Nagesh Adluru (@NAdluru) profile picture

Suffering from constant curiosity and absent certainty but insufferable optimism.

Joined November 2014
yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model. We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann algebras (disproof of Connes' Rigidity Conjecture) to better bounds for high dimensional sphere packing, for circuit complexity, for monochromatic triangles in multicolored graphs, and more. More thoughts here: openai.com/index/ten-advance…
276
939
538
6,629
4,294,054
Very interesting development!!...
Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better. This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.
3
4
54
4,491
AI has now solved a major open problem -- one of the best known Erdos problems called the unit distance problem, one of Erdos's favourite questions and one that many mathematicians had tried. openai.com/index/model-dispr…
72
601
239
3,509
1,513,647
Replying to @ScienceMagazine
We need to confront an uncomfortable truth about how scientists are trained and rewarded. I cannot count the number of PhD students and postdocs I have seen, often men but not exclusively, who advanced without truly knowing how to run experiments themselves. Technicians were hired to do the actual lab work, while they stayed removed from the bench. When experiments failed, they had no idea why, because they had never learned how to troubleshoot, adapt, or sit with failure long enough to understand it. Science cannot be learned secondhand. You do not understand an experiment unless you have failed with it yourself. Yet many of these individuals were highly articulate. They talked well about science, read papers, held meetings, and cultivated visibility. Some supervisors enabled this, others rewarded it. As a result, people who were strong communicators but weak experimentalists moved up the ladder, protected by hierarchy rather than evidence. Meanwhile, others worked relentlessly at the bench, shouldering the real labor, burning out, and often remaining invisible. This creates a deeply imbalanced system. We are producing scientists who can speak confidently but lack mechanistic understanding. That is dangerous, especially in fields like molecular biology, where reductionist thinking dominates. Studying a single pathway in a petri dish after disease onset will not solve complex, systemic diseases that begin decades earlier. Biology does not work in isolation, and neither should our training. These realities are rarely taught or even acknowledged in labs. Failure is hidden. Systems thinking is sidelined. Biases, including gender and power biases, remain pervasive. Science was stronger when rigor, depth, and hands-on mastery mattered more than optics. Change will not come from protecting students from failure, but from teaching them how to engage with it honestly, early, and fully. If we want better science, we need better training, better incentives, and far more integrity in how we define “success.”
3
4
11
465
Nagesh Adluru retweeted
I enjoyed this conversation with an inspiring high school teacher in rural Pennsylvania and his award-winning student, now an NSF postdoc. We discussed math education, the role of mentors, and much more. youtu.be/DhdEZl64VqY
2
11
47
9,788
We’re announcing a research collaboration with @CFS_energy, one of the world’s leading nuclear fusion companies. Together, we’re helping speed up the development of clean, safe, limitless fusion power with AI. ⚛️
87
583
182
3,904
1,607,405
Nagesh Adluru retweeted
🔥 USC Students: Math-Focused Research in Generative AI 🔥 We're looking for USC graduate and undergraduate students* with a strong background in mathematics, physics or CS to work on generative and predictive AI models, vision-language models, and bridge models for medical imaging. Experience in computer vision, stochastic calculus, or continuum mechanics is a plus. Please DM me if interested. 🔥 We're seeking exceptional students in math, physics, or CS to work on the mathematical foundations of generative AI, vision-language models, and medical imaging. 🔥 Current topics include: 🌀 Score-based diffusion, flow matching, and Schrödinger bridges, applied to structural, diffusion and functional MRI 🔁 Invertible models via optimal transport and SDEs; privacy and memorisation in generative models 📚 Vision-language models with links to RKHS theory, Stein discrepancy, subordination, and representation learning 🔥 Ideal background: 🔥 Stochastic calculus, PDEs, variational methods, computer vision 🔥 Interest in functional analysis, information geometry, Malliavin calculus, deep learning theory, NLP 🔥 We are building models grounded in rigorous math; how they work, fail, and how to improve them. 📩 DM me if you're interested! 🔥 Lab video youtube.com/watch?v=fOORfzGj… 🔥 Shorter video featuring VLMs youtube.com/watch?v=o0Stq2GL… 🔥 CS version youtube.com/watch?v=nCfaZepD… 🔥 More stochastic PDEs youtube.com/watch?v=uh1Jw3eH…
9
20
2,295
Nagesh Adluru retweeted
New laptop desk to maximize procrastination
1
2
34
1,401
Nagesh Adluru retweeted
New way of raising money for the lab :)
3
3
111
3,290
Nagesh Adluru retweeted
Richard Feynman on studying and talent.
95
1,396
180
9,552
1,118,905
Most academic writing is *optimized* to be bad — good writing is all about making ideas as simple as possible but career incentives push authors to make their ideas seem more profound and complicated than they are.
