Re: [computer-go] Modern brute force search

[Don Dailey <drd@xxxxxxx>]

> A professional checkersplayer such as Tinsley could play the game near
> perfect.

Yes, Tinsley was on the verge of perfect play, but not quite.  Chinook
may now be perfect.  But there are probably no humans still alive that
can play like Tinsley did, he was the exception.

But it's amazing how difficult  this very "simple" game is.  It's just
barely simple enough that a human  can play very close to perfect, but
not just any human.    That's why I start with checkers as a jump off
point for what can be done.   

Before  Chinook had the  big databases,  it reported  these incredible
deep search depths,  but still would make blunders  that Tinsley could
take advantage of.

Here is  the interesting point of  what I am  saying: Those incredible
search depths that Chinook could do, which could lose to Tinsley, were
many ply less than chess programs do right now!  And yet current chess
programs  are very  close  to the  playing  ability of  the very  best
players!   

Does everyone  see my point?  That's why  I say there must  still be a
very long way to go for  computer chess, and I extrapolate to GO.  

I know it's  possible that GO is actually a  simpler game (for humans)
than Chess  is.  Maybe it's actually  easier for computers  too and we
just haven't  figured this out yet.   Maybe there is  some strange and
abstract  representation of  GO  that  makes it  the  game easier  for
computers yet to be discovered.   Nah!

- Don


_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/