Re: computer-go: Two ways to program a GO-engine

[Heikki Levanto <heikki@xxxxxx>]

Nicol N. Schraudolph <nic@xxxxxxxxxxxxxxxxx> wrote:

>>     That means there is no good ALGORITHM for that. But it doesn't mean 
>> there is no WAY to do it. We just have to go beyond today's technology / way 
>> of thinking :)

> Hmmmm, "a way to do something" is a pretty good definition for "algorithm".
> To get past that, you have to invoke one of:

> 	1) P = NP,
> 	2) nondeterministic (quantum) computers, or
> 	3) computers run by magic/god/intuition/psi forces/new physics/etc.
> Looks like a tall order to me.


Again I humbly propose another alternative
  4) Lower our aim to "just" play passable go at some given level (say 5d).


It is well known that the traveling salesman problem is theoretically pretty
hard (sorry I don't know the proper terminology, I am just a programmer).
Yet computers routinely find acceptable solutions to this, helping many
real-life salesmen to save time and miles...

- Heikki

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