More OTT AlphaGo hype?

[John Fairbairn]

AlphaGo's achievement is remarkable. I have no quibble with that. But I do have a quibble with the over-the-top reaction of some humans. It's as if it even invented sliced bread.

In the latest AGJ, we see the following: "Fan Hui also showed some moves from AlphaGo’s games against itself. Some of the AI program’s moves can really challenge our go beliefs, he said, citing the unorthodox attachment at left. But AlphaGo rates this the best move here, which may mean the Chinese fuseki will not be the same in the future."

The position shown (where the key moves are black stones on C16 and J17 and white's attachment on D17) didn't look unorthodox to me, and certainly not new.

Sure enough, the earliest example (of 15 to 28 depending where you draw the quadrant cutoff) in the GoGoD database was 1924 (Hayashi Yutaro vs Honinbo Shusai). The latest example was Yi Se-tol himself playing it in 2015 - in the China Weiqi League of all places, so it must have been know to Chinese commentators.

Maybe a little restraint is in order?

[Knotwilg]

I think it's fueled by our craving for real novelty, in a timeframe where nearly all top players seem to have a similar game.

There is actually a great potential for novelty built into the game of Go, which has been left unexplored: what happens to Go on larger board sizes? And what happens to AlphaGo against human players, if its expert database has been erased and must fight human intuition point blank.

[Kirby]

Somehow, I'm reminded of Hikaru no Go.

[Satorian]

so it must have been know to Chinese commentators.

But was it acknowledged to be the best move?

[Uberdude]

I also remember Lee Changho playing that attachment in a rather similar position just a few years ago (but with white sliding rather than taking the 4-4 corner of the Chinese formation, and an extra stone on the side out from the 3-4). That was against new pro Cho Insun and I wondered if Lee was perhaps being a bit more adventurous against a weaker opponent ("I can get away with it"). https://gogameguru.com/tale-meijins-lee ... cho-insun/

I certainly think there is hyperbole around AlphaGo playing moves "humans would never think of" which humans have indeed played, particularly bad in the non-Go press. AlphaGo isn't constrained by conventional wisdom and seems to be good at finding the best move for the situation. So whilst I don't think there is any of its famous moves a human pro couldn't find (the r10 invasion on the right in game 1, early joseki peep or shoulder hit of game 2), with perhaps the exception of the armpit hit in the early fight of game 3 (which I've heard was the favourite move of the series of some pros, and possibly mine too, it will be intriguing to see Gu Li and Zhou Ruiyang's thoughts on that!), I don't think there is a single human being who could find all of them (except Go Seigen maybe, RIP, it's a shame he didn't get to see AG).

[jeromie]

I think we're witnessing a shift among some professionals from viewing their community as the highest standard of go knowledge to recognizing a higher authority on "correct" play. Whether this is currently warranted is up for debate, but it does lead to some bold claims about AlphaGo revealing the right way to play.

I don't think it's unprecedented; it has also happened when a single player (e.g. Go Seigen) has dominated the professional scene. His commentary on the game was received with a higher degree of authority than other professionals' thoughts. He earned that respect, of course, but because he was a man his ideas could always be challenged. When the natural human tendency to appeal to authority is combined with popular perceptions about the infallibility of computer systems, it can lead to a blanket acceptance of AlphaGo's evaluation and a higher regard for moves that it approves, even if they are not entirely novel.

[Javaness2]

I'm sure some restraint is in order; not that this should get in the way of a good presentation. What was the previous assessment(s) of this move in Chinese commentary - beyond it being unorthodox?