AI

I hope you'll join me on Tuesday, Oct. 20, for a conversation with Patrick O'Neill about his new working paper, Presumed Effective: The Manufacturing of an Evidence Base for AI-in-Education Through Flawed Meta-Analysis. (Link in Comments) When Patrick sent me the paper this summer, I was quite taken aback. He was reviewing a group of 14 meta-analyses of AI in Education with thousands of citations between them; one of them had already been retracted and he argued that the other 13 had serious, fatal flaws. How does a field that I've been part of for 20 years publish 14 faulty meta-analyses on one of the most important topics of our generation? What are the consequences for education research and policymaking if our scholarly review process cannot assure the highest possible standards for reviews and meta-analysis? How do school folks make evidence-based decisions if the self-correcting mechanisms of science are working much too slowly (and the self-policing mechanisms not working fast enough)? I'm very grateful that Susanna Loeb and Elizabeth Tipton for joining us and helping provide analysis and context (which may be different from my own!). If you see this, I hope you'll share it with a colleague who is an editor or reviewer for an education research or education technology journal or conference. I think if we can get a critical mass of scholars to look carefully at the evidence here, we can work together to ensure a better future for reviews and meta-analyses in this field.

I really don't think Science should publish any articles on education anymore if they think 0.07 is a "sizable signal" -- I want to be upset but I'm more embarrassed for them. It's so embarrassing I normally wouldn't even call them out except for some reason we use a publication run by chemists and physicists to determine the education of our children, and it has been doing damage there -- embarrassingly obvious damage -- for the past 20 years. If you are a person who has zero idea how to do educational research but a large dataset that makes the uninformed go "ooooohhhh" it has always been easier to get published in Science than a real educational journal and they really need to just stop before they cause more harm.

Hey Hope your weekend went well wherever you are. Off to the UK for a work trip Friday so loving hanging out around home with friends and fam.

At the center of the country’s debates over artificial intelligence is a simple but hard question: What kind of technology is this? Is A.I. a kind of “alien mind”? Are we unleashing a new species on the planet, one that will transform human society so completely that historical analogies to past technologies simply don’t hold? Or is A.I. more normal than that? Arvind Narayanan is a professor of computer science at Princeton University and the director of the Center for Information Technology Policy. And he’s an author, alongside his colleague Sayash Kapoor, of the extremely influential essay “A.I. as Normal Technology.” In that essay, and then in a Substack under that name, they lay out their case that A.I. is in fact something we’ve seen before — or at least, it’s close enough to past revolutionary technologies that we have a road map to deal with it. So I wanted to bring Narayanan on the show to hear that perspective.

Build a free MBA with NotebookLM. 197 free lectures from Harvard, MIT, Yale, Stanford and NYU in 10 notebooks, plus the skill to build your own university.

I have trained 290 NotebookLMs, but I realized some of them really hit the spot.

CETI is a nonprofit organization applying advanced machine learning and state-of-the-art robotics to listen to and translate the communication of sperm whales.

The chatbot interface barely scratches the surface of what AI can do. It is also being used to hide genuinely harmful and even criminal activities.

Cal Newport joins us on Intentional AI to talk about automating tedious work, protecting judgment, and keeping room for deep thinking. We also discuss digital minimalism and why a robot’s convincing answer can still be wrong. Episode Links

h/t Mike Caulfield and Anna Mills

Here are four things your students don’t understand about AI—and what you can do about it.

After almost two years, our paper "The shrinking landscape of linguistic diversity in the age of large language models" is out in Nature Human Behaviour, and it made the cover! Three studies, seven datasets, 880,000+ texts. The short version: when an LLM polishes your writing, the content survives but you don't. What we found: - After ChatGPT's release, the variance in writing complexity dropped across Reddit, arXiv, and Patch News. Writing converged toward a common style. - LLM rewriting cuts writing-complexity variance by 21–50% across datasets and models. - Classifiers that recover an author's age, gender, personality, empathy, moral values, and political affiliation from their original text stop working after a rewrite. The signal doesn't just weaken. It shifts toward dominant categories. - For psychology and the behavioral sciences: a century of work treats language as a window into the person. That window is closing. Text-based measurement (clinical screening, personality assessment, corpus studies of culture and morality) assumes the writing carries individual signal. As LLM-assisted text fills our datasets, that assumption needs checking, not inheriting.- - For AI: homogenization isn't a style complaint. Models that flatten their inputs degrade the personalization built on top of them, feed narrower text back into the next round of training, and quietly push writers toward dominant categories they didn't choose. Preserving variation should be a design target, not a side effect. Paper: https://lnkd.in/gDGU55vZ Research Briefing: https://lnkd.in/ge2Ka4ka Our earlier theoretical piece in Trends in Cognitive Sciences: https://lnkd.in/gQtPFtf4 Congratulations to Zhivar Sourati, who led this, and to Farzan Karimi-Malekabadi, Meltem Ozcan, Colin McDaniel, Alireza Ziabari, Jackson Trager, Ala N. Tak, Meng Chen, and Fred Morstatter. | 109 comments on LinkedIn

Large language models (LLMs) used as writing assistants homogenize writing style. Studies across 880,000 texts show that, following the release of ChatGPT, writing became less varied as artificial intelligence (AI) use spread. The LLMs preserved meaning but compressed the variety of language and biased author traits inferred from the words.

Sourati et al. find that variance in writing complexity dropped on social media, news and scientific writing after ChatGPT’s release, a shift from individuality towards uniformity. LLM polishing strips cues to gender, age, ideology and moral values.

2.2K likes, 82 comments. "I Fully Automated My Video Editing Using ...

729 likes, 168 comments. "I Told an AI Doctor My Ears Hang Low"

Jonathan Haidt sounded the alarm on phones in schools. He thinks what’s coming will be even worse.

AI can now build a decent presentation in less time than it takes to refill your coffee. Give it a report, meeting notes, and an old deck you like. A few minutes later, you may have a polished set of slides with charts, colors, layouts, and formatting that look surprisingly good. That is impressive. ItContinue reading "When AI Can Make Almost Anything, Judgment Becomes the Job"

Advanced models can help researchers sort through mountains of biological data.

Anthropic is moving your Cowork tasks to the cloud, and it’s a bigger deal than it looks. Here’s what changes and what to do about it

One of a number of ideas I'm thinking about ahead of 2026. The personal context passport lets people carry a portable, user-owned profile they can share on their terms - from low-stakes convenience (travel, fitness) and evolving into granular, time-limited permissions for high-stakes public services

h/t temit g.?

In my first note on OrgLens I asked whether an AI agent could diagnose an organisation using Stafford Beer’s Viable System Model (VSM). The answer was yes, more…

And that’s important for nonobvious reasons

3 likes. "The Inconvenient Library: Friction and Dissonance as Cent...

Large Language Models (LLMs) are becoming increasingly used to support higher education assessment, yet evidence of their capability on reproducing authentic institutional grade-band labels remains...

238 likes, 35 comments. "Is AI Turning Us All into the Same Person?...

I wouldn’t grant a stranger access to my email. So why would I grant a robot?

The use of artificial intelligence by hospitals and health insurers is escalating their longstanding feud over paying for medical care.