Re: [computer-go] Pattern matching - example play

["Frank de Groot" <frank@xxxxxxxxxxxxx>]

From: "Don Dailey" <drd@xxxxxxxxxxxxxxxxx>
Subject: Re: [computer-go] Pattern matching - example play


> That's kind  of why Franks approach  has some appeal  to me.  Learning
> from human  games are  probably just fine  for now, but  ultimately it
> seems like it has to be very limiting.



When the learning system works well, it is a virtual guarantee that the
computer will start out a few stones stronger than the strongest human Go
player, and from then on the gap only widens (due to the fact that humans'
brains are not subjected to Moore's Law).

A learning system is a hybrid between learned knowledge based on
programmer-taught concepts PLUS search.

Basically, the evaluation function is based on learning, not a programmer
messing around.

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