Skip to main content

View Diary: Obama's Soft Power - a primer on Aikido (303 comments)

Comment Preferences

  •  I had never heard of that game before, (5+ / 0-)

    so thanks!

    In reading the wiki piece about it - I am intrigued by the idea that it might also help us understand our role as citizens in this. Perhaps we can be the "stones."

    •  I love (7+ / 0-)

      how only a beginner can be beaten by a computer at go.  Chess is the computer's game, as it's linked with logic and mathmatics, but once a human gets a handle on the more organic go, only another human player is a worthwhile challange.

      I think I have trouble answering "Why is torture wrong?" because I'm dumbfounded by the question.

      by Anjana on Mon Jun 08, 2009 at 08:15:19 AM PDT

      [ Parent ]

      •  Why is that, do you think? n/t (2+ / 0-)
        Recommended by:
        Anjana, Deoliver47
        •  AI challenge is much much harder than chess (6+ / 0-)

          Short answer is there are an exponentially larger number of moves (19x19 board versus 8x8 board) and the game is more abstract as far as the value of different moves.

          It's basically way harder to program a computer to be really good, and there's so many more moves possible that you can't just program some decent AI and then "brute force" the rest like with chess.

          Most people think it'll be over a hundred years before we see the "Deep Blue" of Go.

          •  And when it comes out (2+ / 0-)
            Recommended by:
            Larsstephens, Thassa

            I'll totally buy it.

            Unless it turns out to be a real-world Skynet.

            I think I have trouble answering "Why is torture wrong?" because I'm dumbfounded by the question.

            by Anjana on Mon Jun 08, 2009 at 09:06:07 AM PDT

            [ Parent ]

          •  Go is construction, chess is a war of attrition (5+ / 0-)

            Go begins with an empty board on which each player builds their own territory with their (black or white) stones.  In an match between equals or a well handicapped one, the outcome of the territory count is usually a matter of a few points.  

            I think the AI comparison to chess can be put this way:
            Chess is a branching game where the computer, through brute force, can eliminate branches of the problem that don't lead to capture of the king. So it has a single objective.  Go is more full-board based in terms of the outcome -  the strategic objective is empty intersections to be surrounded everywhere on the board. Tactical situations are nodes that radiate influence strategically.  The AI problem (and I'm a go player but not a programmer) may be relating the tactical situations to the strategic objective - partly because on the 19x19 board there are so many possiblities.

            Maybe I prefer go to chess because I'm an architect.

        •  Branching factor (6+ / 0-)

          Branching factor wikipedia article

          This is the number of possible moves available for a given board position. This number tells you how quickly the problem "blows up", so that is becomes impossible to "look ahead" for too many moves.

          The branching factor is about 35 for chess. This means that to look 3 moves ahead (3 for you plus 3 for your opponent) you have to examine about 35^6 or 1.8x10^9 board positions. Alpha-beta pruning can significantly reduce this number, so that for chess, computers can run fast enough and search deep enough that only the best human players using specific anti-computer strategies can win.

          In 19x19 go, the average branching factor is about 300, about 7x10^14 board positions is required for that same 3 deep look ahead. Computers aren't that fast. Humans are much better at pattern recognition that computers are (why captchas work for validating human input), and pattern recognition, while important in chess, is absolutely crucial in go.

          Bigger hardware is not going to solve go. Better pattern recognition (known in AI as the evaluation function) might.  

          Hopelessly pedantic since 1963.

          by admiralh on Mon Jun 08, 2009 at 10:12:29 AM PDT

          [ Parent ]

          •  pattern recognition (6+ / 0-)

            quite so on pattern recognition.  go players see the board in snapshots that they've learned by playing and studying.  "look-ahead" is called reading and is often very deep, but one can only be effective in reading the ramifications of a particular snapshot or tactical pattern insofar as that relates to full-board strategy.  the importance of pattern recognition helps explain why artists can be as strong in go as mathematicians.

            back to Obama -  he must have a huge inventory of problem solving snapshots that help him politically.  and his objective is not to dominate, but to mediate.  

            there are many stories of powerful professional go players beating amateurs in teaching games by just a few points - on purpose, instead of wiping them off the board.

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site