Solving a FrontierMath hypergraph problem cost under $100,000.
It would be great to know how much it currently costs per major mathematical research result from AI. People keep using the API cost for a successful try, but forget to add the API costs for all the unsuccessful ones! By that logic, it only takes me one arrow to shoot a bullseye.
No Digg Deeper questions have been answered for this story yet.
Obviously these things will get cheaper, but it is very useful to understand the current costs.
At the moment, I think the current cost per big result that the mathematics community cares about is greater than a lifetime's salary for a top mathematician.
At the moment, I think the current cost per big result that the mathematics community cares about is greater than a lifetime's salary for a top mathematician.
My best guess is that leading labs spent >$100,000 in inference to get their IMO gold medals. Last week's result on the Unit Distance Problem is the first really meaningful autonomous maths research result—but that means the current cost for 1 of those is the entire spend so far.