[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: computer-go: perfect players



Nicol Schraudolph <schraudo@xxxxxxxxxxxxxxxxx> writes

>Hmmm, I actually believe the opposite: to me Go seems to tie much better
>into the human "hardware" (mostly: our visual system) than chess, and I
>see this as one reason why humans can play 19x19 Go instead of reaching
>their limits at 9x9, whose complexity might be closer to that of chess.
>I therefore expect building a computer system that can beat a human in Go
>to be much harder than in chess.  At the opposite end of this spectrum
>lie games like othello/reversi: the cumulative color reversals there
>are really hard for humans to keep track of (visualize), so computers
>have an easy time beating us at that game.

I agree.

Computers are not particularly bad at Go, people are particularly good
at it.  This is because people can visualise Go sequences easily, as the
pieces don't move around much.  Chess has pieces that move repeatedly,
Othello has pieces that change state repeatedly, Wari has most of the
board changing with one move.  All these changes are no problem for a
computer, but an obstacle to humans.

The aspect of Go in which computers compare best with people is in
solving ishi-no-shita problems.

Nick
-- 
Nick Wedd    nick@xxxxxxxxxxxxxxxxx