currently working on our PhD (imaginary issues no one cares about except a really really really specific audience)
- Tweets 7,378
- Following 264
- Followers 332,823
- Likes 3,377
how is this different from a monkey obtaining a millennium problem proof using an AI?
Two high school students, Aayush Bathija and Prince Rohatgi, with AI assistance, have reported progress on a Lorentzian polynomial problem studied by Fields Medalist June Huh. A striking example of how AI is beginning to enter mathematical research at an unusually early stage.
arxiv.org/pdf/2609.05341
19
28
735
86,806
my question now is: how is it to be a math undergrad? do you just pass the assessments to Claude and get by? is it useful for the exams? please tell me
1
21
13,866
I'm sorry you just don't decide *not* to do a math phd, the math phd decides whether it wants to be done by you or not
1
3
137
15,215
It is important that we all remain really calm about the Navier-Stokes situation.
The problem has not yet a complete proof, and the recent publication from OpenAI is still in revision (and will be for many months). Mathematics is all about human understanding. If a proof is not understandable to humans, then it cannot be considered a proof by any mathematician (can you hear a tree fall in a forest if nobody around?).
The work of understanding, expanding mathematics, enriching human knowledge is what academics need to do now with this document. AI is a tool, undeniably, but human knowledge is the objective.
Another chapter of this story concerns the authorship of the alleged result(s). For years, mathematics has largely been a collaborative field, of course clashes between egos have happened here and there as anywhere else, but mathematicians are known by being really open about recent discoveries and for celebrating original contributions (after all, important theorems are named after the mathematicians who proved them). Working together is a must, while racing against another researcher or team of researchers often ends in a bitter way. In particular, Anthropic and OpenAI do not look like they are collaborating in this. For them, this is a race, a way of pleasing stakeholders, increasing percentages, revenues, numbers... they are, fundamentally, companies. It would be good to remember this.
23
147
1,068
39,251
