Regular game of go in which each player receives multiple AI-provided options for the next move.
Potential variations within this:
Choose only from AI moves, or allow anything (I think I prefer the latter, but both are interesting)
Vary number of moves
Vary quality of AI moves – are they the top three suggestions? Two good moves and one not so great? Lot of things one could play with here.
Rather than doing this for every turn, offer each player a limited number of “phone a friend” opportunities to get an AI suggestion (or several) on a particular move.
Apart from being an amusing change of pace, I could see this being useful for learning. It’s similar problems I’ve seen (e.g. in old Go World magazines) where you’re given three different moves and must choose the one that a pro really played.
There is a global policy to not use AI during games (Even unrated) on OGS.
There are very few exceptions like the Alan Turing tournament.
The idea of using AI is still sometimes in the air, as it appears in a few threads.
In my opinion, it would be best to keep it very clearly separated from the main site, with clear announcement if this kind of help is further introduced.
This seems to me less like a go variant and more of a “change of meta rules”. At least for me, a variant involves some change of the actual game rules, and the agreement to not use AI is not technically a rule of go. Your idea seems similar to the Alan Turing tournament, where AI assistance is explicitly allowed. That being said, as long as you and your opponent agree to play this kind of game, I’d say go nuts. If you want to play this on OGS, maybe you should tell a moderator beforehand.
I think it’s important to appreciate that the options in those kind of problems were chosen by hand, and with the intention of creating a good learning experience for the reader. AI is not programmed for that, and it will not explain its choice to you.
Board is separated on 9 parts. Human players only choose on which part to play. Bot plays inside selected part, it may be hard to predict where exactly. So tactic is replaced with AI, but human that better understand strategy wins.
I think that if the players are actually constrained to only choose one of the AI moves, you could argue that’s a legitimate variant of the game rather than a mere change to meta rules. One can quibble about semantics, but the playing experience would be very different, which seems like an important thing.
Your point about this not being the same as problems designed by hand with pedagogy in mind, though. This would certainly be different than that. Of course, in the future, we may see AI that’s better able to explain the principles behind its play, which would open up stronger possibilities in that direction.
When topics are started, rules written in the TOS, … It seems it’s not that obvious, and thinking how to keep boundaries clear not a superfluous matter.
I doubt. For some moves, yes (if you have some background). For others not at all.
If you lack the tools to go search why a AI move is better as your move, you will end having losing your time and even motivation to progress.
Another thing is the slippery side where the “guess” is replaced by the “copy” in the teaching process.
The fun in it is like riding a powerful motorbike when you know only about bicycle, it’s a bit limited.
I get what you’re saying, but how different is this from playing through an professional game, without commentary, for study? I do plenty of this and find it quite helpful. It immerses me in the feel of games played at a higher level, and I can put as much effort as I care to into trying to understand the purpose of each move. Similarly if I’m choosing among several AI-suggested moves, I’ve still got to ask myself what each move means in order to establish preferences among them.
I don’t see how the difference is relevant here enough to diminish the point at hand. Both AIs and pros can play at a high level, and without commentary lower-level players are in a position to do their best to understand the meaning of their moves, and try to learn from that.
In principle you could do the same variant with pro-suggested moves rather than AIs, though of course that’d be completely impractical.