RE: https://mathstodon.xyz/@tao/117185702292629986

Whether one is anti-AI, pro-AI, or "it's complicated", Terry makes a critical point here that is generally missed.

Related: For some it's too obvious to mention, but nonetheless not generally known: generating *new* conjectures, including creating whole new subfields to explore, is *really* important, and humans are good at it, but AI largely can't do it.

["not generally known" because the average e.g. journalist doesn't understand that, or its importance, often even when explained]