Low Rank Jack

"Enhancing Your Beamer Presentations with SVG Figures and Inkscape Layers" https://laurentjacques.gitlab.io/post/enhancing-your-beamer-presentations-with-svg-figures-and-inkscape-layers/

"Countries that slide from democracy toward autocracy tend to follow similar patterns. To measure what is happening in the United States, the Times editorial board has compiled a list of 12 markers of democratic erosion, with help from scholars who have studied this phenomenon. The sobering […]

This day 30 years ago - on Oct 26th 1995, #CPAN (the Comprehensive Perl Archive Network) was announced to the world by Jarkko Hietaniemi – CPAN's Self-Appointed Master Librarian (OOK!), to the comp.lang.perl.announce newsgroup.

Happy Birthday, CPAN! […]

Largest study of its kind shows AI assistants misrepresent news content 45% of the time – regardless of language or territory. An intensive international study was coordinated by the European Broadcasting Union (EBU) and led by the BBC […]

@laurentduval Hi Laurent, I kind of remember you explained me once that, for the Hadamard transform, the concept of "frequency" should be careful defined and you possibly gave a reference then that I forgot. Do you remember it? Thank you ;-) (I post this question here as it could be of general […]

I've been using Goodnotes for handwritten notes and PDF annotation on iPad, but the latest version is full of all kinds of weird AI things. It's no longer as responsive as it used to be, and it is hard to trust that no information will leak to third parties. Any good alternatives?

(sigh) How many (conference) articles mention something like "due to the space limit, we leave the proof to the supplementary material" with this “supplementary document” nowhere available!

Back to the wonderful CIRM (my last time here was in 2018), in Marseille, France for the workshop on "Blind Inverse Problems in Imaging", organized by Luca Calatroni, Emmanuel Soubies, and Pierre Weiss. I'm eager to attend the seminars tomorrow. More information here […]

This is amazing.

"Finally Over: Google Blocks Sideloading of Android Apps" https://www.androidsage.com/2025/08/26/google-blocks-sideloading-of-android-apps/

A variation of the "distracted boyfriend" meme. In this version, everyone is in wearing togas, and the boyfriend has been turned to stone as he looks back at a passing Medusa.

So help me, I laughed
#shitpost #meme #distractedBoyfriend

“Owner of …” or “Moderator on…”

#Mastodon does NOT need to verify you in ANY way!

The only verification happening is when you create an account and confirm your email, that’s it

Staff members can recognized by a badge on their profile right under the username(see image)

If someone does ask […]

[Original post on mstdn.social]

Just arrived in the wonderful city of Lyon, France, to begin a sabbatical year at ENS Lyon, working with Rémi Gribonval and Paulo Gonçalves' OCKHAM team [1]. The location is magnificent, and I am welcomed in the Collegium, the institute for advanced study of Lyon University [2], located inside […]

Linux franchit les 6 % de parts de marché sur desktop : vers une adoption de masse ?
https://www.blog-nouvelles-technologies.fr/336627/linux-6-pourcent-part-marche-desktop-2025/

"C’est une statistique qui aurait semblé impensable il y a encore quelques années. Selon une étude menée par […]

Andrew Kadel @DrewKadel@social.coop My daughter, who's had a degree in computer science for 25 years, posted this about ChatGPT on Facebook. It's the best description I've seen. Something that seems fundamental to me about ChatGPT, which gets lost over and over again: When you enter text into it, you're asking "What would a response to this sound like?" If you put in a scientific question, and it comes back with a response citing a non-existent paper with a plausible title, using a real journal name and an author name who's written things related to your question, it's not being tricky or telling lies or doing anything at all surprising! This is what a response to that question would sound like! It did the thing! But people keep wanting the "say something that sounds like an answer" machine to be doing something else, and believing it *is* doing something else. It's good at generating things that sound like responses to being told it was wrong, so people think that it's engaging in introspection or looking up more information or something, but it's not, it's only, ever, saying something that sounds like the next bit of the conversation.

The only thing ChatGPT ever does.

A dice, lost inside a old style wooden school bench. The picture is shot from inside the bench drawer. A shaft of light falls on the dice from a slit on the top right. The dice is in front of the camera, displaying the score six on its only visible face. The wooden interior of the bench is raw and dusty, with paint streaks and tone variations on the wood. Five circular spots of light, coming from a hole not visible on the picture (the ink pot well lit by several light sources in the room), are marking the wooden floor where lies the dice.

