Re: computer-go: Pattern matching

["P.J.L.Cuijpers" <P.J.L.Cuijpers@xxxxxxxxxxx>]

> It may be the case that knowledge like office skills and perhaps even go skills
> may be so emulated.  It may be the case that computers can write sonatas in the
> style of Bach, or play games of go in the style of Takemiya, or of Go Seigen.

Just for your information:
A year or so ago I suggested a small research project at a group working
on
compression-decompression techniques here at the Eindhoven University of
Technology
(Netherlands). They have been working on compression of an English
Bible-text and
have the "world record of compression".

The idea of compression is to convert a non-random "string" into a
smaller string 
that is as random as possible. Decompression is the inverse process.
On my question "what happens if I feed a self-generated random string to
a decompressor" they answered they had tried this and out came
"bible"-like texts !!
The I suggested to try the same on Bach-sonates and try to make new
music this way.
I never got to actually doing the project but a few months later I heard
that
in Japan this very technique was used to identify whether a new-found
piece of
music was indeed an original Bach-sonate or not.

They taught a compression algorithm to be really good in compressing
original Bach sonates and then compressed the new sonate to see if the
compression was good enough to presume it was real.

The people who work at this project say it is not about neural net
training but
I see a great many similarities.

Alas I have no references to articles... :-(

Greetz. Pieter