Re: computer-go: Complexity & SW

[Heikki Levanto <heikki@xxxxxx>]

Fred Hapgood <hapgood@xxxxxxxxxxxxxxxxx> wrote:

> I suspect that if the number of man-years and master involvement 
> invested in go as opposed to chess had been reversed, today
> we would be seeing computer go programs playing at 3 or 4 dan 
> and computer chess software struggling along with USCF ratings of 
> around 1750.

I suspect this would not at all be the case. As far as I can see (which is
not very far), the key to "solving" chess has been that it can be solved
with brute force (or relatively brute quite-much-force). It is of course
tempting to speculate that something as powerful would have been found for
go. But so far we don't even know where to look for such a solution...


> If someone was to offer a a million dollar prize for a program that 
> could play 1 dan (amateur) I'm sure it would be delivered
> in two or three years, regardless of how complex GO is in any
> ultimate sense.

I think the Ing price was a considerable amount of money (something in that
neighbourhod), and it was available for several years. Its requirement was
to beat a junior champion (or Taiwan?), corresponding to (say) 4 dan
amateur. The prize expired this year, but it had been up for quite some
time. I don't think anyone seriously believed it coule be claimed.  (Someone
with all the facts, please fill in and correct my mistakes!)

No, I believe that go is fundamentally more difficult to program than chess.
At least on today's computers and todays computer science... That is why it
is such a challenge!


Regards
	Heikki





-- 
Heikki Levanto     LSD Levanto Software Development   heikki@xxxxxxxxxxxxxxxxx
               "In Murphy we Turst"