Cho Chihoon vs AI(DeepZenGo)

[PlaySlow]

Link's to Antti and Michael's commentary.

Game 2

https://www.youtube.com/watch?v=A_f4ZYlPsLY

Game 1

https://youtube.com/watch?v=QDVUXWKshC8

[Gomoto]

Thanks for the links!

[Bonobo]

Since I’m not aware that we’d already have an SGF for the 2nd game, here is one, posted by Shigeru Mabuchi on the “computer-go” mailing list (http://computer-go.org/pipermail/comput ... 09633.html)

http://www.lifein19x19.com/forum/download/file.php?id=7948

Denousen2016-Chochikun-DeepZenGo-Round2.sgf

(1.59 KiB) Downloaded 1593 times

[Bonobo]

Link's to Antti and Michael's commentary. [..]

Thank you … linking them inline and adding the other videos with Myungwan Kim’s comments also so that everybody can watch directly here:
________________

Game 1
Nihon Ki-In video with comments by Michael Redmond and Antti Tormänen

AGA video with comments by Myungwan Kim

Game 2
Nihon Ki-In video with comments by Michael Redmond and Antti Tormänen

AGA video with comments by Myungwan Kim (yeah, it’s really game 2 )

[Bonobo]

I think the SGF with comments by Michael Redmond hasn’t been posted here yet, has it?

http://www.usgo.org/news/wp-content/uploads/2016/11/2016.11.20_Zen_ChoChikun_G2-redmond.sgf

(From http://www.usgo.org/news/2016/11/deep-zen-go-wins-game-2-final-game-tuesday-night/)

[pookpooi]

With this result I estimated that DeepZenGo is about the same level as AlphaGo v13 run on single machine version. But my friend said that Cho Chikun is much stronger than Fan Hui, so it'll take AlphaGo v13 on distribute version to win against him even just one game out of three. Can Zen reach v18 strength before Dwango contract terminated? (March 2017)

[Uberdude]

With this result...

For those who missed it: Cho won the 3rd game. Zen made a big moyo with Cho taking solid territory and Cho played an endgame move that looked a bit small when Zen ignored to close the centre in a large scale, but Cho's move had some added value in giving the option to take one of Zen's groups down to 1 eye in sente and then Cho made a cut and Zen suddenly resigned (I think Zen can still live but Cho gets good sente profit).

[EdLee]

Zen suddenly resigned

A little bird said Zen's self-assessed win rate was ~60%, and it was Kato sensei who resigned for Zen, not Zen itself.

Mr. Iyama was still studying the position (inconclusive).

The Chinese top pros watching were also studying the position (also inconclusive).

( My gut feeling is it was a human error to resign for Zen ;
Others more resourceful please fact check... )

[pookpooi]

Interesting, look like Hideki want to prevent long moment of awkward that will lead to 'AlphaGo resigned' , but he should trust his program a little bit more. #RobotRight #EqualityForAI

[Uberdude]

http://eidogo.com/sgf/ABYz5kZ1.sgf

[kimidori]

I watched the AGA, Kim Myungwan considers that B is slightly ahead before the sequence (after moves 150), and to him Zen already had no more than 10% chance to win at the moment. He also evaluated the sequence a bit after Zen had resigned, and he seems to agree that the sequence "B profits and takes sente, while W lives" is the best sequence for both, thus Zen resignation didn't really surprise him.

Interesting to know that it actually causes Iyama and the Chinese top pros some headache . Does anyone have their final conclusion of the board?

[DrStraw]

It seems that if the computer is playing then for a human to resign would be against the rules.

[Bonobo]

Game 3
AGA video with comments by Myungwan Kim

[pookpooi]

From Japan Times

“I felt as if I was playing with a human, because (DeepZenGo) has both strong and weak points,” Cho said after the final game.

DeepZenGo “suffered a crushing defeat today,” developer Hideki Kato, 62, said. But he added: “I know exactly what needs to be fixed.”

(And interesting that v13.1 is already testing, Yoji Ojima is very hard worker.)

[Bonobo]

From Japan Times [..]

Thanks a lot, pookpooi!

Here’s the link to the article in the Japan Times you mentioned.