It wasn’t me!

I have a gentleman’s bet with another number theorist that I will be able to write a nonsense paper and get it published in respectable journal. To be honest, I barely have enough time to write actual papers let alone a nonsense paper, but it has crossed my mind from time to time.

The respective mathematician recently inquired as to whether I was the author of a certain preprint, presumably written in an attempt to collect on our bet; I just want to say that it wasn’t me! First of all, I wouldn’t try to claim a major conjecture, the aim would be to write a paper that zero people actually read (including the referee). My impression from a quick browse is that the preprint in question is AI generated nonsense. Apart from the irritation of having AI generated nonsense on the arXiv, it is not very interesting, and so I won’t say too much more about it.

It actually points to something about AI and mathematics that sticks out to me like a sore thumb whenever I try it out. Namely, AI does a remarkably good job of phrasing things with the full confidence of someone who knows what they are talking about, even when (which is mostly the case) they are saying utter nonsense. In other words, it does a great job of capturing the tone but not any of the actual structure of mathematical discourse. I hope that, at least, is something that will change as AI improves! Maybe also that’s why this fake paper is annoying; as a community, we give each other the benefit of the doubt that we are trying our best both to give correct proofs and honest exposition. That makes it especially annoying when someone deliberately tries to exploit or undermine that generosity. When the work generated by AI remains this obviously stupid it is less of a problem. But if AI improves AND people try to use it maliciously to publish fake proofs, that could potentially be a problem.