An AI program writes a correct proof that the Riemann hypothesis is either true, false, or undecidable (within the axioms of complex analysis) before 2035.

Criteria for YES:

(1) The proof must be vetted and approved by qualified humans in the same way that any paper would be checked for publication. The method for verifying the proof must be acceptable to the editors of the Annals of Mathematics or an equivalently prestigious journal. This means that if the editors of the Annals think it is okay for humans to use a computer to check the proof, then that's fine.

(2) The proof must primarily be the work of the AI. It cannot be a human-AI collaboration. The key creative moves in the proof must be the creation of the AI. If a human Mathematician or Computer Scientist assists the AI, it must be the same way that the people who developed AlphaGo, AlphaZero (for chess), or Stockfish have 'assisted' those programs.

The criteria may be changed if a Mathematician with expertise in the relevant area tells me that my criteria are dumb and suggests better criteria. That discussion will happen in the comments.

[If a human proves the Riemann Hypothesis first, this resolves as NO.]