Re: [computer-go] Pattern matching - rectification & update

["Frank de Groot" <frank@xxxxxxxxxxxxx>]

From: <Compgo123@xxxxxxxxxxxxxxxxx>
Subject: Re: [computer-go] Pattern matching - rectification & update


> Frank, would you willing to test your program live on IGS. I'm sure there
are
> many 5 - 6 dans on this mailing list. Let's setup a date and time.


I am of course very interested in any kind of public exposure but at the
moment it still is a research module.
Like I said it needs a lot more work before it is a program.

That will take a few months and I am planning to send the program for free
to Go servers, Go magazines, dealers in Go products etc.
It is a disk full, 600 MB compressed because the search database indexes are
so big. (The program is able to find IMMEDIATELY any game with any pattern
from 200,000 games and that takes hundreds of MB in indexes.)

Also, I have no experience with Go servers, neither does my program yet have
an IGS interface.

Furthermore I have no idea how this pattern-predictor plays Go. It seems to
be better than I am, but that means nothing, as I don't play Go, I just know
the rules and have tried the game a few times.

My experience is that when one delivers a program that has some bugs, that
Beta testers might tell the world: "This program is nice BUT.." so I prefer
to work until people say: "This program is nice, has some limitations that
will be improved but it has no bugs".

I have been working for a few years on it and the end is near anyway, at
least for the pattern module.
There is a lot more to this pattern module I havenæt talked about yet.

Any pattern can have a *description*. You type it in, it is stored in a
database and shown when you hover over its central point with the mouse.

When 5000 people download it and give a few patetrns a name and my program
emails that back to me, I can build a database with all Fuseki & Joseki
named and redistribute it.

But then I thought: Why not let the program upload the pattern descriptions
to my server and download it back when needed, so it will be a kind of
"distributed intelligence".

Also, I want people to be able to define their own patterns.

Finally, to remind you, it is NOT a Go playing system.
It does not know how to play Go.
It can predict pro moves.

It can not predict deeper tactical moves, only "shape".
So if a strong player would evaluate it, (s)he would say: "This Go program
sucks, Frank's software sucks" because people still don't seem to understand
that a program that appears to play Go and predicts half of a pro's move is
NOT a "Go program" by any means.
It is just as little a Go program as "Eliza" is a psychologist.








_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/