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 Colonel Don Gordon    
↓Joint Special Operations Command
↓Fort Bragg, NC  28307∞

Dear Colonel Gordon:

	I'm sorry the little I had time to say about non-monotonic
reasoning wasn't clear.  However, I think it is important for
applying AI to the problems you discussed in your presentation.

	A reasoning system is non-monotonic if it allows conclusions
that might not be reached if additional facts were available.  The
example I gave was of so-called "default reasoning", where something
is to be assumed unless there is evidence to the contrary.  In that
case, a bird is presumed to be able to fly unless there is evidence
to the contrary.  A military example might be that an enemy weapon
is assumed to have ammunition unless there is evidence that they
are out of ammunition.  An AI system must be capable of making
default assumptions subject to retraction if the relevant evidence
becomes available.  This involves different AI technology than when
non-monotonic reasoning is not required.

	Your presentation was quite striking, and I believe that
it is very important to overcome Soviet superiority in numbers
of troops and weapons.  I also believe that AI can help, and I hope
you will succeed.

	Your presentation raised another issue where there are
difficulties.  You gave the tic-tac-toe move tree as an example
of a conflict situation.  Such simple examples are fine, but
there is one characteristic of military conflict which is not
modelled by move trees and for which there is at present no
treatment in AI research.  This is the fact that in war, the
actions of the two sides are concurrent rather than consecutive
as in games.  Indeed since each side is composed of many subunits,
the concurrency is very great.

	I have no immediate suggestion other than to watch out
that you don't try to apply AI models that are inadequate.
I am sending a copy of this letter to Bob Kahn of DARPA as a
suggestion that he encourage AI research aimed at conflict
situations where there are concurrent events.

.next page

	I hope this will be helpful in some way, and I will help
further if I can.  At present this amounts to putting concurrent
action somewhat higher on my list of problems to think about.

.sgn