Author: Thorsten Czub
Date: 03:51:09 10/16/00
Go up one level in this thread
On October 16, 2000 at 01:06:12, Robert Hyatt wrote: >Maybe. IE I wouldn't mind writing a program that everyone said played like >Karpov. He was one of the best ever. Boring? Yep. Consistent? That too. >I pulled for Kasparov when they first met, because I liked the wild style. >But I began to see the beauty in boring, consistent, mistake-free chess as >well. :) and even if you play like kramnik, you are not ACCURATE at all. even karpov makes mistakes. a mistake free chess program ? Only a database could be without mistakes. in life you have to risk. otherwise the others will overtake you. >This is true. And I would bet that if he were playing today, most of his >wild sacrifices would be blown apart by computers. and ? this would still not make the computer programs play stronger. what do you want ? a program that beats chess-players ? a program that beats chess programs. A program that beats all ? The holy grail of computerchess cannot come from programs like fritz/genius or whatever. programs have to be both, accurate AND understand sacs like Rc6. otherwise they will never learn how to play chess. they will always remain dump and fail when humans play moves like Rc6. how can you handle those moves when you don´t know what they mean. > If he were even able to >reach those kinds of wild positions. I would certainly like to see it happen. >We've had a couple of what I would call tal-like anti-computer players on ICC. >IMOrlov is one that comes to mind. Wildly tactical. Kills humans, including >lots of GM players (at fast time controls). Computers hand him his head in >a sack, however. ah - come on. the todays programs are stupid. the only reason the humans lose to them is, because any human makes mistakes and has problems to play accurate. but not because the humans play weaker chess. kramnik is unable to win a won game against kasparov. not because he is an idiot or plays weak chess, he played a very good game, kasparov looked like an idiot of 1700 in that game, but then kramnik lost fade. and THIS is the reason dump programs beat grandmasters. not because these programs play strong chess. they use the mistakes and the inaccurate moves humans do. but still a few humans are capable to kill machines. and when the last human beeing has no chance anymore, and your mega stupid program has overtaken world-chess, it will still be beaten by Rc6-programs ! >That is the holy grail of computer chess. But until we have some idea how >the human does it, it is not going to be easy to do. ? we know how they do it. they KNOW that Rc6 makes sense. they learned typical situations like this. like a child learned to talk by copying. > Chris liked to talk >about "fog". I personally consider that "hot air". because you don´t understand anything. its hot hair when you talk about accurate chess. because accurate chess is impossible. only when you have tablebases. CSTal was not hot air. it was the effort to teach a chess program to leave the path of the quiesence-paradigm. to drive directly into the fog where you have unbalanced positions with unclear scores. with different material in all kind of evaluations. and cstal was able to decide using knowledge which positions are good and which not. it learned to handle chaos. this is not hot air. and it did not fail. cstal was a slow program. and only doing 15.000 NPS on a 400Mhz machine. but gambit-tiger does 120.000 NPS. this is the new paradigm. not hot air. by insulting or attacking those , by throwing with mud, you will not be able to handle Rc6 situations. > Probably more useful >for marketing than anything else. I don't believe it is necessary to emulate >a human to beat a human. BUT it is necessary to emulate a human to beat the dump machine that beats the humans ! > The computer doesn't work like the brain in any way >I can think of. I don't envision a program doing the same. Producing the >same result? Of course. But in the same way? We don't even know what that >"way" is after 30 years of people studying it from every angle imaginable. than you have to learn what chris/christophe wanted to tell you. >It all depends on the definition of "strong". Which was my point. IMHO, >"strong" is not understood by programs yet. Better? Maybe. But not good >enough by any measure. Against humans? Probably works better than when playing >a computer. But even GMs will recognize if "strong" is not well-defined. I >have had some great attacking versions, some planned, some by accident. And >all scored early successes against humans. But most got "figured out" by the >better players and they then had a feast... you must try. not give up. do you have an idea how much time we spent in cstal ? >Hard to adapt until it is available. But once you release it, just sit back >and watch. :-))) >As I said, I am _not_ trying to put the thing down at all. Just bring a bit >of reason into a discussion that makes it sound unbeatable. Everyone likes that >style of play. The SuperConstellation used to do it all the time. But then >the weaknesses begin to show up. It is bad enough to create a positional >weakness for yourself. But to do so a piece down can be quite dangerous. :) :-)))) right.
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.