Report on ICOT
Mark E. Stickel
June 9, 1992
ics. MGTP thus joins the very short list of theorem provers that have solved open
problems. The contribution of ICOT's parallel hardware to this proof is noteworthy
as well. One of these open problems was solved in 3 hours on a 256-processor PIM;
on a single processor of the same power, this would have required waiting a month for
the solution. This represents a qualitative difference in the theorem proving process
that makes otherwise nearly unthinkable tasks doable. Inoue and others have also
been doing excellent work in demonstrating the usefulness of MGTP-style reasoning
for nonmonotonic reasoning, diagnosis, etc. Many AI reasoning problems seem natu-
rally formulatable for execution by ground MGTP. Upside-down meta-interpretation
permits the bottom-up MGTP to be imbued with goal-directedness.
Outside of theorem proving, I think ICOT's contributions are many. ICOT's scien-
tific contributions, particularly in the area of logic programming languages, practice,
and theory, are competitive with that of other research institutions around the world.
ICOT is an internationally recognized research center. Through foreign visitors coming
to ICOT, ICOT researchers visiting overseas, ICOT's organization and participation
in conferences and workshops, publication of research results in technical reports, con-
ference proceedings, and journals, often in English, ICOT and Japan are participating
strongly in the international computer science community. ICOT's making the software
developed in the Fifth Generation Project freely available is an extension of ICOT's
deliberate policy of openness and is a noteworthy and generous contribution to the
scientific community.
I think ICOT has been very successful in training its researchers in logic program-
ming, parallel processing, and the methods and values of computer science research.
Perhaps they can bring about greater commercial use of logic programming when they
return to their companies. Still, the barriers to new programming methodologies in
industry seem high. With my long experience programming in LISP as well as Prolog,
industry's failure to recognize the value of alternative methods has long been obvious
and disappointing. Considerations other than technical merit often determine what is
successful, such as MS-DOS and C.
ICOT's objectives seem fundamentally right. Parallel processing is the right way
to provide lots of computing power cost-effectively. MIMD architectures are more eas-
ily used for a variety of applications than more restrictive computational approaches.
A focus on symbolic computing applications is a needed counterpoint to the usual
emphasis on supercomputing for numeric, scientific applications. The KL1 program-
ming language is a major accomplishment of ICOT. It is an elegant parallel logic
programming language that facilitates writing parallel programs while easily avoiding
synchronization errors. Writing PIMOS and earlier operating systems entirely in logic
programming languages is an massive demonstration of the suitability of such languages
for low-level systems programming as well as high-level applications programming.
- 113 -