Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A new Deep Blue ?

Author: Robert Hyatt

Date: 05:33:20 07/31/01

Go up one level in this thread


On July 31, 2001 at 07:01:15, Uri Blass wrote:

>On July 31, 2001 at 03:23:47, Christophe Theron wrote:
>
>>On July 31, 2001 at 02:59:10, Joshua Lee wrote:
>>
>>>Is it possible to use the VIA 800Mhz chips which don't require a fan (i think)
>>>in parallel to where any program now could outsearch any Deep Blue of the past?
>>>
>>>It might take a few thousand chips but if it looks deeper it should mostly be
>>>better.
>>>
>>>I do know that there were supposed to be hundreds of parameters so i would bet
>>>it is better positionally than most comps now
>>
>>
>>???
>>
>>What they call a "parameter" is for example the value of a white rook on A1.
>>There are 6 piece types, 2 sides, and 64 squares. That makes already 768
>>parameters.
>>
>>Hey I've got thousands of parameters in Tiger then!
>>
>>They were better in search, that's for sure, because of their processing power.
>>
>>Better in evaluation? That remains to be proved.
>
>Better in search?
>It also remains to be proved.
>
>More nodes per second does not mean better in search because other programs may
>have better search rules.
>
>I believe they were inferior in evaluation.
>The size of the evaluation function is not the important thing.
>The important things is to get the right numbers in the evaluation function.
>
>If you have wrong tables the fact that you have big tables is not going to help
>you.
>
>Uri


If their evaluation numbers were so wrong, and their search was so weak, _I_
for one wish I could make those same mistakes.  What other program has come
close to beating a current world champion in a match?  Particularly when it
was obvious that Kasparov wasn't "playing for a draw in any game."\



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.