computer-go: WinHonte paper

[Darren Cook <darren@xxxxxxxxxxxxxxx>]

For those of you not reading rec.games.go, a new program called WinHonte
was released in March (http://www.jellyfish-ai.com/winhonte.htm), and
can be downloaded for free. It is apparently one of the stronger free
programs.

There is a 1999 paper by the author (Fredrik A. Dahl) here:
  http://citeseer.nj.nec.com/dahl99honte.html

It is interesting, not just for the use of neural nets, but also for
some other ideas, e.g. using proof DAGs to cache results of tactical
search, and using pattern matching to alter the results of tactical
search by deliberately pruning moves that work but in a real game would
have far too bad a side effect to make them worth playing (I can imagine
that would have a bad effect on life/death evaluation however).

Does anyone know of any more recent publications by the author?

Darren

P.S. Can anyone explain diagram 4 to me: I don't see how the white cut
at B can do anything, and why black shouldn't have played A instead of
his strange first line move.