It is currently Thu Mar 28, 2024 8:41 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 2 posts ] 
Author Message
Offline
 Post subject: Shuei and AI
Post #1 Posted: Sun Oct 13, 2019 11:42 am 
Oza

Posts: 3647
Liked others: 20
Was liked: 4626
You may recall that I presented a series here recently on Ohashi Hirofumi's work in comparing some ancient Japanese greats with a version of Golaxy specially trained with no komi. The greats came out rather well, perhaps not so much as compared with AI bots (with Dosaku as a possible exception) but as compared with modern pros.

Jowa and Shusaku were the other two player's in Ohashi's selection, but I have been slowly ploughing through proof-reading of the much updated version of my Games of Shuei, and as I go through each game I pick odd moves to see what the bots say. In general Shuei also comes out rather well, but there was an example I just looked at which seemed worth talking about for several reasons.

One reason is that there is recorded account of when Shuei played a "joseki" move and startled his watching pupils. They wanted to know how he came up with that move. In essence, he simply told them to study old games for themselves. The full story is in my biography of Shuei, Meijin of Meijins.

The other reason is that there was a brilliant article by Murashima Yoshinori which explained the best way to consider joseki. I'm not going into that here, but I spent two pages on it in the book. What I will say though is that Murashima recommended not viewing josekis through the prism of profit and thickness, but in terms of the fighting shapes they produce. (The rest of my two pages explains how, and why Murashima came to that conclusion.)

Murashima's ideas happened to be a perfect explanation also of Shuei's apparently new move. But I decided to see also what Lizzie had to say, and in the process I came to the conclusion that Murashima's theory seems to represent an ideal description of AI play, not just in josekis but in the fuseki. "Fighting shapes" is the key phrase, but other key phrases of his are "avoiding futility", and positively and negatively charged moves, respectively defined in terms such as whether they make a centre-facing attacking shape or a defending shape (i.e. utilising the edge). Murashima also explained how to acquire the skill to use these concepts.

The Shuei move in question was the one marked below. It was a 2-stone game (one White stone has been captured) and if you have the GoGoD database and wish to see the outcome, it is game 1899-01-15b.



For the Lizzie test I added a White stone in the lower right to un-handicap the game. Lizzie had eyes for nothing else but Shuei's move. Eight other moves showed up on the radar screen as barely visible red dots and didn't merit more than tens of playouts. The best of those, however, was A and that could definitely be called a "fighting shape" move. I wonder whether in fact all of the fancy AI high moves may come under that category (another pro suggestion for these moves has been as "network" creators).

But what is really interesting about this position is that White has thickness. This may seem at first blush to defy some of the naïve theories that talk about connected stones, eye shapes and lack of cutting points. White here has a major cutting defect at B, for example, which leads to a 20-point+ corner for Black. But what is important here is that Shuei saw the future thickness, and I put it to you that a major skill in AI play is precisely seeing such future thickness - to a level even beyond Shuei, no doubt.

"Future thickness" may even be a term we should get used to talking about, although given the trouble we have with existing thickness that may be even harder than talking about Brexit :)


This post by John Fairbairn was liked by 2 people: Bill Spight, gowan
Top
 Profile  
 
Offline
 Post subject: Re: Shuei and AI
Post #2 Posted: Sun Oct 13, 2019 12:45 pm 
Lives in gote

Posts: 502
Liked others: 1
Was liked: 153
Rank: KGS 2k
GD Posts: 100
KGS: Tryss
Katago with 0.5 komi and 2 handicap agree with G13. It estimate white at -26 pts, up from -28pts at the start of the game.

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 2 posts ] 

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group