We have exciting news! #CI2026 will be hosted by the École Polytechnique Fédérale de Lausanne (EPFL), in the period of April 27-30, 2026. More information in Tom Beucler's post: www.linkedin.com/posts/tom-be...

Do you like to train neural networks to solve all your nasty probabilistic inference and sequential design problems? Do you love letter salads such as NPs, PFNs, NPE, SBI, BED? Then no place is better than the Amortized ProbML workshop we are organizing at #ELLIS UnConference.

5,290 papers will be published at NeurIPS this year. If you could still get the proceedings in a book, it would be like 50,000 pages long. That’s 14.5 papers a day to read each and every day until the next NeurIPS

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

Indeed 🥲 xeiaso.net/notes/2025/w...

📣 NEW -- In The Economist, discussing the privacy perils of AI agents and what AI companies and operating systems need to do--NOW--to protect Signal and much else! www.economist.com/by-invitatio...

Accepted papers at ProbNum 2025 have now been published as a volume in Proceedings of Machine Learning Research (PMLR). Thank you to all the authors for their beautiful papers and to the PMLR editor Prof Neil Lawrence @lawrennd.bsky.social for making this possible. proceedings.mlr.press/v271/

For context that's a third the price the RIAA expected a single human to pay for just downloading 24 songs a decade ago (after cutting the ruling down from $1.5m, or $62,000 per song), and that's not adjusting for inflation:

terminal colours wizardzines.com/comics/colou... (from "The Secret Rules of the Terminal", out now!)

My latest coding adventure is now up on YT -- some experiments with ray-tracing glass and water, and the cool caustics they can cause youtu.be/wA1KVZ1eOuA?...

The thing that's so difficult in fighting pseudoscience is that its followers fundamentally don't believe in expertise. Exceptional people? Sure. The naturally gifted? of course. But I've talked to people face-to-face who simply don't believe people can get better at things.

📣 Please share: We invite submissions to the 29th International Conference on Artificial Intelligence and Statistics (#AISTATS 2026) and welcome paper submissions at the intersection of AI, machine learning, statistics, and related areas. [1/3]

"Io stimo più il trovar un vero, benché di cosa leggiera, che ‘l disputar lungamente delle massime questioni senza conseguir verità nissuna"

Research Fellow position open at @gatsbyucl.bsky.social to work with me and Jason Hartford on Causality in Biological Systems! Apply at link, deadline is 27 August: www.ucl.ac.uk/work-at-ucl/...

just learned that CSS (as of "Baseline 2023") now supports nesting: caniuse.com/css-nesting I never managed to start using SCSS so I'm excited to be able to write things like this without a preprocessor: .cal-link { right: -18px; @media (max-width: 800px) { right: 0; } }

Perplexity is repeatedly modifying their user agent and changing IPs and ASNs to hide their crawling activity, in direct conflict with explicit no-crawl preferences expressed by websites. blog.cloudflare.com/perplexity-i...

I've strayed into another language family, and made Finnish the subject of this month's post. Under the surface though, it's a piece about prehistoric languages, and how millennia-long linguistic contact can make foreign languages seem more familiar. dannybate.com/2025/08/03/y...

This is an amazing way of writing “most of the world”

Schemes to validate adults and their IDs are possible but they will NOT be run by those capable. All private data exposure is itself a breach already.

By getting express trains off the existing network, HS2 clears space for others - particularly FREIGHT. Fewer lorries on roads = major CO2 win. The northern legs, presently "cancelled", are the crucial part. We need to safeguard the routes and BUILD THEM. www.theguardian.com/business/202...

EurIPS is coming! 📣 Mark your calendar for Dec. 2-7, 2025 in Copenhagen 📅 EurIPS is a community-organized conference where you can present accepted NeurIPS 2025 papers, endorsed by @neuripsconf.bsky.social and @nordicair.bsky.social and is co-developed by @ellis.eu eurips.cc

I saw an infographic a couple years ago about how to remove source identifiers from links and why it's important, but I can't find it again and too many people I know are sending me links with them so here's an infographic straight from the oven

New blog post! You can train a neural network to *just do things* -- such as *predict the optimum of a function*. But how do you get a big training dataset of "functions with known optima"? Read the blog post to find out! (link 👇)

#PrideInWords www.aidanem.com/tag/pride-in... Etymology of LGBTQ

This is a less-sympathetic iteration of a major problem in the information ecosystem, particularly re: mis/disinfo: quality sources are paywalled or otherwise inaccessible, while conspiracy and garbage is free and easy to access. Guess which gets consumed more?

