Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Today's chess programs are trivial opponents for master players

Author: Dann Corbit

Date: 20:46:40 01/29/99

Go up one level in this thread


On January 29, 1999 at 23:09:04, Timothy J. Frohlick wrote:
[snip]
>The more you say that something can't be done, the more you will have people
>trying to do it. I predict that we will have chess computers similar to stand
>alone units today that can beat Kasparov at 40/2 within the next 25 years.  At
>that point we will give up trying to beat the machines and just bumble along and
>enjoy a nice friendly game with a buddy over a hot cup of coffee.
I agree strongly with your sentiment that whatever people say is impossible is
likely to turn out wrong.  All we can say is we can't get there from here right
now with what we know at the present time.  Personally, I love to try do do
things that "can't be done."  I tried for hours to trisect an angle until I read
a careful proof that showed why it won't work, without making a mark on the
straight edge.  But at that point, I just thought to myself, well, at least we
*can* solve it by marking our instrument.  I once had someone show me an article
that proved you can't solve the equation y=x^x for x, given y.  I had a working
program in just a few minutes.  The mistake the originator of the problem made
is that you can't solve it using only elementary functions.  But you can
numerically arrive at any precision you like by a simple binary search of the
solution space.  With 80 tests you can have Intel 10 byte long double precision.
 I later added a 4th order Schroder's method solution which converges to the
correct answer almost immediately.  A friend named Lee Killough came up with a
method many times faster than the one I thought up.  So if you think something
is not feasible, maybe it's just not feasible yet.

Another thing to think about is that human chess players may experience a
revolution in *their* ability to play.  So we may come up with a gazillion mips
machine that the new SGM's can beat ten at a time blindfolded.  So we cannot
imagine that a machine will always be better, once we build a fast enough one.

>The other alternative is to have chess computer drag races and keep on building
>better and faster machines--lightwave computers cooking in the UV range and
>computing 10 terabytes per second with 100 trillion move opening books and
>similar end game 8 to 9 piece databases.  Of course, it won't prove anything if
>our own abilities turn to mush.
If our abilities turn to mush we won't have the wherewithal to build the
machines.



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.