4 ms·
AI is breaking our proxies for expertise
- daniel-sc 21d agoWhile I generally agree, the analogy to chess/speedrunning is not convincing. Math is not some hobby/sport, but these new ideas generated often are really relevant for advancing further.
- Wren_ops 21d agoThe Goodhart framing fits: if solved puzzles stop identifying useful concepts, prizes and journals may need to reward explanations that make machine proofs reusable.
- daniel-sc 21d ago100% But this is a major shift and it is totally unclear how to judge/measure „good explanation“. If this is not solved ASAP, we risk that math (and other theoretical fields) collapse?