Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!news.mathworks.com!tank.news.pipex.net!pipex!peer-news.britain.eu.net!newsfeed.ed.ac.uk!ainews!aifh!ianf
From: ianf@aifh.ed.ac.uk (Ian Frank)
Subject: Re: Computer scores historic chess win over Kasparov
Message-ID: <DMvF1p.IMA@aisb.ed.ac.uk>
Sender: news@aisb.ed.ac.uk (Network News Administrator)
Reply-To: ianf@aifh.ed.ac.uk (Ian Frank)
Organization: Dept. of Artificial Intelligence, Univ. of Edinburgh.
References: <4fjhff$b6p@Venus.mcs.com> <4frqsh$i33@news-f.iadfw.net> <4fvef9$2hs@news.bu.edu>
Date: Fri, 16 Feb 1996 13:34:36 GMT
Lines: 30

In article <4fvef9$2hs@news.bu.edu>, meissner@cns.bu.edu (Karl Meissner) writes:
> One of the developers of Deep Thought had an interesting comment, 
> that every time the speed of the chess hardware doubles, the machine's
> chess rating goes up by 200 points  (because of deeper and deeper look-ahead)
> By Moore's Law, hardware doubles in speed every 18 months.  Kasparov, 
> who claimed he would never be beaten by a machine, is in trouble.  If
> not this machine, then the next or the next after that...

This may be an over-simplification. The developers of Chinook (the
checkers playing program) for example, have documented how the benefit
of increased search, whilst intially providing a linear increase in
performance, tails off drastically after around 15-ply. I recommend
the following paper on the subject:

@INPROCEEDINGS{schaeffer-93,
 KEY          = "schaeffer-93",
 AUTHOR       = "Schaeffer, J. and Lu, P. and Szafron, D. and Lake, R.",
 TITLE        = "A Re-Examination of Brute-Force Search",
 BOOKTITLE    = "Games: Planning and Learning, Papers from the 1993 Fall
                 Symposium",
 PAGES        = "51--58",
 ADDRESS      = "AAAI Press",
 YEAR         = "1993"}

It's available on the Web, from the Chinook home page, or from my home
page:

http://www.dai.ed.ac.uk/students/ianf/Research/

	- Ian.
