Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!csn!ccncsu!handel.cs.colostate.edu!gordons
From: gordons@handel.cs.colostate.edu (vahl scott gordon)
Newsgroups: comp.ai.neural-nets
Subject: Chip vs Chess Master
Message-ID: <13807@ccncsu.ColoState.EDU>
Date: 28 Mar 91 05:27:41 GMT
Sender: news@ccncsu.ColoState.EDU
Reply-To: gordons@handel.cs.colostate.edu (vahl scott gordon)
Distribution: usa
Organization: Colorado State University
Lines: 33


The NOVA program has been a topic of discussion on rec.games.chess.

I enjoyed it quite a bit, but would have liked to see more discussion on
various approaches to solving the chess-play problem.

There have been non-brute-force approaches, with less success.  Probably
the best has been the program "AWIT" from the University of Alberta
(T.Marsland) which achieved a high class B/ Low class A rating.  It
searched very few nodes, something on the order of 500 or less, as I
remember.  It was not, however, a learning program.  Even HITECH (written
by Hans Berliner, also at Carnegie Mellon) uses a not-so-brute-force
approach, with more chess knowledge.  HITECH is currently the second
highest rated chess computer program, and is at the Senior Master level.
From a knowledge point of view, Deep Thought is actually one of the
least interesting.

I am not aware of any neural programs, or even neural programs for
subsets of chess.  I think it would be a great area of research (for
SUBSETS of chess, that is).

I do think that they overplayed the surreal aspects of the game just a
bit... Kasparov is an excellent tactician, and while he doesn't analyze
so many lines as Deep Thought, he probably analyzes KEY lines DEEPER.

Lastly, I think it would be interesting to hold a tournament to see which
program was best for, say, analysis of some ceiling of positions... say,
1000.  That way you'd have to put some chess knowledge in, and try to
find the best lines to consider.  It would be a better window into
human intelligence than brute force.  The games would be interesting, too,
because the programs could better explain why they selected certain
moves (i.e., why they chose to examine certain continuations).