should learn more mathematics than is currently applicable to
computer science. In particular, computer scientists should be at home with the
general concepts of abstract mathematics. The areas which are good for this 
are things like modern algebra... It's always true that engineering doesn't 
use all of the science that is available to it. That is certainly true of
expert systems. On the other hand, many of the expert systems use LISP, which
has been strongly affected by logic.  Others use Prolog, which has even more
strongly been affected by mathematical logic. Certainly the concepts of
predicates and defining predicates is common in many expert systems. Some of
the expert-system writing systems (for example, the ART system produced by
the Inferece Corporation), use logic a bit more strongly... In terms of 
early education, one has to emphasize things that have interesting problems
associated with them.  The parts of logic that are taught in finite-
mathematics courses in my opinion are very boring, and will make students 
the enemy of of logic.  What I would like to advocate at least as long
as people are not able to think of good problems, is that the study of 
logic should be postponed until its application to knowledge representation
can also be taught, so people will see it as being an interesting subject.
I don't care if they see it as being useful, but it should at least be
interesting.