Author: James Swafford
Date: 13:06:41 06/10/99
Go up one level in this thread
On June 10, 1999 at 15:43:16, Dann Corbit wrote: >On June 10, 1999 at 15:24:02, Michael Fuhrmann wrote: > >>Can today's best chess engines become significantly more powerful through >>software improvements alone, or have we reached the point where hardware >>improvements are necessary to make that next BIG step? >I don't think this kind of question can be answered reliably. You can't prove >the negative: >"No one will ever invent a revolutionary new algorithm for chess calculations" >It may or may not be invented. > >Who knows if even Hsu's chips are the ultimate in hardware technology. Perhaps >a billion neuron chip will be written for AI which can be combined with standard >CPU technology *and* Hsu's chips. Who knows. Trying to project the future is >not very accurate where man is concerned. > >IMO-YMMV. Well said. Even from a patzer's standpoint, it is evident to me that the typical evaluation function is, at best, crude. I believe the amount of information we examine in our evaluations is somewhat limited by the speed of our hardware. (IE it is important to maintain tactical ability). As hardware gets faster, we can either: 1) Do nothing to our evals, but hopefully search deeper, improving tactical ability 2) Evaluate more, but maintain the same tactical ability 3) Some hybrid of 1 and 2 Conversely, I think that as we search faster and deeper, some terms may become obsolete. (Such as bishop traps on A2/A7/H2/H7) If you really think about it, all positional terms would be obsolete if we knew the solution to chess. Interesting times, don't you think? I think that most improvement will come in the way of smarter programs, but we need more horsepower to do it. -- James
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.