Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: About 'understanding' the game

Author: Bruce Moreland

Date: 21:58:38 06/15/01

Go up one level in this thread


On June 15, 2001 at 19:06:17, Gian-Carlo Pascutto wrote:

>Jose Carlos wrote:
>
>>Everytime this kind of argument arises, I have the same impression: There's a
>>problem of definition. What do we call 'GM strength'?
>>
>> A. If we speak of 'quality' of chess (whatever this can mean), most chess
>>    players will probably agree to one of these possibilities:
>
>> A.1. Computers are not GM strength because they show lack of understanding
>>    too many times to be considered GM's.
>
>This is a tricky call. Who defines 'understanding' a game?
>
>This is a subjective measure. I can think of two examples to illustrate this:
>
>a) on Tim Krabbe's pages he sometimes has the topic 'computers can't play chess'
>and demonstrates positions where the computer does really awful things (in his
>eyes)
>So 'computers can't play chess'.
>
>On the other hand in his analysis he sometimes refers to a move found by the
>computer. Often this is a good move the human would have a lot of trouble
>finding. 'Humans can't play chess?'
>
>b) my own program plays several variants which it has very little understanding
>of. In one variant is just picks the move that offers it the most options not
>to get mated. In another the only heuristic it has is 'put pieces near the
>opponents king'. In a way it does not understand the game at all, by human
>standards. Yet it is at the same level of top human players. It makes awfull
>moves by human standards. Yet it often wins with those moves.
>
>> A.2. Computers and humans do not compare (like the soldier and the tank).
>
>I think this holds a lot of truth.
>
>> B. If we speak of 'quantities' (namely ELO rating), we can certainly compare
>>    humans and programs, and say _with numbers_ if programs perform like GM's
>>    or not.
>
>This assumes you can express GM in term of an ELO rating. I do not think
>that is possible, or at least has any meaning.

It's a way of comparing the relative strengths and weaknesses.  Humans might say
that computers suck because they don't know that a certain locked position is
drawn, and computers might say that humans suck because they overlook simple
mates in 12, but the acid test is a bunch of real games.  When you play games
you figure out how much these differences matter in real games.

This works for me, I think.

bruce



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.