OpenAI solves math and we're unhappy about how they did it

AI companies are out there all the time telling us how they're going to cure all cancers, solve the world's hardest science problems and find new answers to old questions.

We don't believe them, ridicule them, the press writes stories about how these tools can't even count the r's in strawberry.

And then OpenAI use an unreleased model to solve a bunch of math problems. Difficult, complex, old math problems. The kind with prizes attached to them, such is the importance that the math community has attached to them.

And rather than celebrate it, rather than report on it as good news that the AI labs are doing the thing they said they'd do - we're instead treating it like a bad thing.

Mathematicians who spent decades thinking about these problems don't know what to do now. We're upset about how they might have built on top of the work of other math nerds, or about how they might have attained that information. We're upset about how they presented it, because it didn't necessarily fit the framework that math aficionados want to use when sharing proofs. Or maybe we're upset that they solved 100 problems, but didn't solve the one we're particularly interested in. Or we think they cheated because they had an unlimited token budget.

What a bunch of spoiled little bastards we all are.