From newshub.ccs.yorku.ca!torn!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!jvnc.net!netnews.upenn.edu!sagi.wistar.upenn.edu Wed Sep 16 21:22:03 EDT 1992
Article 6811 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!torn!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!jvnc.net!netnews.upenn.edu!sagi.wistar.upenn.edu
>From: weemba@sagi.wistar.upenn.edu (Matthew P Wiener)
Newsgroups: comp.ai.philosophy
Subject: Re: Checkers master beats computer
Message-ID: <88116@netnews.upenn.edu>
Date: 7 Sep 92 17:42:04 GMT
References: <1992Aug30.195848.6377@unocal.com> <87846@netnews.upenn.edu> <1992Sep3.200000.16300@unocal.com>
Sender: news@netnews.upenn.edu
Reply-To: weemba@sagi.wistar.upenn.edu (Matthew P Wiener)
Organization: The Wistar Institute of Anatomy and Biology
Lines: 35
Nntp-Posting-Host: sagi.wistar.upenn.edu
In-reply-to: stgprao@st.unocal.COM (Richard Ottolini)

In article <1992Sep3.200000.16300@unocal.com>, stgprao@st (Richard Ottolini) writes:
>I was under the impression that because checkers moves and strategies were
>simpler to compute than chess,

Yes, checkers is smaller than chess.  It is still a huge computation.

>			        the computer can look ahead so far in the
>game and basically win by brute force.

That is what usually happens.  Ditto for chess.  Nothing close to creativity
or "intelligence" has come up in any of these programs.  Someone even gave
a rather all-too-typical "horizon effect" anecdote, where brute force loses.
Cray Blitz was, uh, enjoying a break in the 1984 US Open playing speed chess
against all comers, mostly experts, masters, and the like.  At the bottom of
its competition (1800), it shocked all kibbitzers (including a grandmaster)
by sacrificing its queen out of nowhere, eventually losing.  On the replay,
it was found out why: the 1800 had a forced mate-in-11, and the program opted
for avoiding the "sure" loss instead of the "uncertain" loss.  All the humans
evaluated "sure" and "uncertain" the other way around.  Not even with the
obvious clue that Cray Blitz chose a major (and useless) sacrifice did
anyone see the "sure" loss.

>					 But human mind can still devise
>strategies to defeat such a computation which is amazing to me. 

The current checkers' computations are not complete computations.  Ditto
for chess.  Moreover, checkers has an "ugly sister" sort of reputation,
and as such, I would not be so certain that Tinsley represents the peak
of human checkers performance.  That he's been on top for 40 years may
speak volumes about him.  Or about checkers.

There were comments in r.g.chess that more extensive book calculations
by Chinook and company may solve checkers within the decade.
-- 
-Matthew P Wiener (weemba@sagi.wistar.upenn.edu)


