Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Smirin vs. Shredder - a question

Author: Roy Eassa

Date: 07:36:32 04/16/02

Go up one level in this thread


On April 16, 2002 at 00:55:45, Vine Smith wrote:

>
>Since no one claims that humans can search 20 moves, or 40 ply, full width, it
>is obvious that 20 move combinations of the sort you refer to in fact are more
>positional than tactical, with entire branches ending with short dismissals like
>"of course not 3.'such and such a move'?? because of the resulting kingside
>weaknesses". But after such combos are more fully analyzed, it often turns out
>that moves like 3.'such and such' are actually good upon further inspection
>because the kingside weaknesses or whatever were not so serious after all. Flaws
>based on positional assumptions seem to arise quite frequently, and to support
>this I quote from the introduction to GM Nunn's "Understanding Chess Move by
>Move":
>"As the power of computers grows, they become more and more adept at finding
>holes in games formerly thought to be sound. My new 950 MHz machine is a monster
>in this respect, and after it had torn many of my intended games to analytical
>shreds I had to go back to my database to look for further examples, only to
>have the process repeated."
>Remember, he was looking only at GM-level games thought to be good. So I think
>it is not impossible to imagine that GMs err tactically quite as often as
>programs err positionally, and that these respective weakness may cancel out.
>Just imagine if computers had somehow preceded humans to the chessboard, and
>demanded perfection according to their standards from the newcomers. Fritz might
>opine "True, these biological units have some interesting ideas, but just look
>how the best of them can blunder pieces, forget openings, overlook mates,
>misplay tablebase positions...how can they ever hope to be our equals?"
>
>Regards,
>Vine Smith


A very well-reasoned set of points you make.  I have no doubt that strong
programs can find tactical errors in most GM-GM games ever played (other than
the short "GM draws").  And GMs would probably find positional errors in most
computer games too.

The only point you make that one might reasonably disagree with is the "cancel
out" statement.  My thinking on that:

In the past 2 years or so, the tactical prowess of PC chess programs has taken
the GM world by storm.  Many GMs have fallen victim.  If you look only at the
results of the past 2 years or so, you might conclude that GMs are doomed.  But
it is VERY early in this game!

IMHO, each individual GM can spend 50 hours or 200 hours (or whatever) and
become a LOT better at adapting his style to avoid the brunt of the computers'
major advantage (plus the GMs can share info in this regard), whereas it will
likely take each major chess programmer many hundreds of hours (at best) to make
a significant change in the way his program handles humans' major advantage.

"Cancel out" implies equality of results.  In the last year or two, the GM-PC
results actually favored the computers (or so I hear).  This year they might be
closer to equal.  In five years a LOT more techniques will be known by a LOT
more GMs than today, plus more young GMs will have spent a larger portion of
their careers co-existing with strong computers.  It seems likely that at least
SOME of these smart and flexible people will put learn and put to increasingly
good use ever-improving anti-computer techniques.

The vast majority of chess players are _already_ way out of their leagues
against computers, and _that_ will not change.  But IMO the best of the best
humans have a LOT of room to improve in their anti-computer techniques and
results -- and with any significant motivation they will do so.  Thus I think
it's way too early to state that NOT ONE human will consistently achieve
positive results against the top PC programs EVER in the future.

(Of course, I personally will definitely continue to be slaughtered mercilessly
by computers!)



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.