Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Go programming

Author: Janosch Zwerensky

Date: 12:17:21 02/13/04

Go up one level in this thread



>>GnuGo is said to be not much weaker than the top commercials (the strongest of
>>which is Go++), but a weakly 4-kyu amateur like myself (...)
>Janosch, thanks! It's great to have somebody here in a computer chess discussion
>who's also so strong at Go.

Thanks, but 4th kyu isn't really strong yet :).

>
>When I started playing Go (after Christmas), I was told that my decades of chess
>experience would work to my detriment.  But I don't think that was the case.

Experience at chess is certainly going to be an advantage relative to other Go
beginners rather than a disadvantage. If it does nothing else, knowing chess
will prepare one for the general type of challenges any deep game poses.

>(...)
>What do you forsee for the 5- and 20-year timeframes?

I've no idea. Programs are certainly getting stronger and will continue to do
so, and since there are no miracles happening in the human brain, I'm certain
that it is possible in principle to build a machine that plays Go well.
Personally, I'd say I'm quite optimistic that it will be doable one day also in
practice, but whether it will happen in the next 20 years, I have no way of
knowing.

>  Will vastly faster CPUs
>make a big difference if the current algorithms continue to be used?

I don't think so. Faster CPU's alone would certainly help improve tactical
reading a lot, and might open the door to shallow full-board reading if this is
desired, and such things might gain programs a few stones. However, even if
running a thousand times faster would allow for simple software changes leading
to, say, three stones of a gain in playing strength, programs like those we have
now would still be 4 kyu at best, and certainly weaker when playing people who
know their weaknesses.
Of course, faster computers will never hurt or hinder the development of better
software :).

> Might a
>completely different approach (I have one idea on the back burner) provide a
>real breakthrough?

It is my understanding that something like this has happened in computer
backgammon, so it might not be impossible to happen in Go. Trying new ideas will
also certainly not hurt but help the development of computer go :).

>  What do you think is really going on inside the mind of a
>9-dan pro?

I've no idea. I actually don't look at pro games, because these guys have a
habit of playing moves I don't understand, which tends to frustrate me. For this
reason, I prefer to go through games of players from the lower and intermediate
amateur dan ranks, because the proportion of plays which I wouldn't have seen
but which I still can recognize as being simple, beautiful and strong is
perceivably higher there :).

Regards,
Janosch



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.