An important reason to fix reviewing (solution is simple, by the way: phase out double blind) is to reduce the whole "If it were not for rectangular-headed reviewers, we were winning the Nobel Prize. Scholarship is dead, etc." shtick. Tired of this every cycle.

Looking for a PhD program? It all starts with great supervision. Choose wisely. www.nature.com/articles/d41...

I am increasingly getting comfortable with the idea that estimation and uncertainty quantification can be treated in quite different ways, which would once have been quite strange to me.

I’m finding that GPT-5 with agent mode is really good for doing quick literature reviews. For example: “Make a report on how people use the Gauss-Newton approximation to the Hessian in VI.”

I ran the following experiment: for each person in the errata, I asked gpt to search for other books in which they appear. I learned about "Bayesian Filtering and Smoothing". Ended up buying it. I'm very impressed by the complexity of the query and the difference in performance –Claude found nothing

Far less focus on consumer product feature, and much more on its development as an enabling technology for industrial automation and robotics. Take a snapshot of one of their plans, for instance. For them integrating AI into industrial production, and building [+]

This is not OK. I don't submit often to NeurIPS, but I reviewed papers for this conference almost every year. As a reviewer, why would I spend time trying to give a fair opinion on papers if it's what happens in the end???

Submit your work on to the @euripsconf.bsky.social workshop on on Differentiable Systems and Scientific Machine Learning! differentiable-systems.github.io/workshop-eur... Submission: 10 October Notification: 31 October Workshop: 6 or 7 December in Copenhagen

Unfortunately, our submission to #NeurIPS didn’t go through with (5,4,4,3). But because I think it’s an excellent paper, I decided to share it anyway. We show how to efficiently apply Bayesian learning in VLMs, improve calibration, and do active learning. Cool stuff! 📝 arxiv.org/abs/2412.06014

Interesting feeling: I'm reading a book and, on the errata page, I see many acquaintances' names. Feels like I'm late to the party. (Clearly one I should’ve read earlier.)

🔬 Are you working in computational imaging or interested in learning? 📖 DeepInverse just released a new 5-minute quickstart tutorial deepinv.github.io/deepinv/auto... which gets you started developing AI models for reconstructing images!

Next week I’m going to be in the Netherlands. Happy to meet people to chat!

Next week I’ll be in Barcelona. I’m happy to meet people to chat, especially if you know of academic-adjacent positions.

I am pleased to announce that together with some friends, we are organising a workshop on Non-Reversible MCMC Sampling, taking place at Newcastle University from 8–10 September 2025. Details on the programme and registration can be found at the workshop website (sites.google.com/view/probai-...).

This will go down as one of the most important AI Ethics piece.

I saw a vegetarian pasta recipe for soy-based marinara, and... I'm sorry, but "soy marinara"? YO NO SOY MARINARA Soy capitán Soy capitán

Para toda persona a quien pueda interesar

Fixed it for you...

I'm starting to write some Lean snippets on my blog to keep track of things I want to remember. This one is about how to prove an inequality involving real numbers (and logarithms): jburroni.github.io/lean/2025/06...

Just completed the formalization of the Kraft-McMillan lemma in lean. Quite an interesting experience and feeling: it is complete (there is nothing missing and all is correct), but clearly not finished (things can be made nicer). It took me quite a lot (1.3kloc). I hope I can get faster.🤖 helped!

welcome to the future, now your error-prone software can call the cops (this is an Anthropic employee talking about Claude Opus 4)

I'm speaking at Programming Language Implementation Summer School (PLISS) next week. Let's see if I can get the attendees to implement a tiny version of Boxer: dl.acm.org/doi/pdf/10.1... It is a fun example of how interactive programming *systems* can do more than *languages*!

Coming to @programming-conf.bsky.social? Submit a Lightning Talk proposal to share & discuss what you are currently working on! 2025.programming-conference.org/track/progra...

The registration for @programming-conf.bsky.social is still open. Just sayin'...

Our new work deconstructing the concept of "AGI", led by the brilliant @borhane.bsky.social in collaboration with @graziul.bsky.social , Hananel Hazan, @shiridoshi.bsky.social +more co-authors below, has been accepted to ICML. 🤗 We discuss the many "traps" that... 🧵 1/ arxiv.org/abs/2502.03689

Upright bikes have been a game-changer for urban biking around the world since they first “exploded” outside of Europe — I wrote this with @modacitylife.com back in 2014, “In Praise Of The Upright Bike” back when we were still spreading the word in those early years. www.huffpost.com/archive/ca/e...

Seeing o3 writing a python script to use pil to zoom-in in an image is the most blade runner thing ever

"Part of the reason that the AI community is so interested in math is that math is hard." @emilyriehl.bsky.social on Testing Artificial Mathematical Intelligence at the Simons Institute's workshop on AI for Mathematics and Theoretical Computer Science. simons.berkeley.edu/workshops/si...

Vivimos un presente donde no es obvio que es posible y que imposible. Esto fue un paseo en bici con Gemini live (cámara prendida, le iba mostrando problemas en la ciclovía)

I’m very very impressed on how “uv” solved the scenario of running complex python scripts. You can share a github repo, “uv run main.py”, pin pan pum, and it is running. No virtual env, no issues with python versions

Here's something funny: - By distance travelled, pure-electric cars make up ~22% of total in 2023 in Norway - By *amount of energy consumed*, it's like, maybe 5% ish This is bc electricity is ridiculously more efficient than fossil or bio :) robbieandrew.github.io/EV/

Looking for postdocs! if you are interested in Geometric Deep Learning, Computational Chemistry, TDA, Diffusion Models, then give it a thought :)

One potential loose relative of Sutton's so-called bitter lesson is the observation that if you're always designing methods (e.g. algorithms, models, etc.) with intuition built from low-dimensional problems, then you might not naturally find good solutions for high-dimensional problems.

I already advertised for this document when I posted it on arXiv, and later when it was published. This week, with the agreement of the publisher, I uploaded the published version on arXiv. Less typos, more references and additional sections including PAC-Bayes Bernstein. arxiv.org/abs/2110.11216

I find it quite interesting how disparate the behaviors of GPT and Claude are with Lean4. (The former fails miserably, while the latter can do—some—things.) I think it says something about the difference in training between the two, but I don’t know exactly what it is.

You publish papers in statistics journals and you're proud of them? But, when your friends ask "are you coming to AISTATS with us?", you feel sad and you regret not to submit papers to conferences? Well, this message is for you! There is now a Journal-to-conference track at #AISTATS2025!

🚀 Excited to announce the launch of a pilot trial for the Journal-to-Conference track at #AISTATS2025, coordinated by @pierrealquier.bsky.social and me The track considers published papers from AoS, Biometrika, JASA, JMLR & JRSS B for a poster presentation at AISTATS 2025 📅 Deadline: 15 March 2025

🚀 "UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate" is accepted at #ICLR2025 A thread! 📜 Paper: arxiv.org/abs/2409.01985 🖥️ Code: github.com/tachella/uns...

People do this often, so we have an information page how to approach this so it goes smoothly en.wikipedia.org/wiki/Wikiped...

I'm looking for an emergency reviewer or two for an AISTATS submission in the area of Bayesian ML theory, which at present has the misfortune of three low-quality reviews. If you're able to help, please DM and let me know! Please link your Google Scholar if I don't already know you personally.