Re: computer-go: Complexity & SW

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

> This is how both go and chess programming works: finding a better 
> set of patterns than those driving the last rev. I see nothing about 
> chess to lead me to think that the task of building a machine good 
> enough to beat a chess master is any easier than building one to 
> beat a 1 dan amateur.  (I would agree that a chess program could 
> conceivably win a game or two against a rank beginner using only 1 
> or 2 ply searches, which a go program could not, but that isn't
> relevant to anything.) 

Human chess  is all about patterns.  But  right now computer chess has
almost nothing to  do  with  patterns.   In some cases   very specific
konwledge is added  to   a chess program  that  might  be  likened  to
patterns, or even be properly called a pattern, but this does not form
the heart and soul of a chess program.

I know  that as a   chess player I   very often see  patterns, typical
setups that  I've seen a million times  before and I know exactly what
kinds of moves to expect  and to execute (and  to ignore).  Even  when
patterns are applied to chess programs they are done in the evaluation
(at end nodes), not to suggest moves and plans selectively.

I'm  sorry to keep disagreeing with   you, but I  disagree about there
being nothing inherent in  Go that makes it more  difficult to build a
machine to  play  at  the top   level.  I also    don't think that  Go
programming is that far behind  Chess  programming, when you  consider
that Go has probably had more man hours poured  into it (to date) that
Chess did when the basics of  writing "good" programs were discovered.
This all happened with Chess  before most  people  has easy access  to
computers.   So when Chess had the  same level of man hours investment
in it that Go has today, we were  probably already seeing master level
programs.  And this was with lower quality tools (compilers, debuggers
and easy to use fast systems) than we have today.



Don