Find a caption...

If you haven't seen this yet, watch.

It's an AI vid of the "grand opening" of the hideously gilded, wildly overpriced, Let Them Eat Cake future Trump Ballroom.

https://www.youtube.com/watch?v=nD_25aSMSFc&t=5s

Fun statistical observation: If X follows a Chi square distribution with k degrees of freedom, while the MGF Eexp(tX)] of X doesn't exist for t>1/2, I computed that, for k=3, and the cardinal hyperbolic sine sinhc(u):=sinh(u)/u that basically behaves like exp(u)/2u if u is large, then […]

SampTA 2025

Currently at the "Sampling Theory and Applications" conference (#Sampta25), enjoying great talks and the beautiful city of Vienna. https://sampta25.univie.ac.at

Microsoft, AWS, and even France-based OVHCloud say that they can't keep the US Government's hands off your data, if requested under the CLOUD Act.

👉 You can't guarantee data sovereignty using a provider that is subject to US law, even if your data is housed outside the US […]

"Why are data nerds racing to save US government statistics?"
https://archive.is/7gF5f

"After watching data sets be altered or disappear from U.S. government websites in unprecedented ways after President Donald Trump began his second term, an army of outside statisticians, demographers and […]

"Archivists Recreate Pre-Trump #CDC Website, Are Hosting It in Europe."
https://www.404media.co/archivists-recreate-pre-trump-cdc-website-are-hosting-it-in-europe/

"A team of volunteer archivists has recreated the Centers for Disease Control website exactly as it was the day Donald Trump was […]

YouTube video by Timothy Gowers IMO2025Q1

From time to time I like to video myself solving maths problems in real time. Since the IMO has just happened, I plan to try a few of this year's questions. Here I get through Question 1. More to come at some point.

www.youtube.com/watch?v=1G1n...

Guardian version https://www.theguardian.com/tv-and-radio/2025/jul/18/the-late-show-with-stephen-colbert-to-end-in-2026-as-cbs-cancels-show?CMP=Share_iOSApp_Other

Please report any account that tells you that you need to verify your #Mastodon account to continue using it through a private message. It is a scam. We do not require identity verification. Real staff accounts either have a special role badge on their profile or are verified through the […]

The card game "publish or perish"

Look at what my friend brought me from the USA!

With many thanks to @fuzzyleapfrog for the recommendation!

