Author: Robert Hyatt
Date: 18:40:13 06/12/99
Go up one level in this thread
On June 12, 1999 at 16:03:53, Fernando Villegas wrote: >Hi Bob: >You are right: evolutive improvements over time are the way things has happened >until now in this field, BUT by the same reason I smell the posibility of an >uncoming breaktrought. I mean, if you take a look at every technology and >science, always great breaktroughts are preceded by a long, sustained period of >time where the main issues of a field seems to be solved and just and only >improving. It was so with steam locomotives -that reached a great level of >perfection-, it was so with pre-einstein physics, it was so with every >machinery, technique and body of ideas that have ever been created. In fact is >that same evolution that prepares the road to breaktrought as much the evolution >develops all chances hidden in a previous revolutionary idea. >Course, we have the issue about how much is breaktrought. In that and in >relation with chess computers, you and other programmers here have the word, not >lay as me, but then i would like to ask you which you believe could be a >possible breaktrought.. Maybe pattern recognition of key positions as GM do? >Greetings from Chile >fernando This is an interesting issue. For me, a 'breakthrough' is not needed, because our present approach has shown that it _can_ beat GM players. We are now almost unbeatable at blitz games, very difficult to beat at game/30 games, and are getting better at 40/2hr. Since I am interesting in beating humans, I don't see any need to become 'selective'. IE I see no evidence that says "before you can beat the best humans, your program has to "think" like the best humans and only look at a few moves per position." I believe that the current approach will evolve into something humans can only rarely hope to beat... It seems obvious that a computer _can_ become better in a revolutionary way, bu examining much smaller trees. IE if we could reduce the branching factor to < 2, computers would ahve a chance to effectively 'solve' the game at today's speeds. But such an approach is difficult. And it is a really long way off from becoming a reality because no one is working on it in a serious way any longer, since the full-width approach has shown itself to be so effective...
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.