COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00005 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source
.place text
	I regret that I haven't time to read this proposal and
the associated papers thoroughly.  I read very slowly and have
to concentrate my reading on papers relevant to my own work.

	However, a preliminary reading justifies some conclusions.
First, the proposal seems promising as a means of improving the
poker playing ability of this program.  It seems that experience
with it is being competently taken into account.

	The general rationale of the research seems to
be to get numerical measures of the situation that can be weighed
in a decision making procedure.  While this is certainly a reasonable
approach to poker, it isn't clear how much human intelligence works
this way or how much intelligent behavior can be achieved this way.

	The proposal should contain a statement of how the research
will contribute to making machines intelligent.  Perhaps such a
rationale was contained in Findler's earlier proposals.  The closest
thing I could find to it was in the last section of %2Studies in
Machine Cognition Using the Game of Poker%1.  While the author is
correct that decision making under conditions of uncertainty is
important, it seems to me that poker provides a situation in which
numerical estimates of utilities and probabilities that have to
be guessed in say military examples are provided free of charge.

	In making these comments, I must admit to a previous
opinion that the ability to identify qualitative factors is more
of a problem in artificial intelligence than the ability to
estimate quantitative factors correctly.