Guillaume Pasquier

Alex, ouvre-toi ! Pas de grande prise de position éthique dans ce billet, j'avais juste envie de partager mon expérience d'utilisation d'OpenAlex (et un peu plus) dans le but d'avoir un suivi des publications de l'IHEID. Ça n'intéressera peut-être pas grand-monde, mais ça donne une idée de ce que…

✊ Ah beh ça fait 10 ans qu'on le dit et qu'on le fait. Faites-le aussi, ça rend l'atmosphère ici bien plus supportable ! #OnBloquelExtremeDroite Source : blik .media sur TikTok

Do research participants support data sharing? A Canadian Neuroscience Institute provided a second consent for research participants about whether their data could be shared for unspecified future uses. Of the 2912 participants, what percentage consented? Guess, then check the answer in the reply

It seems like the #nlproc community is overdue for a conversation about how we should and shouldn't use text processing in reviewing, and also, I guess, what peer review & science even are. #emnlp2026 , 🧵>>

In New Review of Academic Librarianship no less.

On rayonne à l'international. www.theguardian.com/world/2026/o...

95% in: the share of an article's cost paid before submission, mostly with public money. 95.9% out: the share of RELX's 2025 operating cash flow returned to shareholders. Two 95s. Opposite directions. This #OAWeek, let’s take a closer look at the numbers: sparcopen.org/our-work/cos....

"Les notions de désescalade & de stricte nécessité de l’usage de la force ont disparu du vocabulaire gouvernemental" "La France compte le plus grand nombre de mutilés et de tués par la police en «maintien de l’ordre» de toute l’Europe" A lire dans @letemps.ch www.letemps.ch/opinions/la-...

Are you an MLIS student or new-ish librarian worried about navigating convos around AI? Join @juliesetele.bsky.social & me for a free webinar hosted by Rutgers LIS students. Let's talk strategies for finding your (AI skeptical) people & finding ways to push back. 📚 scarla.rutgers.edu/event/talkin...

honestly the thing that I find most galling about all this is the massive overlap between people trying to apply human rights models to computer programs and those arguing that some categories of humans -whether by race, category, or caste- don't have them.

Over a year ago I met an Anthropic employee at a conference and at one point she mentioned that one of the things she was working on was whether or not it was cruel to ever turn off an AI model. I thought she was joking. She was not.

Is America in a demographic crisis? Demographers don’t think so. I conducted a survey of U.S. demographers, and this is what I found: 3%: “The low birth rate is a demographic crisis” 25%: "serious concern” 53%: "should be monitored, but no intervention necessary” 20% :“not a problem at all.”

Canada is refusing to issue ISBNs to books primarily generated with AI. This illustrates both the urgency of the problem of the growing flood of slop books, and also the toothlessness of addressing it on the honor system without there also being consequences for lying

Harvard just dropped a study AI productivity in software engineering and it found the average increase in productivity from using Claude code is.....zero and for exactly the reasons you'd think fion.ac/jellyfish.pdf

The Metric Is Not The Mission: The Finished Internet The Metric Is Not the Mission is a ten-part examination of how Big Tech moved from building and expanding the open internet to increasingly shaping it... https://www.techdirt.com/2026/10/05/the-metric-is-not-the-mission-the-finished-internet/

Et c’est reparti pour les chiffres et graphiques foireux sur l’éducation. - la dépense d’éducation française « parmi les plus hautes de l’OCDE en % du pib » alors que ça inclut les pensions de la fonction publique, considérant que les retraites de France telecom sont une dépense d’éducation

Rappel : en droit français (donc, dans un monde idéal, en langage journalistique, et dans un monde parfait, politique), le terme "violence" implique une atteinte à une victime humaine. Brûler une poubelle ou abîmer un bâtiment, ce sont des "dégradations". Les mots ont un sens et une portée.

