Re: [gnugo-devel] endgame module for GNU Go

[Martin Girard]

On Sep 6, 2004, at 18:22, Eric wrote:

> --- Paul Pogonyshev <address@hidden> wrote:
>
> > Eric wrote:
> >
> > > --- Paul Pogonyshev <address@hidden> wrote:
> > >
> > > > * I don't think it is worth the inconvenience of
> > > > plugging in a planner into GNU Go.
> > >
> > > Yea, unless it makes GNU Go useful for something
> > > besides your own personal entertainment.
> >
> > GNU Go is meant to be "useful" in terms of playing
> > Go.
> > It can be used for research (or do you mean any else
> > use?), but that's not something "common" GNU Go
> > users
> > are interested in.
>
> Okay Paul, please allow me to be more blunt. I have
> absolutely no interest in making GNU Go "stronger". My
> aim is to make GNU Go "work".
>
> By "work", what I mean is that I want a champion,
> human Go player to value and enjoy playing GNU Go, and
> not just have it as a curiosity for novices.
>
> As an artificial intelligence engineer, I am sickened
> by the many articles that have been written which make
> the claim that computers will never be able to play a
> decent game of Go. And I aim to put a stop to them.

I understand exactly how you feel. I subscribed to this mailing list 
more or less for the same reason.

> Also, Deep Blue (or Big Blue, or whatever is the name
> of it) didn't help the reputation of AI either by
> massively storing and doing look-up of all the chess
> moves, and then calling it an AI approach. Because
> it's not. Procedural programming is not AI. They will
> never be able to try that trick with the game of Go.
> There are way more possibilities than in chess.

And that's the very least there is to say about it. Chess is very 
unidimensional in terms of tactics compared to Go.

> Similarly, GNU Go will never conquer the game of Go by
> relying exclusively on a procedural, pattern matching
> approach. It's a nice start, and it is impressive how
> far you folks have gotten with it. But it aint gonna
> get the job done.

I agree up to the procedural part, although you seem to underestimate 
pattern matching. Isn't it more or less how humans approach the 
problem? Besides, there is some work into teaching neural networks to 
play Go, and so far results are encouraging.

> So, with that being said, I hope you can forgive my
> loss of patience. I have a lot of work to do, and I
> came to this list for technical advice. Instead, I get
> hung up with all this logistical chit chat about FSF.
> Anyone who has ever taken an Introduction to Computer
> Science course knows how FSF works. It is not new.

Given GNU Go is Free Software, this is very relevant, don't you agree?

Also, keep in mind that GNU Go IS a procedural computer Go player. If 
you dislike the concept that much, wouldn't it be much wiser to start 
your own project? What's the point of building on top of something 
you'd like to scrap?

> > > > Most users would not even know what a planner
> > was
> > > > and that GNU Go had a use for one.
> > >
> > > Duh. That's the whole purpose of the effort.
> > Nobody
> > > knew what personal computers were useful for
> > either,
> > > until Bill Gates starting building them.
> >
> > I didn't make any sense out of the second statement.
> > Did Microsoft already claimed that it was Bill Gates
> > who invented computer or what?
>
> Bill Gates wanted to build a computer that
> non-programmers could use. Everybody gave him hell,
> because it was so counter-intuitive. They didn't
> understand what use someone who couldn't program a
> computer would possibly have for one.
>
> All of a sudden, everybody understands now.

You definitely give that guy way too much credit.

> Best regards,
>
> Eric
>
>
>
> _______________________________________________
> gnugo-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/gnugo-devel