LLMs are 5/10 writers. So the fact that you can reliably improve on the explanations in papers and books by asking an LLM to summarize them is a huge condemnation of academic writing.
33
66
8
622
65,052
Science cannot thrive if, for whatever reasons, its practitioners have to play it safe. It’s telling that eccentrics and mavericks were tolerated far more in Victorian England and Jeffersonian America (Stubbins Ffirth, anyone?) than in the halls of 21st century American science.
2
2
28
2,561
Nagesh Adluru retweeted
There may not be a Highlights party this year, but the Highlights magazine marches on Congratulations and thank you to Maria Eugenia Caligiuri, Mathieu Boudreau and the entire @mrm_highlights team for keeping the flame alive for ten whole years!
Excited to announce the release of the 2025 edition (Volume 10) of the MRM Highlights magazine, with an introduction by Highlights editor @MaryJCaligiuri. Available in print at the ISMRM conference. blog.ismrm.org/2025/05/11/mr… @ISMRM #ISMRM #ISMRM25
1
1
5
526
Nagesh Adluru retweeted
Out now in MAGMA, our contribution to the debate about harmonization vs. standardization of MRI measurements Thank you to @agahkarakuzu @BlosteinNadia, Alex Valcourt Caron, Arnaud Boré, @FrancoisRheault and @maxdescoteaux for translating our vision into words and pictures
Containers, vendor-neutrality, data standards… These concepts are gaining traction in MRI research. But how can we create interoperable and modular measurement workflows in search of reliable quantitative imaging biomarkers? Out in MAGMA #ISMRM2025 link.springer.com/article/10…
3
7
685
I think that part of the discord here is due to different meanings of the word "speed"; one meaning is "talent" (that is, rate of acquisition of skill over unit of effort) and the other is "fluency" (having acquired a level of skill, through deliberate practice, so that execution becomes effortless). These may superficially sound closely related but they're completely different. Talent, to the best of my understanding, is in large part determined by one's genetics, and in large part by one's environment. There's *a lot* we can and should do about the latter to make sure everyone realizes their full potential (whatever it may be), including expanding opportunity with high quality instruction and high expectations! Fluency is something *everyone* can achieve; it just takes work, painstaking, difficult work, day in and day out. (It's obvious why it'd be convenient for people to avoid admitting that this is as important as it is.) Some people will become fluent with less work vs others (see: talent); that's life, and life's not fair. That doesn't mean that it's not possible for *every* student to become fluent at every skill asked of them, it's just a matter of doing the work. (Again, some people will have to exert more effort than others to achieve said fluency; life's not fair! Doesn't mean it shouldn't be expected of them -- that'll make life even less fair!) When Ralph says "speed has been used to exclude people", I believe he is saying (please correct me!) that, in the past (and surely some small number of bad teachers today), if people saw students with less perceived talent, they would have lower expectations of them, and hence wouldn't provide them with as much opportunity and encouragement as others. In that sense, Ralph and I are completely aligned: have the highest expectations for *every* student! (I'm partial to the Ratatouille message: "Not everyone can become a great chef, but a great chef can come from anywhere!" So if your only selfish goal is to get the most research mathematics into the world, all societal issues be damned, then you too should want to teach *everyone* at the highest level!) When I say "speed is important", I'm referring to people telling kids that they don't need to know their times tables (etc) by heart to succeed in STEM. There might be a tiny of fraction of people who can squeak by without fluency in that one skill; but I guarantee you that every STEM profession is fluent in a TON of math. Fluency is very easy to measure: show me that you can solve a good number of problems correctly in a short amount of time (hey, we've just invented the "timed test"!). The best part: you don't have to agree with me! I'm not telling you what to do with your children or students; I'm just explaining what I insist on and expect from mine. Take it or leave it.
How about: misplaced or poorly implemented teaching that misuses speed and memorization has and continues to push away and exclude people from continuing their mathematics and education #iTeachMath
5
3
20
4,124
Did you know that even though #tractography is often considered a computationally expensive and offline process, the latest algorithms can now be performed in real-time without sacrificing accuracy? link.springer.com/article/10… @tractography @baranaydogan and Joseph Yang
1
9
24
1,203
📢 Deadline extended! 📢 The inaugural IST Conference in Bordeaux is giving you more time to submit your abstracts — new deadline: May 30, 2025. Don’t miss the chance to share your work! #tractography #IST2025 event.fourwaves.com/istconfe…
6
9
724
I just put up a new video on the fundamentals of quantum computing, building up to a step-by-step walk-through of something known as Grover’s algorithm. Here's an excerpt of a quiz used to introduce the topic.
89
440
44
4,510
236,923