MoGo vs Human

  

MoGo vs. Human


Back to forum


Thibault de Vassal    (2008-07-23)
MoGo vs. Human

In 1997 Janice Kim (1 dan) beat Handtalk, then the strongest Go program, despite giving the program a 25-stone handicap.

On Thursday, August 7, Kim MyungWan 8p will play MoGo, probably the world's strongest computer Go program. MoGo will be running on a supercomputer boasting over 3,000 processor cores !! The game will be broadcast live on KGS - http://www.gokgs.com/download.xhtml

The human is "8p", meaning 8-dan professional; not quite 3 stones stronger than a 1d pro player, who in turn would give an amateur 1d at least 6 stones. Edward Lasker said that 3 stones handicap at Go is comparable to knight odds at chess.


Svante Carl von Erichsen    (2008-07-23 17:33:36)
Janice Kim

Janice Kim was a 1 Dan professional in 1997, she is 3 Dan professional now. Is there more information available somewhere, e.g. regarding handicap and time settings?


Thibault de Vassal    (2008-07-23 18:40:31)
1 dan Pro

Thanks for correcting me :)

I did not find more information about the match yet.


Tano-Urayoan Russi Roman    (2008-07-26 09:38:07)
Some info

I found this "According to the AGA eJournal?...a human will play go against MoGo running on a 3000+ node computer. According to the computer go mailing list.. it is planned to be 19x19, will start with 5 blitz games to decide a basic handicap, and then a 'real' game will be played."


Thibault de Vassal    (2008-08-08 13:19:10)
MoGo wins ! (estimated 2 dans)

I just read it in the American Go E-Journal, MoGo computer program defeated Myungwan Kim 8P by 1.5 points in a 9-stone game billed as “Humanity’s Last Stand?”

The professional player estimated MoGo’s current strength at “two or maybe three dan”, “made some 5-dan moves” (the program used 800 processors, at 4.7 Ghz, 15 Teraflops on a borrowed European supercomputer)

Strangely, Kim easily won two blitz games with 9 stones and 11 stones and lost one with 12 stones and 15 minutes by 3.5 points before this one hour game.

http://senseis.xmp.net/?MoGo

http://en.wikipedia.org/wiki/Computer_Go


Robert Mueller    (2008-08-09 17:34:44)
3,000 processor cores?

... supercomputer boasting over 3,000 processor cores ... I want one! :)


Andrew Stephenson    (2008-08-09 21:55:37)
Mogo

Did the human have a handicap in the 1 hour game? Does the player moving 2nd generally get some points/stones to compensate between equally rated players?


Don Groves    (2008-08-09 23:12:15)
MoGo

MoGo had a nine stone handicap in the long game. I just read this morning that MoGo can analyze 1.7 million complete games per second using Monte Carlo techniques. Since the event took place right here in Portland, we had some coverage in the local media.


Don Groves    (2008-08-10 00:32:11)
Komi

Yes, Andrew, the second player is compensated with some number of stones, usually between 5.5 and 7.5 (ties are eliminated). This is called komi. I don't know what the komi was in the MoGo match.


Andrew Stephenson    (2008-08-10 07:45:29)
Mogo

So mogo gets to put down 9 stones to start with but I dont know whether he moved first or second - it sounds like he moved first and there was no Komi. Anyway its a big handicap but the breakthrough appears to be that the win was achieved on a 19x19 board in a "long" game (1 hour) Kim didnt use so much of his time but said more time spent would not have made any difference prononcing Mogo invincible at 9 stones and very difficult with 8 stones. The programmers were excited because they said 1 year ago they needed 18 stones now 9 and maybe a year to lose the other 9! If they can maintain this rate of improvement then they are suggesting that in a few years mogo could be the strongest go player in the world. Interestingly there is a reversal here with chess: programs being stronger against humans the shorter the game (ie blitz) but Mogo did better with more time! I guess this is about the time Mogo needs to assess the long term consequences of each move.


Thibault de Vassal    (2008-09-20 03:29:51)
MoGo - Rematch

According to the AGA E-Journal : "Myungwan Kim 8P will take on Mogo in a rematch set for this weekend's Cotsen Tournament in Los Angeles."

Good news :)


Don Groves    (2008-09-20 05:03:31)
MoGo

Hi, Andrew -- It would be an amazing feat if MoGo programmers could eliminate its nine-stone handicap in only one year of development. With each stone eliminated, the combinatorial aspects increase exponentially and the human advantage in pattern recognition and game understanding increases in proportion.

FYI - In a handicap game, the weaker player always goes first (ie., plays black). The handicap stones are placed on predetermined points and that constitutes black's first turn. White plays next and they alternate the remainder of the game. Komi still applies unless otherwise agreed.