Author: Will Singleton
Date: 05:29:11 10/07/02
Go up one level in this thread
On October 07, 2002 at 06:45:59, Rolf Tueschen wrote: >As always let me make some short and sharp remarks. > >Computer chess is the simulation of chess. While the opening moves and the last >technical endings can be played with perfection, the simulation is still far >from its optimum in the middle game. Right. I would add that good opening book construction is very difficult, and that most books out there are mediocre to bad. Also, the early endgame is not helped by egtb, and is a huge source of poor play. > >What Vladimir Kramnik has shown with his masterpiece, the second game, is the >unpleasant truth, we should never forget. Machines have no understanding for the >beauties of chess. Either they play with perfection, because the solution is >already there,or they play like a newborn kid. Well, I noticed a couple IM's were saying they couldn't see how Kramnik would make progress, even in the rook endgame. I think this was just before h4. But it's true, machines can look pretty stupid at times. That's the challenge of chess programming. > >The confusing of a training tool with a genuine chess player is the reason for >the speechless amazement of many computer chess lovers. But would they be as >astonished if I would present a "philosopher" with the implementation of the >complete Encyclopedia Britannica and tried to enrol "him" in Harvard or in the >peace conferences at the Lake of Geneva? > >If you are absolutely determined to participate in human chess, although the >mainpart of chess is far from being solved, you must not be surprised if a good >human chess master is reveiling the nature of the whole fantasies from time to >time. > >Because you can fool chess amateurs with the mere superiority of complete >opening dictionaries, you can also fool chess masters from time to time, if they >go for some as-if in the 19th century excursions into the land of combinations, >but you won't be able to always fool the best chess thinkers, or let me better >say chess artists. Because they don't need bad books or certain ideosyncratic >weaknesses of the machines, because they feel and understand the myst of the >imperfect simulation and then sure they have the necessary technique for a >challenge over the whole game, and not only some isolated parts amateurs are >familiar with. > >This is the explanation for the actual situation of computer chess with all the >problems the programmers of the super computer software already had in the 80's >until DB2 in 1997. As I predicted since 1997, the human chess masters have >understood the message of the old trick with the traditional secrecy. Because >without a feeling for the "architecture" of someone's "chess" there is no way to >prove the human superiority in five or eight games. But if you have it, then one >or two games are well sufficient. As Kramnik proved yesterday. > Now, this I'm not so sure about. If you read Kramnik's comments, he indicated he was "shocked" by the tenacity of Fritz's counterplay, and he felt at one point that he was playing for the draw. Fritz saw some things that Kramnik hadn't anticipated, and it made him sweat. So, while the GM has the feel and intuition, the machine isn't "hindered" by this. And so it will play it's own game, which is different, and therefore remains dangerous. Will
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.