Birgit Schmidt

hallucinated references will land you a 1-year ban from arxiv now. wow

Attention @arxiv authors: Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated. 1/

AI people don't understand what copyright is. Cameras, lenses, paints, brushes, software, and paper are tools and materials, none of them 'contains a copyright'. They also can't generate previously copyrighted things spontaneously for you. Books, tutorials, and museums are all things meant to be learned from. Art, art history, and culture are all things you can learn from as well, and if you aren't plagiarizing or using those materials directly in your work without permission, there is no copyright issue. AI is built directly from other people's work without permission, the models contain weights that essentially serve as compression and CAN generate copyrighted materials, both intentionally and unintentionally. It also competes directly with the people it steals from, which is NOT fair use. OpenAI CEO Sam Altman is also on record saying generative AI couldn't exist without copyrighted materials. Pretending that literally every art tool or resource or source of inspiration is the equivalent of mass theft of the entirety of mankind's creative output is not just disingenuous, it's propaganda. Do not listen to these people.

My latest *Locus* column is "Reverse Centaurs," and it sets out to unravel a paradox: how is that some AI's users describe their experience as a hellish ordeal, while others delight in the ways that AI is changing their lives for the better? locusmag.com/2025/09/commen… 1/

It Took Many Years And Billions Of Dollars, But Microsoft Finally Invented A Calculator That Is Wrong Sometimes - Abigail Segel, Defector defector.com/it-took-many-y…

New Normal Curves podcast out today! We discuss an MIT paper that recently made headlines like "ChatGPT rots your brain." Spoiler alert: The paper has a lot of problems. Listen here: bit.ly/chatgptmit or watch here: bit.ly/chatgptmitvid

Stronger together: European e-Infrastructures are uniting for better collaboration and services. At the recent Assembly meeting in Amsterdam, @EGI_eInfra, @Eudat_eu , @GEANTnews , #PRACE, & #OpenAIRE shared goals, progress, and a joint vision for the future of research in Europe. Read the recap: connect.geant.org/2025/08/06/eur… #OpenScience #EOSC #eInfrastructure #ResearchCollaboration

From today's @ftopinion @FT #TomLehrer

How great was Tom Lehrer? In 2020, he moved his entire catalog into the public domain because he felt he'd made more than enough money off of it. Enjoy every Tom Lehrer album: tomlehrersongs.com

Tom Lehrer has left us, aged 97. An incredible talent who, in 2023, put all his songs and recordings into the public domain, making them free for anyone to use for anything, forever. What a remarkable bloke.

My cartoon for this week’s @GuardianBooks

#RecommendedRead: Kherroubi, G.I., Erdmann, C., Gesing, S., Barton, M., Cadwallader, L., Hengeveld, G., et al. (2025) 'Ten simple rules for good model-sharing practices', PLoS Comput Biol, 21(1): e1012702. doi: doi.org/10.1371/journa….

Presentation slides for my talk yesterday at #SciPy2025 Barba, Lorena A. (2025). Embracing GenAI in Engineering Education: Lessons from the Trenches. figshare. Presentation. doi.org/10.6084/m9.fig…

Incredibly important report on AI training & copyright released today by the European Commission. It gets lots right, slamming the idea that creators should be required to opt-out of having their works used for free by AI companies. It says: - Generative AI must be based on opt-in consent - Current EU law does *not* permit generative AI training on copyrighted work without a license. Some claim Article 4 of the CDSM does, but this view is mistaken. - Requiring creators to opt-out or have their work trained on for free would be unfair and contravene international law - AI systems don't learn like humans, from a technical or legal perspective It suggests a solution of a statutory exception for AI training with an unwaivable right of equitable remuneration for authors and rights holders. I'm not sure this is the right solution - but it's great to see the Commission coming out so strongly in favour of respecting creators' rights, and making it clear that tech companies' dreams of unpaid use of copyrighted works cannot be permitted. Read the full report here: europarl.europa.eu/RegData/etudes…

This is what I wrote in the Preface to my book 6 years ago. Peter Norvig told me that he almost stopped reading after this part, but people whom he trusted told him that it was worth continuing and he hasn't regretted he later told me. Isn't it still true today with LLMs, VLMs, and all other LMs? 100% true. While you are in distribution, everything works like a charm, but should your foot just slightly leave this nice comfort zone, things get ugly quickly. This is how machine learning is, and as long as you remain in this paradigm, it will always be this way. If you disagree with this state of affairs, you have three choices: 1. Become a scientist, and through decades of research (and sacrifice), and if you are incredibly lucky, formulate a new AI paradigm. 2. Become a liar like those infamous CEOs and their sycophants AI influencers. 3. Become a blind believer like the followers of those infamous liars and their sycophants.

The correct take: - “stochastic parrot” is a top notch turn of phrase that does accurately convey something about what LLMs do - stochastic parrots can be smarter than most of the people who like that phrase think - but not as smart as the people who don’t like that phrase think

Explore what’s new for researchers in the EOSC EU Node in the upcoming Ask Me Anything session on July 8th (14:00-15:30 CEST) to unpack: - New user policies & credit models - Training resources on OpenPlato - Federated services - Q&A session 🔗open-science-cloud.ec.europa.eu/events/eosc-eu…

Everybody, feel free to sign this even if you are not NL-based and ‼️ share ‼️ (anonymous signature is poss): Open Letter: Stop the Uncritical Adoption of AI Technologies in Academia (link below)

#RecommendedRead: Ronald Snijder has written two blog posts (1) (2) outlining the impact these bots have had on the OAPEN Library and DOAB, where automated systems have drowned out genuine users: -oapen.hypotheses.org/1265 -oapen.hypotheses.org/1450

"AI’s use by high-school and college students to complete written assignments, to ease or avoid the work of reading and writing... puts the process of deskilling at education’s core. To automate learning is to subvert learning" Great piece by @roughtype newcartographies.com/p/the-myth-of-…