Author: Robert Hyatt
Date: 19:56:31 01/12/01
Go up one level in this thread
On January 12, 2001 at 21:34:53, Peter McKenzie wrote: >On January 12, 2001 at 10:02:44, Robert Hyatt wrote: > >>On January 12, 2001 at 00:41:33, Garry Evans wrote: >> >>> A short while ago, i asked you on ICC, would you acknowledge that computers are >>>of Grandmaster Strength if Rebel Won the Match against Van der Wiel, your answer >>>Was yes!! So would you please honour this agreement and acknowledge here in >>>Public that computers are GM Strength? >> >> >>2-3 years ago my estimate was that the programs were at about 2400-2450 on >>the FIDE Elo level. I would probably change that to barely 2500 for today's >>much-faster hardware. I wouldn't begin to suggest they are beyond 2500 >>yet, however. They _still_ have a lot of weaknesses. > >Hi Bob, >I don't usually participate in this sort of discussion but hey, its a slow >progamming day :-) >Personally I'd bump that 2500 up to around 2550, which I guess is 'GM strength' >whatever that means exactly. > >I think its easy to over estimate the strength of humans, because they are >capable of playing very profound chess. However the practicalities of playing >chess free of tactical mistakes are definitely non trivial, even for GMs. >Relentless tactical pressure definitely works against GMs, a fact clearly >exploited by players such as Kortchnoi and Fischer. > >Also, we now have comps that are more than capable of exploiting small >positional advantages and grinding out points that way. > >I hear that GMs will 'learn to exploit computers', as if chess computers were >just invented yesterday. Of course they will score the occasional impressive >anti-computer victory, but I think these are becoming increasingly more >difficult to pull off. Perhaps the trend is more a case of the programmers >learning to exploit the GMs? > >cheers, >Peter I don't think we will really see how "bad" computers can be until we see the day when computers play in human events with regularity. IE until a GM is _forced_ to address the issue of computers, he isn't going to do so. A good curve-ball pitcher is simply bound and determined to throw his curve, until he finally realizes that there are a few batters that are going to knock him off the mound. Then he begins to learn which batters like the curveball and he throws them sliders or fastballs or changeups or whatever. But until _he_ (he being the pitcher) finally accepts the fact that he simply can't throw a curve past some batters, he is going to keep trying. And keep watching as his pitches get knocked into the parking lot. But sooner or later, he will begin to "throw to the batter" and not "to the catcher" and then he becomes a _real_ pitcher. And those batters that can _only_ hit curve balls begin to have real problems since it is very difficult for them to adapt to sliders or whatever... the human GM players haven't gotten to that point yet, although if you watch on ICC, you see a few "new breed" GM players. I watched Mecking rip a well- known program several games (and about 100 Elo points) to pieces the other night. Because he played the right kind of positions. I have watched GM players play Crafty 10 games in a row, finally quitting when they get a draw on the 10th game. Against the computer they are beginning to play very deliberately toward drawish positions because that raises their ratings (since the comps on ICC are usually rated above them). Humans will eventually respond when the challenge is recognized. Right now computers are a novelty in the GM tournaments. I doubt computers will become very commonplace there, which means they will continue to do pretty well vs the humans. Until they invade the human's territory enough that the humans decide to take action.
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.