In terms of sheer playing performance, Monte Carlo and DCNN leave Gnu
Go in the dust.
But even Alphago has the fundamental flaw that it doesn't "know" what
it's doing, because neither DCNN nor Monte Carlo embodies the basic
concept of commonsense intelligence, namely being able to construct
and reason about a meaningful computational model of the dynamics of
the world.
Here is one example of how a program design which does embody such
commonsense can find a better move than Alphago did at move 79 in game
4 of its match with Lee Sedol, because it does "know" what it is
doing.
The program is designed, but it's up to others to turn it in to
software. Could it beat Alphago across the board, and with much less
hardware?
The design includes new algorithms for computing territory and
influence, group strength, and strategic and tactical reasoning. It
can talk about what it's thinking so would thus be useful to people
learning Go.
text: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2818149
video:
https://www.youtube.com/watch?v=Nns70047Rxo&index=27&list=PL4y5WtsvtduqNW0AKlSsOdea3Hl1X_v-S
_______________________________________________
gnugo-devel mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/gnugo-devel