Arnab Bhattacharyya

Addicted to chasing mysteries

[Arnab Bhattacharyya @abhatt2]

Two weeks ago, we brought together some of the most exciting young researchers doing work at the foundations of AI & ML. The event was amazing. Thanks to all the participants and attendees. Already looking forward to next year!

[Arnab Bhattacharyya @abhatt2]

We are organizing FAIRS: Foundations of AI Rising Stars (https://t.co/RwZuii0IMl). If you are a PhD student or postdoc working on foundational aspects of AI and is/will soon be on the job market, please apply. Deadline is 15th July.

[Arnab Bhattacharyya @abhatt2]

@letonyo No, I'd bet this happens every year.

[Joseph Conlon @JosephPConlon]

Some thoughts on AI, theoretical physics, careers and the question of whether this is a good time to do a PhD in theoretical physics (which I will just call ‘theory’ for shorthand). Stimulated by a conversation with one of my undergrads. 1/9

[Arnab Bhattacharyya @abhatt2]

@aryehazan Congrats! This is fantastic!

[Arnab Bhattacharyya @abhatt2]

@ameya_pa My understanding is that ETH being false is not earth-shattering news, but the fact that we now have an argument to show this is.

[Arnab Bhattacharyya @abhatt2]

@thesasho @ksmeel Honestly, it's disappointing that someone like Ben would write such a click-baity post....

[Arnab Bhattacharyya @abhatt2]

@thesasho @ksmeel @ksmeel I agree with Sasho here. It is of course important now to try to improve the algorithm as much as possible. But the point is that up until now, we didn't know whether it is possible to beat exponential time even in principle. And now we do!

[Arnab Bhattacharyya @abhatt2]

@FrankT56966988 Much less

[Arnab Bhattacharyya @abhatt2]

For a theorist, a good day in the office used to be proving an interesting lemma or two. Now, a good day in the office is solving an interesting open problem or two.

[Arnab Bhattacharyya @abhatt2]

@jasondeanlee Got mine approved just today!

[Bernardo Subercaseaux 🐒 @b_subercaseaux]

We have finally found a solution to a geometric question that has puzzled us for years!

Is it possible to have a set of points in the plane that is *k-everywhere-unbalanced*, meaning that, for any line through two of the points, the number of points on each side of the line differs by at least k?

It was originally conjectured by Kupitz (1979) that this is only possible for k = 1, meaning that all pointsets would be roughly balanced under some perspective. But in 2002, Alon found a beautiful construction on 12 points that is 2-everywhere-unbalanced, in the GIF below.

The obvious question is whether k=2 is the limit. And despite amazing theoretical work by Conlon and Lim (2025) showing that this is possible if we allow lines to bend, nothing was known for k >= 3 with straight lines... Until now!

[Arnab Bhattacharyya @abhatt2]

@PI010101 https://t.co/ePfGp9rlnq

[Arnab Bhattacharyya @abhatt2]

We should give a name to that feeling you get when your AI is not spinning its wheels on some open problem.

[Arnab Bhattacharyya @abhatt2]

@rahulsanthanam @rrwilliams True

[Arnab Bhattacharyya @abhatt2]

@rahulsanthanam @rrwilliams "I see no value at all in accelerating the process in this way." I understand your general point, but regarding acceleration, math research used to be a race until a short while ago. Tao is equally guilty of acceleration in that sense!

[Abhishek Saha @ObhishekSaha]

When the proof of Fermat's last theorem was announced by Wiles in June 1993, newspapers around the world covered it extensively. I was a ten-year-old boy in Calcutta, India, and the story was splashed across my local newspaper. A few months later, I went to a bookshop with my grandfather and bought two popular books that tried to explain what all the excitement was about.

Yesterday was the single greatest day of mathematical breakthroughs in history. This is not an exaggeration.

In terms the collective significance and depth of the results released, this was at least ten times, perhaps even a hundred times, as momentous as the proof of Fermat's Last Theorem.

So where is the media? Where are the headlines, the breathless newsrooms, the door to door conversations? Where is the sense of incredible wonder this moment deserves?

[Arnab Bhattacharyya @abhatt2]

Integer multiplication!

[Arnab Bhattacharyya @abhatt2]

Sidorenko's conjecture is false!

[Arnab Bhattacharyya @abhatt2]

Poly length 3-query LDCs are still open!

[ELLIS @ELLISforEurope]

🚨 Apply now for the #ELLISPhD Program!

Work with leading AI/ML researchers across Europe, with exchange visits & joint supervision. The application deadline is 31 Oct.

Get all info & apply now at https://t.co/VEawFTFALc

[Arnab Bhattacharyya @abhatt2]

I'm searching for PhD students to join me in research on foundational questions in AI & ML: https://t.co/7VXUxBNbkZ. Deadline for the first round is 30th October.