[Computer-go] a pro game which is computer-unreadable
terrymcintyre at yahoo.com
Wed Nov 30 05:20:24 PST 2011
I have attached an SGF file of this game.
In addition to the "losing" ladder, there is a later fight in the lower left corner which is interesting. B won this game by resignation.
I once had a conversation with a pro who explained how to discover solutions to tsumego. He put it this way: "I try a solution. If it doesn't work, I ask what would make it work - perhaps an extra stone would take a liberty from my opponent. I think of a way to make that extra play, and try again."
This sort of meta-thinking could have led Lee Sidol to play out this "losing ladder" on purpose; losing the ladder provided an extra stone which enabled Lee to capture three white groups, completely changing the status of the lower right corner.
Any tree-following program is likely to get lost in a thicket of branching possibilities; a human player will consider the main line, the ladder, to be most probable, follow it (depth-first) to the end, and explore the effects at the end; then back up and consider a few likely branches, looking for a tsumego which might refute the main line of play.
If meta-reasoning can be fed into the search, it could propose using the "losing" stones to good effect elsewhere on the board.
This could be hacked, possibly, by storing information which tweaks the probability of proposing certain moves. Looking at the lower right, the dead b group can be revived if one of the surrounding w groups can be captured, which can be accomplished by reducing the liberty count by one. Perhaps moves which do so should be tested a little more often than otherwise.
Terry McIntyre <terrymcintyre at yahoo.com>
Unix/Linux Systems Administration
Taking time to do it right saves having to do it twice.
> From: Olivier Teytaud <olivier.teytaud at lri.fr>
>To: computer-go <computer-go at dvandva.org>
>Sent: Wednesday, November 30, 2011 1:35 AM
>Subject: [Computer-go] a pro game which is computer-unreadable
>sorry for taking some of your time with non-technical long-term AI/GO dreaming,
>but if sometimes you find Go fascinating you might like the video below :-)
>As many of you I guess, I've spent time trying to design some sort of learning in MCTS, so that
>monte-carlo simulations would be "adaptive" to the current situation. This idea looks like a very
>natural solution to the problems we have for reaching human top-level.
>I've met this incredible game; I'm not a Go player, but like many not-so-strong players at first view the moves by black
>look like a big mistake (misunderstood ladder):
>In fact, it's (as far as I see...) a very clever idea by black (Lee Sedol, pro player), in spite of the fact that it's a failed ladder.
>We tried various things for having machine learning in MCTS:
> - Contextual Monte-Carlo for online learning simulations: http://hal.inria.fr/inria-00456422/
> - poolRave (using RAVE values in simulations): http://hal.inria.fr/inria-00485555/
> - Bernstein Races for offline learning patterns http://hal.inria.fr/inria-00622150/
> (a synthesis of these papers in http://hal.inria.fr/inria-00544758/ )
>and many of you have published related stuff;
>but when a computer will be able to understand a situation as the game above, it will be very impressive to me :-)
>Go looks like a combination between feeling and mathematical reasoning. One day the people of the Go-sect will convince
>me that this game has something really special :-)
>In particular, my feeling is that a 10kyu can not play this pro game, but a 10-kyu can understand a posteriori. It's difficult the discuss
>the possibility for a computer to understand a posteriori, but with a little bit of provocation from this point of view computers
>are not yet 10-key :-)
>Computer-go mailing list
>Computer-go at dvandva.org
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 1579 bytes
Desc: not available
More information about the Computer-go