+1 to chess being a less intensive game to compute than Go. But beyond that, Lichess has a network of contributors who donate compute via fishnet
One could imagine this happening for OGS too. But I don’t see it happening any time soon.
the subscription model is working fine
the backend is closed so integrating a feature like this would need help from one of the few who have access to the server code (though someone could build KataNet fully outside of OGS and integrate it later)
the Go community is smaller than the Chess community, so there are just fewer people who can build cool features like this
There are a lot of desktop tools that make it relatively easy to do AI analysis locally with KataGo, Leela, etc.
If you try any of those, you will also see very quickly how slow analysis takes on your consumer GPU, as hard evidence why this is a computationally harder problem than local chess analysis.
Or potentially just enabling the interactive analysis without having to trigger a full game review could be another approach also.
If you consider that (as mentioned above) that you need a full game review run to start triggering the interactive analysis or one might need to start a game review either.
@Uberdude@benjito@lazerwalker thanks for your input. However, my question was purely around user experience and not related to computational complexity at all.
I’m still wondering whether I’m just missing a simple way to get AI analysis for a specific position for games already available on OGS.
(Regarding computational complexity it would be interesting to know by what factor Go is harder than chess. You could compare the power needed to get to top-human level in chess and in Go.
Would it take twice as much power for Go? Or 10 times more? Or 100 times?)
A combination of the suggestions of @GreenAsJade and @shinuito is what I am looking for: A way for instantly switching on AI support at a specific board position when looking at a game.
I don’t understand your complex process. Isn’t the AI analysis instantly available as soon as the game is finished? Maybe the AI is switched off at your account?
Computational complexity influences what’s possible. Two of your complaints are limited by compute:
As @GreenAsJade pointed out, OGS can add the button to add AI review for any game (great idea!), but it will still take some time to generate, and someone needs to pay for the compute.
Analyzing a single move is cheaper than analyzing an entire game, of course.
Your original post frames the question in comparison to Lichess: “Lichess offers fast and free AI review, why don’t we?”. That’s why so many folks compared the computational complexity of Go vs. Chess.
A more constructive conversation (and I think this is the direction this thread is headed) is had when we look at the Go tools themselves and ask how OGS can better utilize them.
It also occurs to me that your flow can be improved without any added features.
I’ll also add that if one really does want to add a game to their library there is no need to download the sgf. On the sidebar there is an option to “add to library”.
This doesn’t work for me unfortunately. I do not get AI support for the main line after starting a review.
However, this could be a workaround: If I pass twice from a position on the main line, I do get AI support from there!
Unfortunately, while this is adding the game to your library, AI review doesn’t start if you open the game from your library. Even if you deviate from the main line, AI doesn’t kick in.
