I have mentioned several times here that I find the way Japanese pros are talking about the insights AI has to offer rather different from the way people here talk about them. Obviously to a large degree that must reflect the pro/amateur difference, but you may recall that I have also mentioned recently something I have noticed for myself from compiling Go Wisdom (GW) indexes for pro commentaries I am now presenting in books. Whether these two observations actually go hand in hand I can't really say, but at the very least the GW indexes have made me even more alert to usage of go words.
Two aspects of GW struck me. One was that it is important to think about nexus (which is the plural of nexus, by the way) rather than individual terms. The other was that the frequency with which particular terms came up differed markedly between "them" and "us." I am here concentrating solely on Japanese texts for "them" because I am getting too old and lazy to read the latest Chinese and Korean texts, but I have read plenty of C & K pro commentaries in the past, and as a linguist have always been sensitive to language, so you might indulge me and accept that "them" means CJK pros here and "us" means amateurs - rather than oriental/western. As to specific words, I find that the main difference lies in the pros focusing more on overconcentration and efficiency, but there are other terms where differences emerge - see below.
In the latest Go World Shibano Toramaru has been invited to talk about AI, and some of what he had to say thoroughly astonished me despite the fact that I was already primed to expect a them/us difference, so I thought it might be of interest here if I added a few remarks to show more clearly the differences I am talking about.
Shibano chose to talk about the joseki in the lower right below.
This White attachment is not new. It goes back to 1925, and for some odd reason it was tried several times by different players against Go Seigen - as if they were looking for a secret weapon specially tailored to him. Given that Shibano entitles a section in his piece "An arrangement emphasising speed," that doesn't have to be far fetched. Go was famous for his emphasis on speed, and his attaching opponents may have been trying to pre-empt him by upping the pace before he got to control it. And FWIW Iwamoto scored a rare victory over him in their eleven-game match that I am currently preparing for publication. (Hence my special interest in Shibano's article.)
A further interesting aspect of Shibano's prefatory remarks is that he called the attachment White 6) a "probe." Probes were one thing that really caught my eye in compiling Go Wisdom. There was (for me) an unexpectedly high frequency of mention of probes, and perhaps the most characteristic occurrence was when a mature pro (e.g. Honinbo Shusai) is commenting on the games of a still immature pro (e.g. Hayashi Shuei). It seems the true masters simply play more probes, and playing more probes is one more step towards mastery. But the recommended timing of probes was unexpected, too. Here, Shibano is talking about a probe on move 6. The earliest I have seen is move 4. There is also the unusual concept of a counter-probe.
But the human attachment essentially died out after WWII, possibly in connection with the widespread introduction of komi - until AlphaGo revived it (in self play) in 2017! Since then it has become a fad among Korean and Chinese players especially, with several very recent examples. It's still very rare in Japan, though Iyama did try it once and won, all of which perhaps makes Shibano's interest in it a significant pointer for his go.
The GoGoD database shows four Black replies (not counting tenuki). A is by far the commonest and then the order is B down to D.
The bots have tried all except D.
What did Shibano say about B? He gave the following result, which astonished me (I'll leave you to work out the actual moves - part of the Go Wisdom philosophy is that "effortful practice" and hands-on experience are essential for improvement.)
Shibano said this was fine for White, and added that if Black now plays A (then White B, Black C), White will be very happy to extend to D.
Several surprising things here swirled round my brain. One was the early sacrifice of two stones to get a wall that didn't even have an extension yet (i.e. the eventual extension is ultimately gote, the opposite of "speed"). Most obvious of all, of course, was that HUGE Black corner "territory." But the biggest surprise was that Lizzie agreed entirely with Shibano. The initial win rate did not budge at all up to the position shown (i.e. White's play was "perfect"), and if Black did continue with A etc, that was a significant mistake (53.5% down to 57%). Lizzie says Black's next move has to be on the right hand side, and parity is maintained.
There was a glimmer of understanding once I'd read Shibano's comments. Of course I had already realised that White has two unchallenged stones on the left side to compensate for any Black gains on the right, but my intuition - wrong intuition obviously - is that a flock of birds in the hand is worth two fledglings on the left side.
The glimmer was Shibano's comment that Black in the first position above was already solid, and so he does not mind making him more solid. In other words Black is overconcentrated (bear in mind my exhortation to think in terms of nexus rather than individual terms - "solid" is part of the overconcentration nexus).
Shibano says quite a lot more about this variation, but moving on now to Black's underneath attachment, he shows the following position.
Here the two White limpet stones have not only overconcentrated Black but are "efficient" because they inhibit a Black invasion at A.
Of course, Black does not have to be a passive onlooker, and he can try to make White "heavy" (part of the overconcentration nexus), as follows:
Shibano says this does not really work for Black because, even if White's stones were heavy, there is no way Black can mount a serious attack on them. Two things leapt out at me here. One is that he makes no mention of any Black territory in the lower right!
The other is that he refers to the resulting White shape as a "kamae" (and not thickness). It may surprise you how common this term is, as GW reveals. Westerners either call it a moyo or don't call it anything all - don't even recognise it is a concept at all, perhaps. It is a "construction" move. You are building a barn. A moyo is just putting stakes in the ground. I believe early farmers discovered the value of building barns and granaries before they thought about tarting up their houses. There's a message there, and I think it's been ignored in the west - with one possible exception. Robert Jasiek has mentioned a concept I think he also called "construction." I haven't read his relevant books and he chooses not to share his ideas here, so I can't say whether it's the same thing (I suspect he may be focusing more on bases - garden sheds rather than barns, but even that is still worth noting).
Another example Shibano gave was option D:
Here Back is unequivocally overconcentrated, even if to a very small degree. Black has a strong "kamae" in the lower right but, again, Shibano does not resent making him stronger where he is already strong, and White's stress-free shape on the right allows him to look forward to a relaxed game, especially given the large komi.
For completeness, a brief mention of option C (which is where White invariably plays if Black tenukis BTW):
Shibano suggests White initially ignores Black's reply. The probe has done its work and, seeing that Black has declared his digging-in strategy, White now concentrates on denying Black expansion possibilities. The means grabbing big points (the two knight's moves on the top side are especially highlighted by Shibano as "good judgement"). White's initial attachment then eventually becomes palpably efficient (i.e. the attachment has proved worth making, he says) as part of a strategy of erasing Black's emerging moyo. White can now look forward to a satisfactory game because he has a preponderance of stones on the upper and left sides to compensate for Black's already constrained moyo.
I hope this explains some of my earlier remarks while also giving a fascinating glimpse into what pros are making of the AI revolution.
|