It is often remarked that AI cannot explain why one move is better than another, whereas humans can - or pretend they can.

The example below seems to fall into the "pretend they can" category. It is Honinbo Shusai, no less, saying that Black 3 (the triangled stone) should be at A. And then he comments that White 4, which was at B should "likewise be at C."



We can infer that he thinks the upper-left corner demands priority, but he doesn't say why. Nor does he say why the mokuhazushi is correct, although at the time of this game (1911 - between Honinbo Shugen and Kita Fumiko) mokuhazushi was fashionable. It was a no-komi game but that doesn't seem to matter at this stage.

Since AI seems to disagree with Shusai as regards both corner and the point within that corner, I think we can further infer that, ultimately, he can't explain why his choice is better, simply because it isn't.

But he stated it in a national newspaper column and so put his reputation firmly on the line. What is your view as to what was in his mind?

It seems to me that he is trying to say something about the upper side, though I will toss out the possibility that, while doing that, he is also concerned with matters of the mind. By focusing on the upper side, Black would be keeping the initiative (yes, yes, that again - but it's important and it's got nowt to do with English sente). I say this not just on the basis of what may be on the board but on the basis of many remarks by top pros when explaining either why they think a move is good or, more interestingly, why one potential pupil is more promising than another. The most promising one is one who strives to keep the initiative.

However, there is another unusual comment that may throw a sidelight on the matter. The following opening of the same period (Inoue Inseki vs Kita) elicited the comment from Iwasaki Kenzo 8-dan that Black 5 (triangled) could well be at A.



Iwasaki's reasoning was that if Black played A and White came in at the triangled point (TP), Black could extend to B. Then, "as has been frequently remarked hitherto," White would have no really good point to play on the left side between TP and the star-point stone above. I infer that this shows that discussion of the sides was in full flow at the time and, if so, that they didn't completely understand the sides completely then. I would posit that the reason mokuhazushi was fashionable was simply that so much focus was on the sides. That focus, in turn, I would suggest, came about because it was becoming so difficult to make any headway with White in no-komi games, and so new ideas focused on the sides were the start of a desperate search by White for the initiative - a search that really only ended with the introduction of komi. That would, I suggest, also explain something about Shusai's thinking in the first position.

It's a bit of stretch, perhaps, but I am also tempted to believe that this search for play on the sides was only resolved when Go Seigen came up with the notion of Go Seigen groups (my term), the significance of which is that he also roped the centre into the equation - and that is also why his play looks very AI-like.

This implied search for best moves on the side is also mirrored in old Chinese go (with group tax) where they hummed and hahed about the best location: 10-3, 9-3 or 8-3, eventually deciding on 9-3. It is noteworthy that old Chinese go was also no-komi go.

Excellent Point, Micheal Redmond also notes that, unlike humans, AI doesn't change the way it plays depending on whether it's white or black.

Here you give me the idea that sente, first move, is simply one of the many kinds of sub-initiatives on the board, and it's the least important one. The overall initiative is the weighted combined mean of all the sub-initiatives.

I see the Black move at A as a double purpose move, being an initial move in an empty corner and also a very distant pincer of the white stone at P17, or rather limiting White's prospects for development on the upper side.