Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Go programming

Author: Dann Corbit

Date: 14:26:49 02/13/04

Go up one level in this thread


On February 13, 2004 at 08:26:57, Dr. Oliver Brausch wrote:

>On February 12, 2004 at 15:21:58, Dann Corbit wrote:
>
>>>Some people say that no Go program will be stronger than the best humans for at
>>>least another century, if ever!  Today's best Go programs are far weaker than
>>>any Go professional.
>>
>>Hogwash.  Computers will have more compute power than the human brain long
>>before a century passes.
>>
>
>I am not so sure about this, Dann. I have not played Go a lot, but it is
>really a very stratetical game. Chess is both stratetical and tactical, which
>is the fascinating point of chess. Computers are strong at tactics but still
>very, very weak at stratetics.
>
>And what will be in 100 years? Computer 10^6 times faster than now, even
>perhaps 10^9??? yes, but it will be a finite number. So we got some plies deeper
>on Go, some more plies on chess.

Since compute power doubles in 1 year right now (up from 18 months -- it appears
to be superexponential) in 100 years, computers will be 2^100 times faster
(roughly).  Lets say that they only make 2^50 (so that Moore's law breaks down
very badly).

That would be roughly 1,125,899,906,842,624
or one quintillion times faster.

>But the human brain with stratetic power thinks even much deeper than that
>will be. They have a feeling for the position, for the pattern, what computers
>are missing. Perhaps some day there will be Artificial Integences with
>that skill, but I am not sure if they can be called Computers then.

I expect the algorithms to also improve dramatically.  Go programming is in its
infancy compared to chess programming.  People have been writing chess eval
functions since the 1960's.  After 40 years of go programs, they will get a lot
smarter.

>Intersting point to that:
>It was almost 10 years ago, when the best Chess computers started
>to beat GMs. So all people predicted that in 10 years computers will
>be unbeatable by human. And what is now? Computers 100 times faster still
>are beaten by humans.
>And I predict something: In 10 years they still will be beaten by
>humans! (not by me, btw...)

The people can also benefit from things that the computers learn.  But in 100
years, the humans will have no chance at chess, even in correspondence games.

>I think there is a bifurcation: Or you solve a game (like checkers), then
>the computer is unbeatable or the computer will never be unbeatable.

Of course, Moore's law might give out entirely.  But I doubt that will ever
happen within the next few centuries.  They have already discovered new
technologies that will enable enormous bounds in power.

So I think that in 100 years compters will at least be as good as the best Go
players on earth, and maybe a lot better.



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.