Author: Robert Hyatt
Date: 19:01:26 08/18/02
Go up one level in this thread
On August 18, 2002 at 10:54:25, Uri Blass wrote: >On August 18, 2002 at 10:38:19, Robert Hyatt wrote: > >>On August 18, 2002 at 09:06:02, Jorge Pichard wrote: >> >>> Kasparov proved that he can defeat programs at fast time controls when he >>>defeated Deep Thought in a game/90 two games match in 1989. This program was >>>weaker than Deep Junior is today, >> >>How would you figure that? Deep Thought produced a 2650 rating over 25 >>consecutive 40/2hr games against GM players. Anybody else done that yet? >>Not that I have seen. > >If we look at the games we can find tactical mistakes of deep thought that >the top programs of today have no problem to avoid. Maybe or maybe not. I ran a test a few years ago, after Chris Whittington made a comment about how poorly Cray Blitz played in the 1986 WCCC in the game against Bobby. One particular move he criticized I found that every commercial program we could try played the _same_ move. It was an ugly move, but the only way to save a pawn. I then decided to have Crafty annotate all 5 rounds played by Cray Blitz, and I was quite amazed to discover that Crafty found _no_ tactical errors of any kind that it could see in the same time-frame as the 1986 CB, even though Crafty was significantly faster than the 1986 CB (we were doing 400K nps or so then and crafty on my quad 400 was much faster than that.) I don't think the programs back then were that tactically weak. And I could easily point out lots of positional mistakes that _todays_ programs will make that programs from back then might not make. Even though that doesn't prove a single thing... > >One example is that deep thought let judit polgar to draw because deep thought >did not detect repetitions in the last plies and there are more examples. That was a known problem. And yes, it affected a game here and there. But not very often, and it was fixed in DB2. > >I do not know about impressive moves of deep thought from games that the top >programs of today have problems to find. > >I also believe that humans today know better than they knew at the time of deep >thought so comparing rating is not convincing. Depends on the human. IE IM Mike Valvo knew (and still knows) as much about anti-computer chess as anyone on the planet. I can think of others as well, Kaplan being one. > >It is more interesting to compare moves and I know of no impressive move from >deep thought from games(there are moves that the programs of 1990-1995 on the >hardware of 1990-1995 could not find but the programs of today on the hardware >of today are clearly better). > >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.