Author: Christophe Theron
Date: 13:00:54 05/22/98
Go up one level in this thread
On May 22, 1998 at 15:05:37, Thorsten Czub wrote:
>On May 22, 1998 at 13:53:59, Christophe Theron wrote:
>>The point in fact is not being fast, but to search deep enough that some
>>positional issues begin to be obvious. When you get this, you REMOVE the
>>positional knowledge associated with these issues, and try to
>>concentrate on the kind of knowledge that search will be unable to find
>>(long term planning for example).
>>
>>By saying this (remove knowledge), I know that some people will judge me
>>as a kind of heretic. Those people want to believe that the human
>>knowledge and the human way of searching a chess tree is the most
>>beautiful thing in the universe, and the ultimate goal for a chess
>>program. Of course I disagree.
>>
>>
>> Christophe
>
>Christophe - forgetting about some things is the most important feature
>of our brain. The whole life the human beeing has to forget the NOT
>ANYMORE USED or the NOT USEFUL "knowledge".
>We have to forget when the pain in the brain is to high, we have to
>forget the night after we learned for an exam, we have to forget what we
>don't like.
>
>Computers don't forget. They memorize things different.
>If they would be able to save information AND to select only "important"
>things, they would be able to generate SELF and EGO and feelings.
>
>Removing knowledge is the search tree is a very mechanical way of
>emulating
>this main feature of the brain.
>Since it is at least - an emulation - no matter how insufficient it is,
>it is ONE way.
But I'm not trying to emulate the humain brain, nor generate an ego.
Just trying to find a way to make a computer play better chess.
I know that I don't know how to do this perfectly. I don't take as being
the truth that the best way to do it is the human way.
I don't know what the truth is in this field. But if I see a way to go
in the direction of better play, I follow it. It may not work in the
end, and if it is the case I just go back and go in another way.
>I do understand your idea without calling you heretic.
>On the other hand I do not believe that we should try to generalize.
>The fact that Nimzo/Fritz/Junior have conquered the top-positions is a
>rare fact.
>Never before fast-searchers were able to play as strong as the slow
>programs.
>
>The last time this happened was in 1984 with superconny and
>brute-force-module and later called MM2 and in those times the
>MATERIALIST also told all of us that knowledge programs and evaluation
>is wrong and fast-search is leading for the rest of the time.
>This was a lie, as we can see from the history.
>
>This is a competition. Sometimes you are up. Sometimes you are down.
>If the fast searchers can stand as long as the other slow searchers have
>dominated (years !) they are good. I don't think this will happen,
>but...
>
>Your program ChessTiger 11.2 is a very good example how fast searchers
>can play good and nice chess.
Chess Tiger is not that fast.
I've been working on the search for a long time. I have focused on
trying to go deeper. I thought I would never be able to make a decent
positional evaluation. But I discovered that the search was able to
understand positional issues. So it was easier than I thought to work on
the evaluation.
>It has shown in Paderborn that it can play very strong and stand all
>kind of programs (despite Gandalf --- grrrrrrrr !).
>I would not have played so many games with your program if i would not
>like it, since it has no autoplayer function and I have to outplay
>anything manually :-)))
What about Tiger 11.4? Any new games to show me? Not a lot of feedback
from you this time...
>So - to a certain degree, i understand your ideas and support your
>effort.
Different ideas, different people, if we try to communicate we build
something better. That's why I love CCC!
Christophe
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.