Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SW vs HW in Computer Chess?

Author: Knut Bjørnar Wålberg

Date: 13:06:27 10/27/02

Go up one level in this thread



>I am curious what this forum thinks with regards to the following:
>
>1.  How strong is strong enough for Computer Chess?  3000 ELO, 3200 ELO,
>stronger?
>
That would depend on the strengths/weaknesses of the programs. Somewhere I've
seen the claim that the best computer chess programs play ~3000 ELO in open,
tactical positions, but only ~2400 at closed, strategic positions, giving them
an overall rating around 2700. I don't know if this is true, I'm much too weak
in chess to say for sure. However, if it is true, then it's not enough to say
that we want a 3000 ELO chess program, if that means it will be tactically at
3500, but positionally at 2500.

>2.  Do we really need more speed (faster HW, no doubt this makes program x
>stronger) or is sw improvements (on a 3Ghz machine) enough to make most people
>here happy?  Which do you want to pay for HW, SW or both?
>
>For me, 400 ELO above the strongest GM would be plenty.  HW vs SW makes no
>difference to me, I will buy SW and HW upgrades for a while to come, but the HW
>upgrades are harder and harder for me to justify, the SW is easier for me to
>swollow. :)

I think the greatest challenge is in SW. The chess sites I've visited, seem to
have the following view on analysis: A computer can be a great aid, but there's
no substitution for a GM if you really want understand chess. I guess that's
because there's a lot of things chess programs don't understand at the moment.
So the point where it's *enough*, at least for any practical use, is when chess
programs are at least as good as the best super-GMs in all aspects of the game,
not just tactics.

Which actually brings up a question: Is it possible to make a computer program
that is (much) better strategically than a human? I'm wondering because to me it
seems like:
strategic understanding means grasping a lot of concepts
=> better players know more concepts and how to utilize them
=> the best players know and use almost all the concepts
=> a strategic super-computer should know even more

but how can you implement a piece of knowledge that no human is aware of? I
guess with *real* AI a computer could bring completely new knowledge to chess.

But back to SW/HW. Obviously, if a lot more chess knowledge into a program,
you'd probably want faster HW as well, so you can still keep up the ply depth.

Knut Bjørnar Wålberg



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.