Author: Robert Hyatt
Date: 14:03:46 07/15/98
Go up one level in this thread
On July 15, 1998 at 08:58:16, Dan Homan wrote: >On July 15, 1998 at 08:13:21, Shaun Graham wrote: > >> >> >>Well if you posit that fritz is just say 2450, well a computer 2450 is really >>more consistent than a human 2450 wouldn't you think,( it never forgets it's >>opening, doesn't get sick,isn't late, not in time trouble)? This greater >>consistency is what makes up for a higher rating in my oppinion(i admit that >>some test on that idea might need to be done) >> >> > >I think you hit the main point here, but not in the way that you think. >A computer is very consistent. So much so that it is a sitting duck. Yes, >Fritz plays very strong in many positions and many types of games, but it >does not play very strong in all. As Bob pointed out, it plays rather weakly >in some positions and types of games. The problem is that Fritz will >*always* play weak in those positions (unless the programmers change it). >Therefore it is a sitting duck and an easy target for humans who learn what >those types of positions/games are. > >A GM computer must be strong in all types of positions/games *or* must be >able to adapt (learn) to avoid positions where it is weak. Without one of >these abilities (Fritz has neither), no computer can play at the GM level >for a protracted period of time. The reason is the simple one that you >stated... A computer is predictably consistent and humans will learn its >weaknesses. In the long run this will lead to a program that loses most >of the time to IM and GM strength players. > >Note here that I am talking about *protracted* GM level performance, not >simply showing up at a couple of tourneys where the opponents are unfamiliar >with the program's play. > > - Dan I think computers are less consistent than you'd think. Tactically yes, they are like pushing a house. But in the wrong kinds of endgames they just roll over and play dead. Or in the wrong kinds of blocked positions. That is the thing that a human GM really doesn't do. The kind of "consistency" computers have is good for tactics, but bad for playing GM's because as you said, they are "stubborn" and don't get better. They will make the same mistake over and over and over given the opportunity. And a GM will give that opportunity. :)
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.