Author: Mark Young
Date: 12:48:17 06/25/99
Go up one level in this thread
On June 25, 1999 at 15:34:37, greg moller wrote: >On June 25, 1999 at 15:18:26, Mark Young wrote: > >>On June 25, 1999 at 14:40:37, Bruce Moreland wrote: >> >>> >>>On June 25, 1999 at 09:17:46, Robert Hyatt wrote: >>> >>>>This is the point I have tried to explain to you several times. What "GM" >>>>would lose such a game to a 2200 player? Can you think of one? I see _all_ >>>>programs lose to such players on ICC. Some repeatedly to humans that are >>>>not even 'titled' by FIDE. And so long as this continues, I have a _hard_ >>>>time thinking of the programs as "GM" players... >>> >>>There is a point to this but it's important to avoid taking it too far. >>>Computer strengths and weaknesses will not match perfectly with human strengths >>>and weaknesses no matter how strong the computers get. >>> >>>It's like declaring that until computers can solve that famous locked pawn >>>problem where the game is a draw despite huge extra material (the weaker side >>>simply has to avoid taking a rook, and then hide behind the pawns), that they >>>can't be masters. Clearly they are masters of some variety despite being less >>>than instantly able to solve that problem. >>> >>>They are very strong in some areas and weaker in others, and it might not be >>>true that they have to be strong in *all* areas to succeed. It is a matter of >>>determining the degree to which their weaknesses can be exploited by humans. >>>It's getting harder for the humans all the time. >>> >>>bruce >> >>This is why it is important to trust the results of games played vs. humans be >>it a GM or not, or a win or a loss. You can because to subjective and start >>dismissing and justifying away all successes of the chess programs if you only >>key in on the programs weaknesses. We all know the weaknesses of computer >>programs, but if the weaknesses are or are not counter balanced by the programs >>strength it will show in the results. It is ironic that humans try to justify >>their failures, but for computer programs we try to justify their successes. > >It's still possible that the top programs are GM-strength DESPITE these >weaknesses, which would only mean that the strengths are far beyond what GMs can >do, but that's a known fact already, or should be. > >More games against various human styles and levels of strength are required to >ascertain the true strength of the programs. Such games are sorely needed. >Without them all we've got is idle speculation, more or less. I will always take data over speculation, and it is important to take care in lumping all chess programs into one basket, or mixing results of different time periods. 6 month to 1 year is a life time to a chess program because of updates and better hardware. And as we know not all chess programs play as well as others when playing humans and/or other chess programs. > >regards, >gm
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.