Author: Guido Schimmels
Date: 01:26:18 07/30/98
Go up one level in this thread
On July 29, 1998 at 12:03:08, Bruce Moreland wrote: > >On July 29, 1998 at 11:42:07, Don Dailey wrote: > >>I personally believe things should be looked at in terms of "computing >>systems." A computer is a computing system, and a single person is >>a computing system. But there is no reason in principle a computing >>system could not be 2 humans working together, or a single human >>with access to chess books during the game. Any particular computing >>system should be a separate entity and rated differently. Joe with >>chess books, would be a different computing system than Joe by himself. > >Right. I have done some thinking about this because this comes up on ICC. > >You can play against an entire shopping mall full of people and it wouldn't >affect the way you see the game, you still see a board in front of you on your >computer at home. > >I wouldn't mind seeing anything that was consistent: a person, a person with >ECO, a computer, a person overruling a computer, two people, etc., as long as it >is fairly consistent and is describe in the account's notes if it is not one of >the pure cases of one computer or one human. > >I think there are lots of interesting experiments waiting to be done. > >bruce Maybe I shouldn't have used the word " fair", because that was not really my point. I have no problem to agree with all Bob, Don and you replied, don't be surprised. My point is derived from the excessive discussion we had recently if micros have reached GM level or not. I think the answer to that question depends on the playing conditions - and yes there are no fair or unfair conditions - different conditions define a different discipline/scientific experiment. The only reason why humans would compete against computers is because it's interesting. As soon as neither GM's nor sponsors find it interesting, it will stop - and I'm afraid we are not very far from ! That's why I would suggest a new discipline in computer chess where humans are allowed to access external information and a chess-board for analysis purposes. The most elegant way to do this of course would be a laptop with your favorite database installed. Otherwise we are quickly running out of opponents - remember how pissed Anand was after the match with Rebel, and there were others as well ! In this new discipline I guess 2200 Elo players would become interesting opponents - again. Isn't it in the interest of the computer chess community ? We've seen a few matches between computers and humans which made the computers look very strong: Kasparov-Deeper Blue (not a micro, but nevertheless an indication) Hiarcs- Hergott M-Chess - Efimov Rebel - Anand Rebel - Yussupov (did I forget one ?) These matches indicate, the ability of humans to learn and adapt to the situation is vital for their superiority. If you let GM's play new versions of your program and have a master do the learning for the computer before the match and between the games, you take away a big part of this human advantage and your system is likely to play on GM level. Maybe things would look different if these matches would last longer and the GM's regain part of their learning advantage - and again you see, different conditions different results ! But for example I've Rebel Decade 1.2 (thanks Ed !), and score 100% against it, as it has a very slim opening book and no book learning, so I have winning lines for every opening line of Rebel's ! My DWZ (German Elo) is 1820, what does it tell us about the strength of Rebel Decade 1.2, what is its "real" strength, 2250 or 800 (or whatever) ? Meanwhile all of the top programs use some kind of book learning and the huge books provide a big variety, which helps a lot here - but nothing beats the "accelerated learning" done by human intervention. And don't underestimate how Kasparov got psyched out by the Deeper Blue team changing the bishop/knight ratio between the games ! Now I hope I have managed to make my point clear. If we talk about the strength of computers expressed in *FIDE Elo* we have to define the conditions in advance otherwise it's a useless discussion ! Or does anyone doubt Anand will humiliate Rebel 10 at home after a while (as much as he would beat Kasparov if he had access to his secret analysises) ? - Guido -
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.