It’s a well known problem in higher mathematics that even if you’ve solved a problem, often the proofs are incredibly long and complex and require an extensive amount of time spent by peers to review it.
It would be great if someone could explain to me how AI improves this situation. Even if AI thinks it’s solved a problem, unless the proof is incredibly efficient and well explained, it will be difficult to verify the correctness. One hallucination in 300 steps of logic is enough to destroy the entire proof.
It would be great if someone could explain to me how AI improves this situation. Even if AI thinks it’s solved a problem, unless the proof is incredibly efficient and well explained, it will be difficult to verify the correctness. One hallucination in 300 steps of logic is enough to destroy the entire proof.