COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	I have known Raymond Reiter's work since a bit before 1980, when
C00005 ENDMK
C⊗;
I have known Raymond Reiter's work since a bit before 1980, when
his Logic of Defaults approach to non-monotonic reasoning was
published in Artificial Intelligence along with my circumscription
approach.  I have followed his subsequent work always with interest.

His (1987b) survey of non-monotonic reasoning was very informative
and has become the standard survey.  In my opinion is work on
diagnostic reasoning greatly simplifies and enlightens the subject.
It was certainly a new and bold approach, since it was based on
the idea of satisfiability.  In general, satisfiability is undecidable
and this discouraged people from emphasizing it.  However, Reiter
points out that in many important kinds of diagnostic problem,
satisfiability is reasonably tractable, and it provides great
conceptual simplicity.

I cannot comment at all on the computer vision part of his current
proposal.

I think the other parts of the proposal are excellent.

Reiter is one of the originators of work on deductive databases,
and his proposals for making queries more computable have considerable
practical and theoretical potential.

His ideas with Levesque regarding the use of metatheoretic methods
in dealing with integrity constraints mesh with ideas in non-monotonic
reasoning pursed by Robert Moore and which I am also pursuing.
There is good potential here.

I don't understand exactly what he proposes to do about diagnosis,
but his previous success as mentioned above justifies optimism.