Human level far below the optimum

[hyperpape]

I don't understand the point of you quoting me. Anyway, in go as long as neither is playing perfectly the winning % will most likely not be 100. In chess one needs to be very far away from perfect play to not get even one draw in a thousand games against perfect play.

My fault that I'm not explain clearly.
To be direct, do you have any source that conclude the Elo differece between chess AI and perfect play is between 400-800?
Maybe there's a method that can estimated that number even when the game itself is not solved yet.

I recall reading that you can infer the ELO difference from the increasing drawishness of top level play (maybe in something Ken Regan wrote). The idea is something to the effect that we are making progress in bots, but the dominant effect is that top bots are just drawing against each other more and more (already an issue in grandmaster play). So an improvement might be that you still only win one in 10 games or whatever, but you reduce the number of times the opponent wins even closer to zero. I'm afraid I can't really do justice to the argument myself--you'd have to find the primary source.

[hyperpape]

Let's be concrete: the old anecdote was that a top pro said something like "I can play God on three stones." That's more than a 400 ELO difference at the level of top professionals. So the question is whether these new developments suggest that we're more than that far from optimal play.

[Bonobo]

Let's be concrete: the old anecdote was that a top pro said something like "I can play God on three stones."

Side note: I think it was Otake Hideo who allegedly said that.

[zermelo]

Maybe the disagreement is about how to quantify the distance from perfect play. In go, traditionally skill differences are measured in stones and points (of komi) not in ELO. Getting close to perfect play it seems plausible that elos-points ratio grows very high. I can imagine that some player is > 1000 elo from perfect play but 'god' still cannot give the player a one stone handicap. Is the player now close or far from perfect play? I would call it 'close' myself.

[Gomoto]

Now we're talking.

[dankenzon]

It can be answered in resource aspect. IBM immediately abandon computer chess right after it won against Kasparov in 1997. While DeepMind increasing AlphaGo team members even after Lee Sedol match. If DeepMind doesn't publish any paper, and retire AlphaGo immediately in March, it gonna takes more time for sure.

Deepmind did publish their paper explaining the approach they did with AlphaGo.

The whole idea has been immediately added/plugged to several bots. And as far as there's more memory and chance for they to practice more, they will be stronger.

The real question is:
is it an optimal/"non plus ultra" level of Go?

[dankenzon]

In conclusion, computer go is nowhere close to optimal play.

I do not agree.

How could you not agree? Chess softwares are 400-800 elos away from perfect play. Go is a much more complicated game.
The reason why humans could last so long in chess is just due to draws being an option. Even now one might get one.

I think all the games will remain: because at the end even the stronger more computer like human who learns from a computer, may learn their ideas and approach, but will never handle the detail of the game. See Carlsen, Caruana and others: this top players are quite precise and have a way to play different from all the previous schools, but still make mistakes.