Arimaa Forum - Print Page


    
      
        Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> Off Topic Discussion >> [Go AI] MoGo beats 8-dan pro with 9 stones (19x19)
        
(Message started by: aaaa on Aug 8^th, 2008, 7:08am)

Title: [Go AI] MoGo beats 8-dan pro with 9 stones (19x19)
Post by aaaa on Aug 8^th, 2008, 7:08am

http://www.usgo.org/index.php?%23_id=4602
http://games.slashdot.org/games/08/08/08/1243244.shtml
http://www.godiscussions.com/forum/showthread.php?t=6844

Title: Re: [Go AI] MoGo beats 8-dan pro with 9 stones (19
Post by Fritzlein on Aug 8^th, 2008, 7:52am

Wow, big news! Thanks for the link. Although I'm not quite on board with the folks who interpret the comment "it took us a year to drop nine stones" to mean that in one more year a supercomputer will be playing for the Go Championship at even handicap. I take it more to mean that Go isn't completely transcendental. Maybe in fifteen years a computer and the World Champ can play on even footing. Just about the same time it is happening for Arimaa. :)

Title: Re: [Go AI] MoGo beats 8-dan pro with 9 stones (19
Post by chessandgo on Aug 8^th, 2008, 10:43am

hey, thanks for the news ! That's quite impressive indeed.

This said, it does not look like the program is 2 or 3 amateur dan as announced by Kim. I guess a 2 dan amateur would almost always beat a top pro with 9 stones, and in this game Kim made a big blunder. Just browsed the game quickly, and it seems the computer played a few moves of almost no value :)

Yet quite impressive.

Title: Re: [Go AI] MoGo beats 8-dan pro with 9 stones (19
Post by omar on Aug 8^th, 2008, 12:47pm

Interesting news. That was one heck of a super computer; 800 processors running at 4.7 GHz each; wow!!!

The following is just my opinion and others are free to express their own.

When machines like this are used to defeat a human I feel as if it is just a laboratory demonstration and doesn't yet impact Go. Even if the human player had lost in a 9-stone handicap (favoring the human) to such a machine I would still say we have not reached the goal in concurring Go. Unless of course someone is willing to sell a machine with such power in large quantities and at prices most people can afford. I am not saying that this is not a significant achievement by the developers. It certainly is; it's like capturing a piece in Arimaa; though that is not the goal it helps us eventually get there. In fact the achievement happened earlier when programs using the Monte Carlo approach started winning the computer championships. This progress would have gone unnoticed by the general public if it had not been for this demonstration match. Defeating a top human does help to bring attention to the game and the progress being made. I'm sure the media will pick up this story :-)

I have a funny story to share. After Frank defended the Arimaa challenge in 2005, I did a press release about how humans are still able to defeat computers, but no major news outlet picked up on the story. I finally called an editor at AP and he basically said: 'yes we saw that story come by, but decided not to cover it; let us know when the computer defeats the human' :-) I don't think it is true that the general public is not interested if humans defeat computers, but the reverse seems to get more media attention.