The headline is interesting, and adversarial attacks on #AI are a real thing, but this looks like a cheap trick:

https://arstechnica.com/information-technology/2022/11/new-go-playing-trick-defeats-world-class-go-ai-but-loses-to-human-amateurs/

It depends on KataGo not even looking ahead two moves to the reply to its pass. It almost seems like a rules dispute, where KataGo is assuming (as is the case in normal human games) that dead stones will be removed at the end of the game. If the adversary is claiming those lone black stones are alive, some sort of dispute resolution system should kick in.

#go #igo #weiqi #MachineLearning