![](/_next/image?url=https%3A%2F%2Ffirebasestorage.googleapis.com%2Fv0%2Fb%2Fmantic-markets.appspot.com%2Fo%2Fdream%252FBb4M0n3OL9.png%3Falt%3Dmedia%26token%3Dc23de872-3c2d-44ea-a978-c1732b31d094&w=3840&q=75)
Background: P versus NP problem
"AI resolved" means a machine learning system or a human researcher constructively aided by AI/ML techniques* proves P=NP or P != NP or some exotic outcome (such as P vs NP being independent of ZFC) for the first time before the start of 2050.**
*The AI must "come up with" a key part of the proof. If a human researcher has a unique or critical insight that leads to a proof but that proof is subsequently verified or confirmed using AI plus an automated proof system, that by itself would not count. This distinction is obviously a little fuzzy and subjective, but I will do my best to hold true to the spirit of the question...
** If a human researcher resolves the problem without AI, but an AI later discovers a related or different proof, this market will still resolve NO.
Does an AI have to involved in the first proof of P vs NP, or would it also resolve "yes" if the first proof is created by a human, and later, a different researcher comes along with a new proof using AI? My interpretation is that it should resolve "no" in the latter case, since the P vs NP problem would have been resolved by humans, and AI just provided an additional proof rather than the resolution to the problem.
@JosephNoonan Yes, my intention was for it to resolve no in that case. I can see how my wording is ambiguous; I'll modify the description...