Author: Robert Hyatt
Date: 19:06:05 07/08/01
Go up one level in this thread
On July 08, 2001 at 20:59:53, odell hall wrote: >On July 08, 2001 at 10:29:11, Robert Hyatt wrote: > >>On July 08, 2001 at 06:06:17, odell hall wrote: >> >>>On July 08, 2001 at 00:13:06, Robert Hyatt wrote: >>> >>>>On July 07, 2001 at 23:21:52, odell hall wrote: >>>> >>>>>> >>>>> I really don't understand what your saying, since even on icc computers are >>>>>performing like supergrandmasters. >>>> >>>> >>>>I'm sure you don't. But when you have time, point out a single "super GM" >>>>that loses to a 2000 player. The computers are good, no doubt. But they >>>>are not _that_ good yet... >>> >>> >>> I am curious Dr. Hyatt, you did accept deep thought as a ligitamate Grandmaster >>>correct? If this is true, how many games did Deep Thought play before you >>>believed it's GM Status? Why do you Accept Deep Thought as GM and not Micro >>>Computers? Is it because of your personal connection with HZU? Assuming of >>>course that you do accept Deep Thought as GM, I think i have heard you in the >>>past say this, although i could be wrong. I do mean Deep Thought and not Deep >>>Blue. >> >> >>No I didn't. Deep Thought never did the things necessary to actually 'earn' >>the GM title. It played like a GM if you look at its rating over a set of >>games. But it had horrible positional weaknesses. IM Mike Valvo totally >>crushed it in a 2 game match played on r.g.c when DT was active. DB was >>better. DB2 was even better and DB/DB2 had fewer obvious "holes" in the >>evaluation than DT did. >> >>Deep Thought played at a GM level to win the second-stage Fredkin prize. But >>the only requirement the Fredkin committee had to deal with was a rating of >>2550 or higher, over 24 consecutive games. That doesn't even produce a GM norm >>of course. >> >>So depending on how you phrase your question, I could answer yes or no about >>the micro/GM isue. >> >>1. Is the computer a GM? No. Not enough knowledge yet, by a _long_ way. A >>human GM knows so much more about the game. There are positions a GM can >>understand, while the computer is helpless. >> >>2. Does the computer play like a GM? yes. Because over a 40 move game, a >>human GM is likely to make a simple mistake, while the computer is not as >>likely. Of course, the machine will make mistakes, and the human may well >>spot them and end the game quickly. I expect this from Kramnik, but we will >>see. >> >>3. Could the computer earn the GM title from FIDE if given the chance? >>Most likely. If you enter it in enough tournaments, I believe it would be >>capable of earning the required norms. Unless enough computers are entered so >>that the humans begin to study them carefully and adjust their playing styles >>accordingly. > > > If your Saying that Computers must understand Chess, the way a Human >Grandmaster Understands Chess, in Order to be considered a Grandmaster, then i >have to agree with you That Computers will never be Grandmaster Players. I DON'T >think that even Deep Blue understands chess like a human Grandmaster, so let's >just say that Deep Blue isn't a Grandmaster either, according to your definition >of what a Grandmaster is. DB could well be a GM. It has significantly more chess knowledge than my program currently does. Notice that I said "could be". I am not particularly happy with the current idea of a computer GM... that if they simply play in enough events, and eventually produce a couple of "pseudo-norms" then someone wants to say "GM it is" and be done with it. Statistically that leaves me feeling a bit cold. Because I know that the machine is marvelously consistent, and that if I play in 100 tournaments, the odds are very high that I will do well in a couple of them because of nothing more than this consistency mixed with the inconsistency of humans. IE let's place a bet that I can flip a coin and get 10 consecutive tails. I _know_ that if I build a "robot flipper" it will force you to pay off the bet, because if I flip enough, 10 heads _must_ eventually come up, otherwise the coin is not "fair". That is how a computer will get its "norms". It will just plug along and eventually enough humans will "break" against it in the same tournament and it will do well enough to pull it off. I believe _any_ program could do this today. I mentioned that back in the 1980's, Fidelity entered multiple machines in the US Open, at the same event. Some did horribly. One would invariably do well. That was the one you read about on the front of the package. :)
This page took 0.01 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.