AI in Mathematics: Hype, Rediscovery, and Real Progress

I’m back with my friend and colleague Sabine Hossenfelder for another episode of “What’s New in Science”. I think this is one of my favorite dialogues that we have had. Spending time with Sabine was a nice chance to step away from my physics lecture series for a bit. I know many of you have been enjoying the lectures, so don’t worry, they’ll be back soon.

In this episode, we covered the kind of science news I like best: ideas you can argue about and results that make you recalibrate. Sabine opened with describing a clever proposal that future fusion reactors might double as axion dark matter factories, producing a flux of very light, weakly interacting particles through neutron-lithium reactions in the shielding. That led to a discussion about what people mean by “axions,” why particle physicists tend to be more particular about the term, and why I’m always more interested in dark matter candidates that were invented to solve an actual problem, not just to fill a cosmological gap. From there we jumped to quantum mechanics at the edge of common sense, with a Vienna experiment showing interference from a cluster of thousands of atoms, and a friendly disagreement about whether “collapse” is a real physical process or just the wrong way to talk about what quantum mechanics is doing.

We also talked about AI and math, including the recent swirl of claims about machines proving famous open problems, what was hype, what was rediscovery, and what might genuinely be changing in how mathematicians search the landscape. Then we went from equations to extinction, with a fascinating new approach using space dust and helium isotopes to argue that life may have started rebounding after the Chicxulub impact far faster than people had assumed. Sabine brought a surprising example of string theory mathematics finding a practical use in modeling biological networks, and we ended with biology proper in two very different moods: a sobering study in mice suggesting lung tumors can hijack vagus nerve signaling to suppress local immune responses, and then a lighter result about dogs learning words from overheard human conversation at roughly toddler level. My dog Levi, who many of you have seen on the podcast, was asleep next to me while we talked about it, which felt like the right way to end.

As always, thank you for your continued support, and I hope the changing of seasons brings you good time with friends and family.

As always, an ad-free video version of this podcast is also available to paid Critical Mass subscribers. Your subscriptions support the non-profit Origins Project Foundation, which produces the podcast. The audio version is available free on the Critical Mass site and on all podcast sites, and the video version will also be available on the Origins Project YouTube.