I think criticizing the argument. It's so common for a non-positive logical conclusion to be sort of a half hearted reductio ad absurdum (disproving by showing

I think criticizing the argument. It's so common for a non-positive logical conclusion to be sort of a half hearted reductio ad absurdum (disproving by showing the consequence is absurd) that many people mistake this for happening when it's not. So, they argue against the sub-text that may or may not actually be there. Or, maybe they're just copying the joke. (The joke aspect being that even when saying how AI won't dominate everything, he's referring to human works as "training data", so it's at least dominating the framing.)