Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: "I hope the GM doesnt get the program before the match", means >

Author: Uri Blass

Date: 13:56:54 09/21/02

Go up one level in this thread


On September 21, 2002 at 15:30:35, Joachim Rang wrote:

>On September 21, 2002 at 13:39:17, Uri Blass wrote:
>
>>On September 21, 2002 at 13:00:05, K. Burcham wrote:
>>
>>>
>>>
>>>when I first learned that this statement is used and accepted often, I had to
>>>laugh.
>>>when anyone makes this statement, they are admitting that programs have enough
>>>problems that if a GM has the program before the match, that these problems can
>>>be discovered, and used during the match to play for the win.
>>>I find this amusing.
>>
>>The interesting match is when humans do not get a copy.
>>I could understand giving humans a copy if computers proved themselves to be
>>superior in normal conditions but computers still need to prove that they are
>>superior in normal conditions.
>>
>>I can add that the only way to beat humans who get a copy before the match is by
>>some non deterministic behaviour otherwise the human can  learn the right games
>>before the match and the right games can be even games of the program against
>>itself.
>>
>>Uri
>>
>>
>
>this would mean, that one should be able to repeat "killergames" in an official
>match. Due to a big opening book and the right for the computerteam to choose
>the first move, this won't happen.

No

White may get the opponent out of book very quickly by means of 1.e4 2.Qe2
so I suspect that it is enough to learn 20 games to win with white against
deterministic machine.

White get almost equallity after 1.e4 2.Qe2 and it is possible to beat the
computer later because computers are not perfect.

Of course it is possible to build a big book to make similiar ideas impossible
but I doubt if the Fritz that kramnik got has it.

Uri



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.