Author: Uri Blass
Date: 13:52:21 08/30/03
Go up one level in this thread
On August 30, 2003 at 16:20:58, Timmay wrote: >On August 30, 2003 at 15:57:24, Uri Blass wrote: > >>The question is if humans know what moves are the best in quiet positions. >>I believe that there are positions that humans also do not know what is the best >>move and deep search can help computers to find better moves also in quiet >>positions. >> >>I said that it may be interesting to see the relevant positions that you find >>Junior7 better than Junior8 and Tiger14 better than Tiger15. >> >>Maybe it can help the programmers to understand if they do something wrong what >>they do wrong. >> >>Uri > >Are you human Uri? How dare you say such a thing about the human race...grin. >But seriously, without opening books which are HUMAN knowledge gained over years >and years ALL the programs would be considerably weaker. Take away the opening >books of any top program and I bet 50 dollars I could beat it. If you know the program then you may prepare your win before the game but if you do not know the program then I think that the program has good chances to win or to draw. Computers can't >grasp long-term planning which openings of humans consider in its foundation. >1.e4 e5 2.Nf3 Nc6 can your computer find a better move than 3.Bb5? I doubt it. I am not sure if Bb5 is the best move and there are alternatives like 3.d4 It is an example for a position when I believe that humans cannot be sure about the best move. I guess that both Bb5 and d4 lead to a draw if we assume no mistake but I am not sure. >Computers aren't SO superior to humans that their weakening moves serve some >deep purpose that is unfathomable for humans. Not likely! We're both (humans and >computers) tackling the same game, chess. > >Think about human principles that we've developed ever since Steinitz. We've >discovered that its easier to win pawns if they're weak or isolated or backward. >So when Junior voluntarily makes a weakening move it could lose that pawn later >which could make him lose in the end. In the game where Garry beat Junior in the >g4 line of the Slav, Junior did not play good moves and ended up losing (part of >it was the book, but the playing wasn't spectacular either, there were many >positional inaccuracies which lead to the victory). There is not some mystical >thing that computers know that humans don't like you made it seem in your quoted >response. Computers know to search deeper so even without positional knowledge that humans do not know they may find better moves. You're forgetting, when you put code into your program that says >control the center, or make mobility important, who's knowledge is that? It's >human knowledge! The computer bases its move choices on human knowledge which >guides mindless search. So everything stems off the human knowledge of this >game. So you should not downgrade the human factor. Humans also have the ability >to improve! Programs are always frozen by their code. This is only the situation for the top programs of today but I believe that it may be possible that we get in the future programs that can learn. It is not easy but I believe that it is not impossible. If you produce a computer >that thinks it knows what its doing by playing a4 on the first move followed by >Na3, I bet you I can beat it. I am not sure about it. black is better after 1.a4 e5 2.Na3 d5 but there is no easy win for black. Even though its an "infallible computer" which >knows something we humans don't. So don't criticize the human race by saying >such statements as "the question is if humans know what moves are better in >quieter positions." There's no doubt about it, we do! I admit there are cases >when weakening moves of computers according to human standards have been >unpunished in the past, but they've also been exploited you can't forget that. >Best wishes. Interesting conversation. There are cases when humans know better. There are cases when humans believe that they know better when they are wrong. There are cases when humans do not know better. Not in every case weakness can be exploited and there may be cases when humans believe that a weakness can be exploited in the future and their evaluation is wrong. 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.