Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Computer chess and processor speed in 20 years from now...

Author: GuyHaworth

Date: 05:06:04 02/26/03

Go up one level in this thread



Moore recently opined that his Moore's Law holds until 2017, and that the
industry's challenge is to push 2017 out to later.  I have no doubt they will.

If the MHz of standard processors doubles every N (say 18) months, you can work
out where that takes you in 2020 or 2030.

Translating this into ELO advantage in (at least) C-C play, I believe that 3x
processor-power = 1 more ply in the search, which is worth some 50 (and
decreasing) ELO extra.  For this, see Heinz' latest self-play results going up
to 12-v-11 ply-search.  Looks like 10 years gives you about 4 plies extra.

To that, you can add the contribution of improvements in:

a)  parallel-processor platforms: SMP, clustered, and share-nothing
b)  memory technology
c)  ancillary technologies like FPGAs
d)  software for opening-book development
e)  search/select/evaluate algorithms
f)  endgame knowledge:  maybe complete 7-man by end-2013.

So, by the end of 2013, you might be looking at engines 200 ELO better than
today.

However, I hope that this power will be used to create chess-engines that
'explain and instruct' - which implies that they will do so in human terms.  I
see the DJ team stating similar objectives, i.e., that 'beating humans' is not
the ultimate life-enhancing objective.

g







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.