Author: Uri Blass
Date: 16:02:24 12/27/01
Go up one level in this thread
On December 27, 2001 at 17:37:22, Dann Corbit wrote: >On December 27, 2001 at 17:29:34, Louis Fagliano wrote: > >>On December 27, 2001 at 17:13:25, Dann Corbit wrote: >> >>>On December 27, 2001 at 16:14:22, Will Bundy wrote: >>> >>>> >>>> Does anyone have any opinions on Which programs play the Most like a Human >>>>Grandmaster in terms of strategy and Tactics? I am aware that Computers do not >>>>have the Understanding of Chess that Grandmasters have, but what programs are >>>>able to simulate Grandmaster Play the Most? >>> >>>None of them. The play is very different. A computer program searches with >>>almost flawless precision to find tiny fractions of a pawn advantage and any >>>tactical shot like grabbing a pawn. >>> >>>Human GM's don't play like that for the most part. Polgar and Morphy are >>>tactical sorts, but most GM's have a very different style from computers. >>> >>>CS-Tal plays wild "pirates pouring over the walls" attacks. >>>Gambit Tiger apparently is made to understand gambits to some degree. >>> >>>But really, you can't compare the play very well. >>> >>>Both machines and top human experts play exceptional chess games. But the way >>>that they go from "point a" to "point b" is not the same at all. >>> >>>IMO-YMMV. >> >>I'm not sure if I or Dann Corbit understood Will Bundy's question right. When >>Will Bundy said "plays closest to a human grandmaster" what does "plays" mean? >> >>How a computer searches for a move compared to a human GM or how similar the >>move played (by whatever heuristics) is compared to what a human GM would play >>in the same position. >> >>Dann Corbit interpreted the question as how a human GM searches compared to a >>computer searching which is quite obviously very different. >> >>But I think Will Bundy's question was not differences in search techniques >>between human GM's a nd computers but the final result. In other words, if you >>were given some game scores of Kramnik's, Kasparov's, Anand's, Fritz 7's, Chess >>Tiger 14.0's, and Junior 7's games (against other humans and this is important) >>but were not told which was which, could you tell just by playing over the games >>which were human GM vs. human GM games and which were computer vs. human GM >>games and if so, which program would simulate the play of a human GM the >>closest? > >None of them. > >All of the good programs will definitely find tactical shots that the GM's >missed. They would make these moves instead. > >Most of the good programs will be very much fooled by gambits. > >Most of the good programs will be greatly puzzled by positional moves. > >Try LCT II's tough problems, NOLOT, etc for a taste of this. > >Try analyzing this game with a computer and tell me which one actually finds all >the optimal moves: >ftp://cap.connx.com/pub/game-of-the-week/strategy.pgn It is clear that there are cases when you can be almost sure which side is the computer from looking at the game(in cases that one side made a tactical mistake that is typical to humans or a positional mistakes that is typical to computers but not all the games have tactical mistakes or big positional mistakes that humans avoid and there are cases when you cannot know based on a game without analyzing it with the right program which side is the computer. There is a meaning to the question which program can create the illusion that it is a strong human in more cases. 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.