I am working on a #CRAAP / #SIFT training on #medialiteracy. Would anyone have example websites/articles students could have a look at, like a climate change website funded by an oil interest group or something similar? If not, maybe I'll just try to grab a random headline and run with it.

Loll Yudkowsky, Soares and such AREN'T SHILLS? The Yudkowsky who founded the eugenicist, rationalist cult? The one who predected singularity to be 2025? The one in the sex parties that women have to attend to "network? The rape orgies with dubious consent mechanisms that Soares coorganizes?

A few months ago I started wondering if not using LLMs for coding right now would be a detriment to me as a programmer, but I concluded that the idea is nonsensical. I could start Not Writing My Own Code at any moment, and having stopped programming 2 years ago wouldn’t make me a better programmer.

The with diamond OA pledging with for-profit companies is diamond OA pledging is essentially a charitable act. It does not make sense for charity to generate corporate profits. S2O and "unlatching" are essentially hostage situations with good marketing.

My latest books cartoon for @theguardian.com

Still patiently waiting to be "left behind" by AI rather than it chasing me across every damn space like a slasher in a low budget flick.

Comment détecter un article généré par IA sur Wikipédia ? C'est désormais notre lot quotidien. 🧵 Un petit fil, à partir d'un exemple concret.

Certain kinds of smart people tend to have a "sexy apocalypse" problem when it comes to thinking about risk. E.g. skynet turning everyone in paperclips, while bad, is also *kind of cool and fun* at the same time. Where the reality is that most actual disasters are both stupid and disappointing.

holy shit, this is the whole ballgame in a single paragraph gAI systems require mass theft of copyrighted works, for the explicit purpose of replacing the market for the people they stole from, and the people in charge knew what they were doing ahead of time, and did it anyway

Oh. That's a heck of a subsidy.

If AI systems are viewed as too advanced to control, the companies that build them can’t held liable for the harms they cause.

Matilda has reached 170M deduplicated publications, among which 38.5M are full-text searchable. Everything is free, sharable and reusable and will remain as such under our #POSI commitments and following our mission matilda.science/mission?l=en #openscience #opendata

hey why is no one talking about the licensing situation with this A24/SCP thing? are they releasing it under the same cc-by-sa license every SCP project is required to have? are they betting creative commons won't hold up in court? seems like the kind of question a journalist should look into idk

Bah oui les LLM vont tous nous buter Par réchauffement climatique vous étiez où quand on discutait du coût énergétique et environnemental de ces gadgets pourris

AI is doing so much harm to the open web and open science... I now have researchers specifically asking for things to be closed off because they don't want their work to be harvested into Frankenstein's POS. F this S

Le pouvoir de nuisance n'est pas une indication de compétence. On peut à la fois penser que des modèles IA peuvent faire des dégâts massifs à nos infrastructures et notre société et être conscient qu'il y a une chance nulle de créer une superintelligence avec des LLM.

No matter how you refine your questions, the chatbot will generate responses reflecting what the greatest number of users are likely to want to hear, which tends to be close to what they think they already know." A great, painstaking piece. www.hnn.us/article/ai-i...

Republicans who see Canada as an enemy has leapt up this year, similar to how Republicans’ approval of Putin shot up in 2016. Virtually no anti-Canada US constituency before this year, and virtually no pro-Putin constituency before 2016. Out of nowhere. Now millions say that’s their position.

Here is how we built a custom middleware github.com/epfllibrary/... to support our mission of showcasing EPFL’s scientific output in our institutional repository, #Infoscience infoscience.epfl.ch. A short thread on our workflow

Lots of chips and memories being used and getting more expensive. Lots of AI Data Centers being built or planned. Nothing particularly significant has been added for users, in my opinion. That's what I mean by inefficient And I am not even talking about data workers.

$3.7 billion spent on APCs in 2025. That's ~1000X more than it costs to run bioRxiv/medRxiv. Worth thinking about as we consider how to fund infrastructure... arxiv.org/pdf/2608.16322