Author: Robert Hyatt
Date: 21:00:29 04/22/98
Go up one level in this thread
On April 22, 1998 at 18:46:08, Jan-Frode Myklebust wrote: >On April 22, 1998 at 17:18:13, Bruce Moreland wrote: > >> >>On April 22, 1998 at 15:16:28, Jan-Frode Myklebust wrote: >> >>>And a micro might learn that Deep Blue doesn't learn, and play the same >>>game over and over, getting GREAT results against it.. What a stupid >>>computer, needs human tuning before each match :) >> >>If my whole deal would have been to play a match with Kasparov at the >>rate of one game every day or two, I wouldn't have wasted my time with >>machine learning either. >> >>Different problems have different solutions, you can't pick on people >>for picking an implementation that doesn't solve a problem that they >>didn't feel they needed to solve. > >So then Deep Blue isn't that good at chess in general, only at beating >G. Kasparov under special conditions...? At least it can never become >world champion without either learning to avoid playing the same loosing >line again and again, or solve the game... > >Jan-Frode This is an old argument, and one with one fatal flaw. DB does "learn" if you define "learn" as not playing a gross opening twice. Cray Blitz also "learned". And it didn't have one line of code to do so. It "learned" by having me or Bert Gower modify one character in the book file after a bad opening... and it would never play that again... In that sense, they learn like programs have been learning for 30+ years. "automatic learning" is something else... but with a human team working on the program, no one is going to beat it the same way over and over...
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.