Joe Bebel, PhD

Vibe Mathematician. Theoretical Computer Scientist. I think a lot. R&D. Formerly @CSatUSC

[Joe Bebel, PhD @joeintheory]

🚨 Public service announcement: from now on, I will no longer call them “selfies”, I will call them “autophotographs”. Because an autophotograph is literally a selfie. Thank you for your attention. 🚨

[Joe Bebel, PhD @joeintheory]

Poor quality flour?

mill issue

[“paula” @paularambles]

hate walking in sf?

hill issue

[Joe Bebel, PhD @joeintheory]

When the dust settles, we need to understand what roles in society will be filled with AI and what we will still need human capital for, and treat the latter like the important strategic resource it is.

[Joe Bebel, PhD @joeintheory]

In the AI era, we have to adjust how we do math and science to account for these new tools, and the careers will radically change. But if we stop supporting these careers entirely, then we will lose this critical mass of people, and they will find other careers and not come back

[Joe Bebel, PhD @joeintheory]

Back in the Cold War, there was broad consensus that funding careers for pure mathematicians and scientists - even on esoteric, impractical research - was good because they are a critical strategic national resource (as shown by the Manhattan project and space race)

[Derya Unutmaz, MD @DeryaTR_]

Math, or any other field of science for that matter, does not exist, nor is it funded by citizens or taxpayers, simply to provide people with careers!

Society has generously granted us scientists with prestige, status & resources in return for our responsibility to advance knowledge, contribute to humanity & move civilization forward. If AI or super-intelligence is able to do that better than our mission has been completed or we can now change our role to help steer it as long as we are able to add value.

Unfortunately, academia has built ivory towers guarded by gatekeepers who have taken advantage of society’s generosity. Superintelligence is now challenging these gatekeepers and throwing the gates wide open. That is precisely why they are so upset.

[Joel Grus 🤠 @joelgrus]

1. A robot may not prove mathematical results or, through inaction, allow mathematical results to be proved

[Ryan Williams @rrwilliams.bsky.social @rrwilliams]

Funny, I've been calling it the mathslopcalypse. To each their own.
Can anyone honestly say they found a paper among the hundreds in the release that's *not* a huge pain to read (without AI assistance)?

[Ian Nuttall @iannuttall]

Holy shit, this guy found an exoplanet using Claude Code!

Turns out you can vibe code astronomy now…

[Joe Bebel, PhD @joeintheory]

Or maybe problems with commercial or economic value have received much more resources over the years and much of the low hanging fruit has been picked already.

[Ben Podgursky @bpodgursky]

Lucky coincidence the labs are only successfully cracking esoteric math problems and not the problems where a solution would give the winner an asymmetric commerical advantage.

[Joshua Gans @joshgans]

I wanted to screenshot this for those historians of AI and technology in the future. The day after OpenAI's mathplosion, it was nowhere to be found on the front page of the New York Times.

[Joe Bebel, PhD @joeintheory]

Wordle trolling hard today 😂

[Joe Bebel, PhD @joeintheory]

@mat4m0 I was looking for it, but I didn't have the critical part that made it work. Of course, in hindsight, the key concept is not that complex and so anyone working on the problem could have plausibly stumbled on it, but that's not the same as being close.

[Joe Bebel, PhD @joeintheory]

This includes a claimed proof and Lean formalization of the Unique Games Conjecture, which I have been personally working on for the last year and a half.

[OpenAI @OpenAI]

We’re releasing a broad range of new mathematical results produced by an internal frontier model.

We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.

https://t.co/7N6TPlft1P

[Joe Bebel, PhD @joeintheory]

@AnActualBeing Think getting scooped has always been a hazard, AI or not. Certainly there are lots of human researchers that understand UGC extremely well and any of them could have conceivably figured it out on their own too.

[Joe Bebel, PhD @joeintheory]

@ggplot2fanboy When I was a new grad student, UGC was definitely a popular topic, but it also seemed completely intractable to prove. 15 years of progress got us most of the way, and I felt like AI could help finish it (which it did - on its own)

[Joe Bebel, PhD @joeintheory]

@sandyasm In some sense, UGC being true is the "negative result" - it proves the optimality of many existing approximation algorithms. Had UGC been false, then in theory we might be able to write better approximation algorithms.

[Joe Bebel, PhD @joeintheory]

@littmath @daniel_mac8 @ChrSzegedy I think for most people who have been paying close attention, today was a "when, not if" day

[Joe Bebel, PhD @joeintheory]

AI math is real, and it is formidable

[Joe Bebel, PhD @joeintheory]

@gautamcgoel The algorithmists are going to win this one

[Joe Bebel, PhD @joeintheory]

@ElfaroukHarb I probably would have put a 90% true-10% false guess on it. 15 years ago it was much more hotly debated; then a related problem called the 2-2 conjecture was proven true, which maybe tilted it more to true (but still no consensus)

[Joe Bebel, PhD @joeintheory]

@ItsGoharr @AcerFur I don’t think the textbooks have even been updated to the n log n result

[Joe Bebel, PhD @joeintheory]

@AcerFur I was also surprised by this one. I expected improvements to matrix multiplication and possibly L=RL but not improving n log n