Author: Vasik Rajlich
Date: 05:47:03 01/29/06
Go up one level in this thread
On January 28, 2006 at 00:29:41, Andrew Wagner wrote: >The other day, a link was posted to a fascinating video about the history and >future of computer chess (see >http://www.talkchess.com/forums/1/message.html?481541). Lots of very interesting >points were made, particularly about comparing how computers play chess to how >humans play chess. So, this got me thinking. We know that computers have gotten >to the point where they can search millions of nodes per second. Humans, of >course, typically only search a few dozen positions. And yet, while top >computers have beaten top GMs, they're certainly not far above them, if at all. >Why is this? If we could get computers to vastly reduce the number of positions >it looks at, would it play better or worse (assuming it looked at the right >positions)? It seems to me it would certainly search deeper. Can something like >this be done with alpha-beta, or have we reached the pinnacle of what an >alphabeta searcher can do? Is hardware the key to better-playing machines, or is >there a long way we can go yet with improving artificial intelligence? Some very >interesting things to ponder. Just one thing to add: humans are also doing a lot of searching. You might find that at the "machine code" level, human search is rather primitive - it's just that we haven't figure out how to reverse engineer it yet. Vas
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.