computer-go: Agent Go techniques

["Julian Churchill" <jjc97c@xxxxxxxxxxx>]

 Hi there,

 I was just wondering if anyone had attempted to use agent architectures
and/or distributed computing techniques to write a Go program. It seems to
me that what is needed to tackle this problem is more power ! :) This would
also allow the best of all sorts of techniques to be used and combined (in
some manner... ) effectively getting the cream of the crop. It is obviously
not just a brute force challenge such as the code breaking coordinated by
distributed.net and others, and high quality AI search etc... techniques
will still be required. For instance, you could combine a neural net
program, a brute force search and any number of other techinques to
participate in the 'GoWeb'. You would simply feed in the current board state
and the distributed collective of programs would apply themselves, using
their own specialities, to suggest the best move in their estimation. I
think this may work well for Go since different stages of the game seem to
have varying success with many different techniques of finding the best next
move. The only thing would of course be, how would you then choose the best
suggestion to represent the distributed network?

 Cheers,
 Jules