🏅 Top traders

Very unlucky. https://deepmind.google/discover/blog/alphageometry-an-olympiad-level-ai-system-for-geometry/ tests on post-2000 problems and solved 24 out of 30.
Given that a) 1959-1999 IMOs are generally easier than the newer ones; b) they do not train on any existing data at all; there is very little doubt this would have solved 50+ if they had just bothered to formalize the 40+ eligible problems from earlier IMOs.

@dp Someone could run the code now, but it won't change the resolution of this market.

I apologize - I forgot that I said in the comments I would wait ~1 month to resolve. Normally I do not bother waiting, especially for a market like this one with very little uncertainty, but I said I would so I'm unresolving and I'll see you in a month.

@vluzko Ping

Does it have to be published, or does it only have to be verified that it happened before 2024? E.g. if I now try some new idea, it somehow manages to solve 50 IMO problems, I submit it to arXiv on 31 Dec, but it only appears on arXiv in 2024 -- does it resolve YES or NO?

@dp That resolves YES, but I will only wait ~1 month to see if anything comes out.

@vluzko will the AI be required to examine the natural language statement of the problems, or does an AI that works on pre-formalized problems count?

@BoltonBailey For this question it is from natural language descriptions.

@vluzko Wait, no, sorry, confused this question with a different one. For this one I'm using the same criteria as for HTPS, so pre-formalized problems are fine.

A Tarski-Seidenberg tactic could do this (at least in principle). Not sure anyone wants to put in the work to create that though.

Will any AI solve >=50 IMO problems by the start of 2024?, 8k, beautiful, illustration, trending on art station, picture of the day, epic composition