On Wed, Jun 2, 2010 at 11:49 AM, Don Dailey <[email protected]> wrote:
> How can I dispute it,  you do not define terms?   Break it down,   when you
> say "Go is much harder than Chess" what does that even mean?      13 is much
> harder than 27.    What does that supposed to mean?     Dogs are harder than
> Cats.      If it means it's harder to write a go program, then I completely
> dispute it - I can write a go program much faster than a chess program.
>  Both program will play very well if I define "very well" to be random play.
>   My point is that hard or easy is an arbitrary concept and I don't believe
> there is any reason to believe that how humans play is some kind of sacred
> yardstick.    If you make that the yardstick, then GO comes up short.

I agree that people are often lacking in defining precise terms. But
it's also possible to be unreasonably dense by ignoring some basic
assumptions that many, if not most, take for granted. Doing so
deliberately doesn't help the discussion of course.

Would you agree with a slightly more precise statement: Go is harder
to program to match a human expert than chess. It almost doesn't
matter how you define expert. But let's say something like the top 1%
of players or the top 1000 players of the world. As long as you use
the same criteria for chess experts and Go experts. I have yet to hear
about a better yardstick, either from you or anybody else. Not
surprisingly, because apart from computers and humans I don't know of
anything else that we could use. And to compare computers to computers
is obviously meaningless in any absolute measuring sense. So measuring
humans versus computers is the logical choice to start with. And
possibly the only one until you have some other point of reference,
like a perfect player. But only if you can prove that progress is
smooth and not too jumpy. For example: if you have a perfect player
and the 'almost perfect' player A can never win more than 10% against
the perfect player without being perfect itself. But maybe you can
have another 'almost perfect player' B that also wins 10% (or less)
against a perfect player but wins 70% against A. We already see this
happening where difference in level of a group of different MCTS
programs does not translate to the same difference against humans, or
even a different set of of programs.

Maybe some day computers will learn to program. And who knows they'll
find it easier to program Go than chess. Somehow I'm doubtful, but
until we have more evidence about this it's fair to use the evidence
we currently have at hand.

Mark
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to