Author: William H Rogers
Date: 06:38:11 05/25/99
Go up one level in this thread
On May 24, 1999 at 19:50:06, Francis Monkman wrote: > >On May 24, 1999 at 19:06:05, Will Singleton wrote: > >>I think it's extremely difficult for a computer to do GM-type positional >>assessment. The human brain seems better setup to do that type of >>pattern-matching, selective recall, etc. At Deep-Blue type depths however, >>short-term positional understanding becomes less important, since small (or >>large) tactical maneuvers can be found, resulting in either material gain or a >>simple positional betterment. > >It's also worth mentioning that chess strategy in the 20th c. has focussed more >on the dynamics of chess, hidden potential, even 'resonant' and 'unresonant' >positions. Watching parallel searches, one can often deduce more about the >dynamics of a position from the relative behaviour of the lines, than from the >actual move chosen. Current searches obviously reveal information about the >dynamics of the position, but it seems to be wasted. >> >>At a certain depth, the GM can't compete. We haven't reached that yet, but >>theortically it's true. So, yes, I'd say that due to the inherent difficulties >>of the linear computer-model, the only way to beat a GM in the future is to go >>deeper, bean-counting style. > >I've just been playing through several games by Mihai Suba (author of the >*excellent* "Dynamic Chess Strategy"), and I've just been following a forced >line 19-ply deep from start of attack to resignation. And even then viewers were >surprised, as a clear advantage wasn't seen for another 8-ply or so. In terms >of sheer depth, searches have a long way to go! (I sometimes wonder, in >*clearly forced* sequences, why programmers don't just go straight to the >'end of the line', and start the search from there...) > Once a read an article written by a noted child Physicologist about the way children learn. i.e. the way to grandma's house. He noted that they first recognized grandma's house and as they made more trips to her house they learned to recoginzed more landmarks leading up to her house, until they finally learned the way to her house from their own home. The point was, they learned the way in reverse, starting with the end goal and working backwards. This might be another avenue to chess programming, at least in improving end games, then middle games. It's food for thought. Bill
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.