Author: Bruce Moreland
Date: 10:07:36 11/10/98
Go up one level in this thread
On November 10, 1998 at 03:47:08, odell hall wrote: > > After The Last Game of the Match Between World Champion Garry Kasparov and >Deeperblue, At the Press Conference Garry Announced "I want to assure everyone >here that if deepblue were to start playing real chess, I personaly guarantee I >would have torn it to shreds with no question". These words have echoed in my >mind every since the match ended, I had no understanding of them then and I do >not understand them now. What does garry mean by "real chess"?? I thought that's >what was being played in the first place. I wonder if this was just the angry >reaction of a man who has never expierenced defeat in match play, or if the >statement has any truth. I am assuming that garry means that if deepblue were to >play published games then other grandmasters could study the games and find >weakness. This is the way I take it as well, having thought about it for the first time just now. He had no way of understanding its strengths and weaknesses other than by playing with it games that counted. > However My understanding is that what makes deeper special and >radically more sophisicated then it's predessor is it's ability to change it's >style of play in mid stream, a credit to the Brillance of the deepblue team. I don't understand how this one got started. Game one he beat it badly, and in game two he received some pretty serious shocks. It made moves that he did not understand, and which he did not see capacity for in the first game. So this was a very different. But if I remember right, Murray said that the eval weights weren't tuned until *after* this game. So it was the same program that won game two as lost game one. I don't see how anybody who watches programs play can be surprised by this. If you have a position where the program's terms work right, and which contains the right kind and degree of complexity, a computer will do great, but there are also positions where a program will completely crater by move twelve. Same program. This is part of the advantage that programs have. If a GM is playing a human, and the human does something unspeakably dumb, from that point on they understand you and will never fear you again, and for good reason. You've shown what is inside you, and it's not enough to beat them, and it won't be enough tomorrow either. When a computer does something unspeakably dumb, it doesn't have to mean anything, which not only means that it might play way better in the next game, it might play way better on the next move. Humans don't have a good feel for this, so doing something dumb is kind of like a gambit. Your compensation is that they'll hold you in contempt even after they resign. >My >impression is that it was extremely presumptous of garry to say he would tear it >to pieces! What is this evaluation based on? Is he making the statement based on >his past expierences with strong computer programs, in which after several games >he was able to find weakness and exploit them? The fact is that he was unable to >repeat this strategy against deeperblue during the second match, so what makes >him think he could do it at a future time? Personally I don't think Garry would >have a chance against deeperblue in a future match, The Machine has already >demonstrated an unbeatable endgame and ofcourse garry's flair for tactics is >useless against a machine that calculates 1 billion nodes per second! The reason he's saying it is that he felt like he could have done better, obviously. He was under a lot of pressure and he doesn't appear to have handled it well. It is a human strength to be able to adapt and to aspire. If his mindset is better, and if he is better prepared, I think he would do better in a future match. From having worked with micros I think that humans overestimate them and underestimate them at the same time. I bet the same thing is true of DB. It would have been interesting to see Kasparov play in such a manner as doesn't repudiate all of his strengths. It'd be interesting to see if he could roll the thing directly, and I bet it's still possible. bruce
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.