Sanmi Koyejo raises the interesting point at #AAS248 that the AI will be built no matter what, and that one can engage with what the standards of good science are so that AI is evaluated by that, otherwise non-experts may set poorer standards. And that wouldn't be the same as endorsement of AI