Every answer immediately resolves YES if AI wins IMO gold, i.e. once there is a plausible demonstration that there exists an AI system which can get a gold medal on that year's International Mathematical Olympiad, and it is very likely that the year's IMO wasn't in the training data. Also resolves YES on a consensus that a certain AI system could have achieved an IMO gold (e.g., it solves a Millennium problem), but the year's IMO happens to be in the training dataset.
Else, every year on 1 Jan or shortly thereafter, starting with 1 Jan 2025, I will use a public source of randomness to independently roll for a possible NO resolution of each answer. (A source which allows one to specify in advance the locations where the random numbers will in the future be written, so that my shenanigans are ~impossible conditional on me not actually controlling that source of randomness.)
I will bet. In the event of a whatsoever contentious resolution I'll defer resolution to a trusted group of Manifold users, such as a poll of moderators.
@FlorisvanDoorn Yeah, I was planning to do that, though if I do FairlyRandom then I guess I do have that advantage. Absolutely not my intention to profit from betting NO after the roll though, so I hereby pledge not to buy any NO after rolling it, and in any case not to buy any shares for at least 24 hours after the roll. Thanks for bringing it up. :)
Interesting market.
I just made a similar market inspired by this one: https://manifold.markets/FlorisvanDoorn/what-is-10log-of-lowest-number-roll?r=RmxvcmlzdmFuRG9vcm4
@Lovre i recommend using https://manifold.markets/FairlyRandom, the canonical source of randomness for resolution-relevant decisions on Manifold