On 29, Jan 2007, at 9:17 PM, Matt Gokey wrote:
I wrote:
> In computer-go where there are so many wildly different techniques
> being used, some scalable to some degree or another and some
not, it
> doesn't make sense to make generalizations. Whether a specific
> program's scalability results in any improvements (linear or
otherwise)
> with time-doubling depends entirely on the algorithms and
techniques
> in use.
Of course we all know many computer-go programs don't scale
extremely well. MC Go (and variants) scale fairly well but
probably need a lot more knowledge and probably other methods built
into them to become good as MoGo is showing. But this doesn't
really say anything about human play. In fact, since Go would not
succumb to standard game-search techniques, most computer-go
programs used fairly simplistic models, pattern matching, combined
with some reading - roughly attempting to emulate certain aspects
of human play.
So I thought of another way to express some of what I was thinking.
Humans play kind of like GNU Go only lots better and can think
beyond what they've learned and learn as they play. GNU Go can't
get dramatically better by thinking longer, only modestly better.
Instead GNU Go must learn (i.e. be programmed with knowledge/
techniques - #2 and #3) to get that much better (no offense to the
GNU Go team).
Matt, I think your statements, particularly the ones above, are the
most accurate. Not that Don is wrong, just that I think he is over-
focused upon methods that do scale better.
First, as a human player I very definitely feel that I can improve a
few specific moves somewhat with perhaps one doubling of my time,
but after that my mind just fills up and I cannot get any better by
myself, even if I recognize and understand immediately when a
stronger player shows me a better move. But on my own, after a few
minutes I just am not going to find anything new.
Second, with respect to computer algorithms (and I know that the
primary thrust of your argument was NOT directed towards computer Go)
I think that SlugGo shows rather well that some algorithms do not
scale well. SlugGo gives GNU Go about 72 times as much thinking, and
while it could be argued that some of our heuristics and evaluation
functions sometimes lead us to make worse moves, in a statistical
sense when SlugGo decides to make a move that GNU Go considered to
have a lower value, it is often correct that it is a better move
(even if rarely the "best" from the view of a good human). SlugGo is
at best 2 stones stronger than GNU Go against a third opponent. Don's
"curve" does hold much better when SlugGo plays GNU Go, where we have
seen that we can get SlugGo to beat GNU Go with 7 handicap stones
with enough lookahead time.
Cheers,
David
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/
