I don't see that anyone is making that argument here. My own argument is that some, and maybe too many, people are making that argument implicitly, in favour of AI, by being overattracted by the AI bling and throwing the pro baby out with the bath water. They are therefore neglecting a highly valuable resource.It is not about pro versus ai.
I obviously agree with the first part of that sentence. How well they complement each other has yet to be determined, however. I personally haven't seen much in the chess world that suggests chess pros have learnt much beyond the highly esoteric variations of some openings and endgames. Early indications in the go world seem similar to me.They complement each other beautifully.
I am always suspicious of those amateurs who claim to have improved X grades by spending Y hours studying with AI, in go or in chess. My first hypothesis is usually that the improvement is due to the Y hours, not the AI, and that similar improvement could be expected if they spent Y hours using some other means of study. It's the hours that count. Of course if the bling or the availability of AI gives you the motivation to put in the Y hours, then more power to your elbow. But, unless you can prove it, don't pretend it's the machine that did it.
At the pro level, I think you can make a case that the machine did effect some improvement, but not because it was a machine. Rather it was through demonstrating possible new moves and thus erasing blind spots: moves of the type "Oh, I didn't know you could do that!" The early 3-3 invasion was an example of that. But this is exactly what happened when Shuwa started playing 4-4, when Shuho started playing on the 4th line, when Shuei emphasised the centre, when Go Seigen played the keima from 4-4 instead of ogeima, when Kitani played weird josekis, when Takagawa played early caps (just like AI's early shoulder hits), when Sakata played 3-3 in empty corners, when Takemiya went to the moon, etc etc. They weren't machines but they all had similar effects to what is going on now. And not every other pro at the respective times had much idea of how to explain why these new ideas worked. In essence, I'd say it was just a case of pros increasing their vocabulary of go (or making the tool-box bigger, or whatever analogy you prefer). The main complementary effect of AI that I see is that it confirms the acceptability of each new "word" much, much faster than the trial error of previous generations. Pros can therefore now become stronger quite quickly by having a bigger go vocabulary. They have long ago mastered the grammar of go (for most of them it was a mother tongue learnt young). They are now just expanding their range. The fact that they still don't really understand how that helps them (no more than anyone really understood what effect Shin Fuseki had) is independent of that, in just the same way that even a scholar who writes well can't really explain why Shakespeare is a better writer than J K Rowling (or if he really is).
But that's for pros. Amateurs still need to learn the grammar of go, and pros are the best and maybe the only source for that, though they do seem to believe that playing over lots of games (putting in the Y hours) and building up intuition is really the only reliable way to do that. Amateurs learning go by just learning esoteric "vocabulary" items from AI is just as daft as a beginner learning dirty or obscure words in a real language. One thing that gives Japanese people a good smirk is when a swaggering gaijin shows them he can write ringo (apple) in kanji, which not all that many Japanese can do. Of course the Japanese are "only 9-dans" at their own language and the poor gaijin wouldn't be able to tell them anything useful, such as their postilions have been struck by lightning. But it seems there are plenty of amateurs who want to be a ringo gaijin in go.
AI charts are a useful gimmick in presenting go games, of course. But those little green round things may be apples, you know
) is so large.