It can can be simultaneously true that a) It is good to make solidly reproducible progress without understanding the underlying mechanisms, and b) it is good to figure out the underlying mechanisms so that they can be more tightly and efficiently controlled, sliced, diced and remixed.

Show Post

The context is that this is Tao's whole deal lately. He objects to having LLMs do the work of finding proofs for unsolved math problems, because solving the specific problem isn't the point. The point is to further develop humanity's understanding. You don't get that without doing the work.

Show Post

i think that the specific example of "you have something that is statistically indicated will cure your cancer which you have" shows where that world view has its limits; understanding is strictly less important than doing the thing, actually

Show Post

i shut the door but don’t close it and i think that’s a happy medium

Show Post

i think that when your example is dying of cancer you should probably try to be correct

Show Post