michael_nielsen

I think this work is quite irresponsible

Incidentally, I certainly don't disagree - at all! - that this kind of thing will be hugely problematic.

the key problem is that it produces realistic outputs that are mostly driven by priors, not data

(I'm having the problem you allude to: don't have enough data to unpack your argument into what you mean!)

it’s just a compelling fiction. You can’t make data from nothing