Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Muti processors, ok,then to be fair muli GMs in future matches

Author: Uri Blass

Date: 13:24:26 09/03/05

Go up one level in this thread


On September 03, 2005 at 16:08:57, Carey wrote:

>On September 03, 2005 at 15:15:58, Uri Blass wrote:
>
>>Humans do not have to think for themselves in opening position and they can >play
>>moves that they remember.
>
>Generally speaking, even the best player will have an opening 'library'
>substanially less than even a relatively simple program will have.

I am not sure about it and if it is the case then I am not sure if it is an
advantage of the program because the program may play wrong book lines.

I remember that in the last WCCC fruit was out of book relatively early even in
position that is known theory.

>
>Additionaly, players usually don't do this by rote memorization.  They play over
>it (using books etc. as guides) and they learn how to play it and whether their
>style fits the required type of play.  They study it.  They learn *how* to play
>it.

I am not sure about it and I remember that even kasparov in one of his games
lost because he did not remember the right line that he analyzed at home some
years earlier.

>
>So when they do play an opening, it's generally based on what they've learned.
>Not by what they've memorized.

I do not think that you are right.

>
>That's not the same type of thing that most chess programs do with an opening
>'book'.  In that case, it's just rote memorization.  For a human, that'd be
>comparable to consulting an actual physical opening book.


I am not sure if consulting database for a human in that way is going to help
him because the human may not know what to do when his opponent play a move that
put him out of the database.

  Perhaps a few notes
>in the margin about various aspects.  But that would still be just data
>retrieval.  Not playing chess.
>
>
>>Huamns also do not need to think for themselves in endgames and if they >remember
>>the right move they are allowed to use their memory
>
>How is *learning* general endgame rules at all comparable to a program using a
>precomputed 6 piece endgame tablebase?
>
>One is learned and the other is just rote memorization / data retrieval
>containing no chess what so ever.

It is the same except the fact that computers are able to memorize more than
humans but they are also able to calculate forward more than humans so if you do
not allow them to use tablebases and opening book and more than one processor
you should not stop there.

You should not allow them to search more than 1000 nodes per move because humans
are not able to do it and you should not allow them to calculate evaluation
function because humans are unable to calculate the evaluation function that
program use today because it includes a lot of components.

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.