Good point. But at least we know the ML algorithms are related to ML performance, we don't know that about QM.

There's no evidence that "it" is working in the first place, though.