The Mathocalypse
- Source
- scottaaronson.blog
- Author
- 6bitquant
- Date

Context
For Scott Aaronson, the hard part of OpenAI's mathematics release is not only deciding whether each proof is correct. The essay argues that the results are worth most if people come to understand the ideas well enough to explain and reuse them, and it says the race to do that has only just started. Many of the AI-written papers are described as hard to read, and some results come with machine-checkable Lean certificates while others do not.
The essay then contrasts two release practices. OpenAI posted the proofs publicly without explanation, which sets off a competitive scramble among humans to decipher them. In a different case involving Anthropic, the human authors wrote and announced an explained version of a result. The essay says it does not settle which approach is better and asks readers for their view. It also quotes the advisory group's statement that mathematicians must be able to formulate their own questions, and it separately rejects the idea that no one should use AI on open problems.
Checking sign-in…
Loading comments…
