[gnugo-devel] address@hidden: Re: GNU go and Artificial Intelligence?]

[Daniel Bump]

Dear Mr. Rasmussen,

Your query was forwarded to me by Bradley Kuhn. I've
taken the liberty of forwarding it to the GNU Go development
list. You can communicate with the list by sending mail
to address@hidden Let me remind anyone replying to
Mr. Rasmussen to add his email address to the cc line
since he is not a subscriber.

Some of your questions call for speculation, others
for extensive explanation. Briefly, GNU Go uses a
variety of different searches to evaluate various
moves. The term search refers to the generation of
future board positions (thousands or millions) in
order to evaluate the effect of a certain line of
play. Searches are divided into tactial and strategic.
The generation of moves in tactical search is
hand coded. In strategic search, it is driven by
pattern matching. I think you can get more extensive
explanations by communicating with the list. You
may also get some feeling for how the program
works by browsing the documentation at:

http://www.gnu.org/software/gnugo/gnugo_toc.html

An important aspect of GNU Go is that it is Free
Software. This means that we publish our source
code and permit its reuse by other program writers
under the terms of the GPL (GNU Public License).
See:

http://www.gnu.org/philosophy/free-sw.html

I think that GNU Go, in addition to being a program
that competes with the top program, has had a
beneficial effect on the field by being Free. As
a concrete example, GNU Go implements a new
protocol, designed by Gunnar Farneback, called
the Go Text Protocol, for communication with
other programs. It seems that the GTP is gaining
acceptance as the future of interprogram
communication among Go programs. Another example
is that other programmers often use GNU Go as
an opponent in developing their own programs,
and the spirit of openness is contagious.

I am travelling now and will not be back in the US until
Monday. I will not have internet access before then
after this message. Please take further discussion
to the mailing list.

Daniel Bump

> My name is Ole Dahl Rasmussen, I am a freelance journalist and I am
> writing an article on Go and Artificial Intelligence for the
> Scandinavian popular science journal Illustreret Videnskab. I called the
> Free Software Foundation today, Thursday, and they gave me your email.
>
> In connection to this article, I have some questions that I hope you
> would be able to answer, or that you could eventually direct me to
> someone who could.
>
> The questions are the following:
>
> I have heard that computerprograms are not very good at go. Why is it so
> difficult for computers to play go?
>
> How are you trying to learn them? What techniques do you know of to
> learn computers to play go, and what are the special features in GNU go?
>
> In a broader technological historical perspectives there are some
> examples of surprising developments. E.g. it is often said that the
> space-research helped creating the micro-wave owen and the teflon
> material. In ten years, what developments do you think Go could lead to?
>
> You can either answer the questions electronically or e-mail me a phone
> number and a time that would be appropriate ? then I would call you
> back.