Yes, I recall that earlier episode. I would be happy to have a better relationship going forward.
I wrote some explanation generators for Scrabble and Chess AI, but these were much simpler systems that I could break apart. E.g., the Chess engine would play two moves out until a "quiet" position was reached, and then explain which evaluation parameters were affected. But even this was not easy, because sometimes after playing out for a bit the AI would "change its mind" by deciding that the other move was better after all. And evaluation factors were sometimes too numerous to list individually, even after pruning the factors that are too small to change the decision. And explanation by "comparing factors" assumed a linear evaluation function. So I see huge challenges scaling that up. Another approach that works well for tree-search games is just to let the human explore, in the style of an opening library. The engine simply responds to the moves that the human proposes, creating trees and assigning values to endpoints. MCTS programs show a "board control" visual, which might be enough to explain the positional evaluations. Basically: let the human absorb a lot of case studies. There was a paper about postal-go where a human player used multiple programs in this way to construct a phenomenally strong player. -----Original Message----- From: Computer-go [mailto:computer-go-boun...@computer-go.org] On Behalf Of "Ingo Althöfer" Sent: Thursday, March 31, 2016 4:31 AM To: computer-go@computer-go.org Subject: Re: [Computer-go] new challenge for Go programmers Hello all, "Brian Sheppard" <sheppar...@aol.com> wrote: > ... This is out of line, IMO. Djhbrown asked a sensible question that > has valuable intentions. I would like to see responsible, thoughtful, > and constructive replies. there is a natural explanation why some people here react allergic to Djhbrown's new contributions. He had an active phase on the list already from early August 2015 to mid October. Things started interestingly, but somehow he went into a "strange loop", and in the end he was asked to stop posting. http://computer-go.org/pipermail/computer-go/2015-October/008051.html Perhaps all sides can help that things run better now. Ingo. PS. For my interest in computer-assisted human go visualisation questions on DCNNs are indeed interesting. _______________________________________________ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
