Re: Re[2]: computer-go: Pattern matching

["Jeff Massung" <jmassung@xxxxxxxxxxx>]

I am in aggreement that computer systems are very different from the human 
brain, and while they can be emulated, cannot be duplicated with electronics 
(perhaps a biological/chemical computer someday).

A computer program (even cyc) is designed to do ONE thing (ie play go, 
chess, associate sentences with pictures or ideas, pattern match, etc).  
However, no matter what I'm doing, I'm doing many things at once related to 
this.  When I'm making a move in go, not only am I making the best move in 
my mind, but I'm anticipating the next move (and not by the "best line" so 
to speak, but by who my opponent is, how s/he plays, etc).

Even this cyc program, which boasts results such as "ask for a happy person, 
and it retrieves a picture of a man with his baby", will fail when shown a 
picture of a man and his baby and asked, "what do you see in this picture?"  
IMHO, a program like cyc is almost useless (especially when evaluated time 
input vs. output).

The difference between go and chess (I know this has been done many times 
over), is that in chess, there usually is a "correct" move to make, and so 
ply searching is acceptable.  But in go, there is no "correct" move most of 
the time, and the computer would have an impossible time trying to "guess" 
the next move.

Solution: a computer go program must be goal oriented and be able to adapt 
to an individual player by discovering how that player plays (discovering 
the player's goals).  That way, ply searching is possible.  The goals must 
consist of overall goals (win the game, capture territory, capture stones), 
but also subset goals of those goals (capturing the corner helps me with the 
game), and even more subsets (not allowing my opponent to control the corner 
helps me to control the corner).

Well, I know I strayed around alot, but it was fun :)

Jeff

> "There is no inherent difference between the human brain and a computer 
> system
> which prevents knowledge that it [sic] stored in a human brain from being
> emulated by a computer."
>
> This is a _very_ strong assertion.  I think it can easily be disproved.
______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com