#AcademicChatter

Standard error of what now? Scrolling through discourse about the evil vector, I stumbled across a group of people dismayed that the papers they were reviewing for NeurIPS didn’t include proper error bars. In fact, I learned that there’s a checklist that authors must fill out and attach to every NeurIPS submission. You can still learn things on social media. The NeurIPS program chairs introduced the checklist in 2021. They argued that the community wanted “both more guidance around how to perform machine learning research responsibly and more flexibility in how they discuss this in their papers.” They came to this conclusion after listening at the “NeurIPS 2020 broader impacts workshop,” a fully remote workshop held during the cold, dark winter of the second wave of the covid pandemic. They argued that they would experiment with a checklist as a way to facilitate more responsible machine learning. They called the checklist “experimental,” though there is no control group.1 They hoped that “future Program Chairs will continue to improve and evolve the checklist in subsequent years.” You don’t have to be a bureaucracy scholar to know that checklists “improve and evolve” by metastasizing in length and complexity. And that’s precisely what we’ve seen. The NeurIPS paper checklist is now 3800 words long. This is twice as long as the original checklist, and has 3 times as many items to check off. It’s a lot of ridiculous infantilizing boilerplate. Item 1: “Do the main claims made in the abstract and introduction accurately reflect the paper's contributions and scope?” Come on, folks. You are required to check that you have read the code of ethics (another 2000 words here). You are asked to check whether you obtained the appropriate IRB approvals. Of course, this only applies if you are at a university. There’s a call out to the AI Safety dorks: “Do you have safeguards in place for responsible release of models with a high risk for misuse (e.g., pretrained language models)?” Come on, folks! But I’m particularly fascinated by the weird obsession with statistics. I imagine Leo Breiman is chuckling up in heaven that one of the two cultures is trying to strangle the more successful one.2 In 2021, there was a single line about statistics: > 2021: Did you report error bars (e.g., with respect to the random seed after running experiments multiple times)? Honestly, not the worst question in the world. So much of machine learning and AI is based on randomized algorithms, and it’s good to check whether you have a stable result (i.e., training a neural net for classification) or a wildly variable result (i.e., using reinforcement learning to game a robotics simulator). But someone on the steering committee decided we needed a longer rule set and more statistics. The statistics creep began in 2023: > 2023: If you ran experiments, did you report error bars (e.g., with respect to the random seed after running experiments multiple times), or other information about the statistical significance of your experiments? And eventually, due to some unknown person’s lobbying, these were expanded further in 2024: > 2024: “The factors of variability that the error bars are capturing should be clearly stated (for example, train/test split, initialization, random drawing of some parameter, or overall run with given experimental conditions). The method for calculating the error bars should be explained (closed form formula, call to a library function, bootstrap, etc.). The assumptions made should be given (e.g., Normally distributed errors). It should be clear whether the error bar is the standard deviation or the standard error of the mean. It is OK to report 1-sigma error bars, but one should state it. The authors should preferably report a 2-sigma error bar than state that they have a 96% CI, if the hypothesis of Normality of errors is not verified. For asymmetric distributions, the authors should be careful not to show in tables or figures symmetric error bars that would yield results that are out of range (e.g. negative error rates). If error bars are reported in tables or plots, the authors should explain in the text how they were calculated and reference the corresponding figures or tables in the text.” What does this have to do with anything? If I run my code ten times, why should you care if I properly account for normal approximations? Who does this help? Anyway, I love that NeurIPS is proving my central thesis about statistics: Statistics is a bunch of arbitrary rules we use for approval. These arbitrary rules have now found their way into the machine learning publication machine. Whereas I understand why we require statistical tests when approving pharmaceuticals, no one has provided an explanation for why (or if) these statistical guidelines improve the quality of the thirty thousand NeurIPS submissions. Indeed, I can’t figure out _why_ people have become so obsessed with error bars in machine learning. I have been told that it’s because data sets are “small” now. For example, some of the “can LLM solve human tests” data sets have a few dozen questions. The AIME benchmark has 15. But what do frequentist error bars buy you here? This isn’t like Fisher’s friend who tastes tea. If a machine can solve a single one of these problems, it’s interesting! LLM answers are variable by design. Trying to gauge this variability with “Gaussian approximations to the standard errors of the mean” misses the forest for the trees. Indeed, if you only have 15 questions in a dataset, you don’t need statistics. Just look at the answers! We’re not grading the LLM on a curve here. The obsession with statistics is particularly ironic because the advances in machine learning from the past 15 years have been entirely based on optimaxxing vibes. While program committees and responsible ethics boards fixate on procedure, the big ideas have come from “this feels right,” whether they be bigger convnets, ADAM optimizers, attention mechanisms, or anything in RL. Do you think recent trends in transformer architectures like using RMSNorm instead of Layernorm or SwiGLU instead of ReLU are undergirded by deep statistical grounding?3 What gives away the whole checklist charade here is bullet 4. The rules say that you must disclose the normal approximations in your error bars, but _you don’t have to release code_. > "While NeurIPS does not require releasing code, we do require all submissions to provide some reasonable avenue for reproducibility, which may depend on the nature of the contribution.“ Why no code? It’s because a substantial fraction of NeurIPS papers come from private companies. The biggest models come from private companies. The conference exists to enrich Sam Altman. There’s no recruiting fair, no late-night parties, no signing bonuses based on Google Scholar profiles without the corporate commitment. So we can pat ourselves on the back and tell ourselves we’re being responsible researchers as we fill out our checklists and format our error bars. But let’s not kid ourselves about what we’re participating in. Subscribe now 1 I couldn’t find the preregistration plan. 2 There is still no statistically significant evidence to support the existence of heaven. 3 Like you, I don’t know what any of these things are. By Ben Recht

"Standard error of what now?
The NeurIPS checklist corroborates the bureaucratic theory of statistics."
By Ben Recht, https://www.argmin.net/p/standard-error-of-what-now

Mastodon’s latest update readies the app for Quote Posts, revamps design https://techcrunch.com/2025/07/08/mastodons-latest-update-readies-the-app-for-quote-posts-revamps-design/

The French city of Lyon is dropping #Microsoft software in favor of #OpenSource alternatives. It doesn't want to depend on proprietary, closed-source software, and it doesn't want to depend on *American* software. It doesn't want "potential [US] governmental surveillance." […]