We’d like to announce next week’s virtual seminar on Bayesian Decision-making and Uncertainty! Sebastian Ament (Meta) Unexpected Improvements to Expected Improvement for Bayesian Optimization Sign-up link: gp-seminar-series.github.io

more terminal fun: canonical mode wizardzines.com/comics/canon...

Happy to have joined this great team :) In which circunstances can we use abundant proxy preferences to quickly learn true preferences? I'm glad to announce our paper explores and proposes a model for one of these cases. Follow @zhuyuchen.bsky.social for more details. #ICML2025 #ICML

New work! 💪🏻💥🤯 When Can Proxies Improve the Sample Complexity of Preference Learning? Our paper is accepted at @icmlconf.bsky.social 2025. Fantastic joint work with @spectral.space, Zhengyan Shi, @meng-yue-yang.bsky.social, @neuralnoise.com, Matt Kusner, @alexdamour.bsky.social. 1/n

Ever used kernel mean embeddings or MMD? Wished you had a closed-form expression so you wouldn't need to estimate these? See our new paper with Alex Gessner, Toni Karvonen and Maren Mahsereci, which comes with an associated Python package. arxiv.org/abs/2504.18830

These custom feed recommendations are great~ The only thing I missed from algorithmic feeds is making sure I can get posts from infrequent posters and people from different timezones 😇

🔗 IKEA happily used Verdana on the web, because yay it’s on everyone’s machines. Until they expanded business across Asia and the Middle East, and found the supported languages of Verdana were lacking. This lead to Noto IKEA. Robin Whittleton explains www.robinwhittleton.com/2025/04/23/i...

Four "sinoform" scripts inspired by, but not directly derived from, Chinese, written around the 10-16 centuries: 1. Tangut; 2. Khitan Small Script; 3. Khitan Large Script; 4. Jurchen (based on Khitan LS).

Spanish ‘tengo un sueño’ means “I have a dream”. In contrast, ‘tengo sueño’ means “I’m sleepy” (lit. “I have sleep”). Why does ‘sueño’ mean both “sleep” and “dream”, two very different things? It’s because ‘sueño’ and ‘sueño’ are different words – with a different etymology. My graphic explains:

♂️⬅️♊

The BayesComp workshop on 'Bayesian Computation and Inference with Misspecified Models' will take place in Singapore on the 16-17th June. We have an open call for posters/contributed calls, with a deadline on the 1st May. More details on the website: postbayes.github.io/BayesMisspec...

I just came across this paper from ICLR 2024 which proposes an intricate combination of transformers and diffusion models to generate forecasts with these uncertainty bounds (red box), which are clearly inappropriate and could likely be outperformed by modelling the data as a random walk...

Would you present your next NeurIPS paper in Europe instead of traveling to San Diego (US) if this was an option? Søren Hauberg (DTU) and I would love to hear the answer through this poll: (1/6)

If you saw this thread from a few days ago with a sneak peek inside my new book, you might have noticed a discrepancy involving the subtitle.

Italian is a bit of an outlier among the major Romance languages. Spanish has 'escuela', Portuguese has 'escola', French has 'école' - all starting with a vowel - but in Italian it's 'scuola'. Did it never get a vowel? It sure did: Old Italian had 'iscuola'! Here's the story of this little vowel:

Typst 0.13 releases today! This update is all about listening to feedback. Read on to find out which of the most highly anticipated changes we made.

Put another way, screwing up a translation into another language happens. Not being able to render the Arabic script correctly is simply disrespectful. Let’s at least get that right.

We got a chocolate bar from Rüya chocolate (delicious!). Inside they included a translation of “I have a Rüya” into many different languages. There were issues. #language

Excited to share our ICLR 2025 oral "Residual Deep Gaussian Processes on Manifolds"! With @vabor112.bsky.social & @arkrause.bsky.social, we introduce manifold-to-manifold GPs that can be composed together, generalising deep GPs to manifolds. Applications include wind prediction & Bayes opt! 1/n

"On the third attempt, digging between the filing cabinets, they struck lucky..." This is mega exciting as the Basilica is one of the most important missing pieces of Roman London!

This is due to a beautiful and little-known covariance identity by Hoeffding (1940): Cov(x,y) = ∫∫[F(x, y) - F(x)*F(y)] dxdy So the difference between independence and uncorrelated-ness comes down to point-wise equality vs. a (weaker) integral equality between the CDFs. 2/2

1/📯I am hiring! Postdoc position in Probabilistic Machine Learning and Amortized Inference. @univhelsinkics.bsky.social with strong links to the Finnish Center for AI (FCAI) Please see blurb and link in thread below. Applications evaluated on a rolling basis. Please reshare!