Recent searches
Search options
The first $1.048 million progress prize for the Artificial Intelligence Mathematical Olympiad #AIMO Prize has now launched! https://www.kaggle.com/competitions/ai-mathematical-olympiad-prize/overview The challenge is to submit (by June 27) an #AI model that can perform well on a set of 50 test questions, each of which has an answer that is a 3 digit number. (The benchmark model that we tested on was only able to correctly answer 3 of the 50 questions; prize winners will have to beat this benchmark as a minimum requirement.)
@tao this is the kind of stuff that bothers me. when will the ai prize be an algorithm that can successfully mimic a living earthworm? or better a Stentor?
this is too narrow a def of intelligence.
@barrygoldman1 The objective of this prize is not to develop any sort of "artificial general intelligence", but to address the much narrower (but still, in my opinion, interesting) question of quantifying the extent to which current AI tools can actually be used to attack non-trivial mathematics problems (if not research level, at least at the level of olympiad problems).
I am sure there will be no shortage of other AI prizes offered by other organizations that will explore other aspects of artificial intelligence, but that would be beyond the scope of this current prize.
Incidentally, on the current leaderboard https://www.kaggle.com/competitions/ai-mathematical-olympiad-prize/leaderboard , it appears that someone has already beaten the benchmark 3/50 result to obtain a 4/50 score on the public test data set. Not bad for a single day of competition...
@tao well that's fair then. i wish i could find an explanation of the mathematical properties of current 'ai' tools. they seem to be treated as black boxes. not enough controlled experiments on input and output!