Interesting thing that was shared on Facebook:
And I wonder whether such a thing might actually really help learning better judgement?
More information:
Quote:
Not sure though whether I’d really want a heartbeat feedback … I might suffer from a heart attack 
I mean, I begin to transpirate heavily already when the stones begin to fall faster in Tetris …
Longer video:
Sounds cool! Obviously shouldn’t be used except by prior agreement with opponent, and then only in unranked, but still cool
It looks very interesting but unfortunately there’s a lot of situations where this would just be cheating, and it’s of course a worry that that would be its primary use.
It would at the very least make sure you pay attention at crucial moments, and might even give hints on when to tenuki, and of course it’s telling you whether you’re winning or not the whole time which is not a small knowledge advantage to have.
Yeah, I’m definitely in the camp of “Don’t ban something because it can be used to cheat: ban cheating with it”
The AI estimate may be misleading. The player who is ahead according to AI may lose the game after getting killed, whereas a superhuman AI wouldn’t have difficulty making life.
I think there’s a tradeoff between how easy it might be to cheat with it and get away with it vs how it’s generally used and how useful the thing is.
Like for instance if you made a browser extension for OGS to overlay katago’s blue moves on the OGS board, or add a katago score estimate or winrate bar to the display - there might be ways to detect it, the same way you might be able to come up with ways to detect this heartbeat stuff, but it certainly makes cheating while incorporating your own ideas easier.
Essentially if you make a tool where it’s
(a) hard to detect (or it takes some work to figure out how to detect)
(b) improves your strength a bit but not too much
then it starts being the case that it’s better to play with the tool than without it, because that’s possibly what other people will do, at least for those tempted.
However the article the video author links suggests they may be interested in having this as a tool for it’s own separate matchmaking mode, and maybe in that case it’s fine, as opposed to it being available like katrain, katago etc.
Well the thing is if katago is going to go crazy when there’s exactly one move on the board that has to be played, like you can see it here at 4:28 say
in the 3-3 invasion double hane variation - you pretty much can’t allow the big ponnuki, you have to connect in most situations. The same will be true when there’s one way to live with a group that’s about to be killed, and then you know not to tenuki and to start studying the tsumego deeply until you find an answer - which is a big hint I would think.
When I review my games, I look for my mistakes. If OGS had an accurate (4d) score estimate that I could use while trying out lines of play, it sure would help me learn! The current score estimate comes with lots of warnings about how it is an approximation.
The score estimation is strong (9d) when reviewing one of your finished game. Weak only when you use it while playing.
Sure, same for me—but such things would ONL>Y be fair if both players used them. AND I’m not sure such aides should be used in rated games.
So … I think, if this should ever be implemented here, games with this feature should …
Maybe they should even have their own category, next to “rated” and “unrated”:
“Centaur Games” or “Four-Brain (Six-Brain?) Games” or something … but wait, I think this is already used for games where both sides use (agreed-upon) AI help.
Would any such features encourage cheaters?
Considering only the usefulness vs progressing (and not the TOS).
I am not so sure on this myself. I mean what counts is acquiring something and separating the study from the play looks still important for me. I have to engage myself in my moves fully and more as consulting a catalog of suggestions.
Sorry, I used imprecise terminology. I mean when using Analyze after a game has been played, to experiment with alternative moves.
I would never use an AI to help me when I’m playing a real game, rated or not, even if it were easy to do! A real game should be played honestly.
And when using Estimate Score during review or analysis of a completed game you get a strong AI estimator (9dan+).
The intentionally weak estimator is only used in an ongoing game.
With the exception of variants, like for the Alan Turing Nines tournament, I agree
Well then it’s a strong AI (9D)