Map from my copy of an 1832 atlas by educator & cartographer Emma Willard (1787-1870). She argued for equality in women's education, noting women were equally capable of subjects like math & philosophy. In 1821 she opened the first school in the U.S. to offer women's higher education. #WomenInSTEM

Discovery of the week: turns out there exists a literal "I hate [insert my name]" club. Haters join from pro-and anti-AI crowd, as they misunderstand my work, call me stupid & spread online hate. Misogyny is a hell of a drug Kinda assuring that the figures in my papers are giving them headaches

Monument Valley is such a wonderful, beautiful, and wholesome game. Warm recommend ✨💖

Also no about the semantic point; most who work in the area make a figure like the left one. Actually it's semantically and pragmatically vacuous, not just wrong but useless. bsky.app/profile/oliv...

This paper link.springer.com/article/10.1... Proves the Ingenia Theorem, which basically brackets the AI learning problem as NP hard. NP hard means that one needs exponentially growing resources to learn more and more things. Scaling is a chimera and we should stop wasting resources on this.

📝 Guest, O., Suarez, M., & van Rooij, I. (2025). Towards Critical Artificial Intelligence Literacies. Zenodo. doi.org/10.5281/zeno... ✨ Thread with my favourite quotes 👇 1/🧵

“Flying pigs are also possible in principle; possible in principle bakes no bread.” — Jerry Fodor (2005, p. 27) Screenshot from: van Rooij, Guest, et al. (2024). 'Reclaiming AI as a Theoretical Tool for Cognitive Science’: link.springer.com/article/10.1...

Dimension 2: CRITICAL THINKING 🤔 “Critical Thinking is deep engagement with the relationships between statements about the world.” Guest, Suarez, & van Rooij (2025). Towards Critical Artificial Intelligence Literacies. doi.org/10.5281/zeno... 12/🧵

allowing such views to pass uncritically. The core traps here with AI are what @irisvanrooij.bsky.social and I outline (based on our previous work) in a forthcoming short publication, distilled into the table below: bsky.app/profile/iris...

I am echoing @vajra.me's crucial message here by adding: library and archival workers, historians, professors, this is also us. There is no such thing as an 'ethical' use of aye-eye 🗑️ and I'm gonna need some of my industry colleagues to stop drinking the Kool-Aid on it.

There is no ethical use of AI because the continued existence of the AI industry is directly harmful to (alongside everyone else) the interests of everyone whose labour makes books exist—writers of fiction & nonfiction, illustrators, translators, researchers, editors, everyone.

Updated this to be my pinned post for now. TL;DR: I did a lot of research on trying to make it tractable to make tractable models of cognition and, as ti turns out, this is hard. More info on my homepage: irisvanrooijcogsci.com/2023/12/27/c...

26. van Rooij, I., Wright, C., Kwisthout, J., & Wareham, T. (2018). Rational analysis, intractability, and the prospects of ‘as if’-explanations. Synthese, 195(2), 491–510 link.springer.com/article/10.1...

41. van Rooij, I., Guest, O., Adolfi, F. G., de Haan, R., Kolokolova, A., & Rich, P. (2024). Reclaiming AI as a theoretical tool for cognitive science. Computational Brain & Behavior. link.springer.com/article/10.1...

40. van de Braak, L., de Haan, R., Dingemanse, M., Toni, I., van Rooij, I., & Blokpoel, M. (2024). Intractability obstacles to explanations of communication. Proceedings of ICCM2024. osf.io/ep8qc_v1

36. Rich, P., de Haan, R., Wareham, T., & van Rooij, I. (2021). How hard is cognitive science? Proceedings of CogSci2021. Paper: escholarship.org/uc/item/8cr8... Video: www.youtube.com/watch?v=2bdK...

35. Woensdregt, M. S, Spike, M., de Haan, R., Wareham, T., van Rooij, I., & Blokpoel, M. (2021). Why is scaling up models of language evolution hard? Proceedings of the Annual Meeting of the Cognitive Science Society, 43. escholarship.org/uc/item/0217...

34. van Rooij, I., & Baggio, G. (2021). Theory before the test: How to build high-verisimilitude explanatory theories in psychological science. Perspectives on Psychological Science, 16(4) 682–697. journals.sagepub.com/doi/10.1177/...

32. Kwisthout, J. & van Rooij, I. (2020). Computational resource demands of a predictive Bayesian brain. Computational Brain & Behaviour, 3, 174–188. link.springer.com/article/10.1...

31. Rich, P., Blokpoel, M., de Haan, R., Otworowska, M., Sweers, M., Wareham, T., & van Rooij, I. (2021). Naturalism, tractability and the adaptive toolbox. Synthese, 198, 5749-5784. link.springer.com/article/10.1...

30. van Rooij, I., Blokpoel, M., Kwisthout, J., Wareham, T. (2019). Intractability and Cognition: A guide to classical and parameterized complexity analysis. Cambridge: Cambridge University Press. cambridge.org/core/books/c...

Short answer: that's actually not what we claim. Long answer: bsky.app/profile/iris...

"Safeguard current agreements that prioritize people and nature."

Also much harder to develop Probably requires unusually Luddite parenting / schooling

And the courage to use them as a uniquely individual contribution.

In a world where people can offload their cognitive labor to AI systems on a large scale, critical reading, writing and thinking skills are more important, not less.

#Skystorians *mic drop*. I was just telling STS colleagues how I was struggling with a puzzle from something I had seen in an Inquisition edict. Keeping it vague cause will write a blog about it. I had a pic of the edict but it was not clear which texts it referred to. Usually they would mention 1/2

What is shocking is that researchers still think in this year of the lord that 'AI tools' will be objective and neutral.

Greta Thunberg arrested for holding a sign opposing genocide. When peaceful protest is a crime, democracy is in deep trouble. Happening under a Labour Government.

These might be useful too bsky.app/profile/oliv... Thank you @jedbrown.org!

Exactly! Thank you for this thread. This be useful for people to tease apart different forms of AI, by focusing on their relationship to humans: bsky.app/profile/oliv...

Wildly different things, tasks, techniques, subspecialties being lumped into "AI" and then being conflated with each other, doesn't help. Different types of models vs the techniques to train them vs the tasks they are supposed to accomplish, all under "AI".

And the same pattern with cancer research. These people aren't interested in reducing suffering, they find it rhetorically useful for their empire-building.

For generations, when you learned to program you were constantly being forced to be better at paying attention to the words on the screen. Noticing you'd made a typo, noticing you'd used the wrong argument, noticing shit in general. Over time, you got better at noticing the words on the screen.

Learning how to pay attention to small details, how to deliberately strain yourself to notice tiny details that are off, is a skill programmers learn the painful way. And, using a chatbot deskills you in this very extremely important skill of being able to pay attention to small details.

It's funny that everyone who has *ever* programmed, knows how important attention to detail is. Hours wasted because you made a typo, forgot to add a character, etc. The tiniest thing you literally don't notice is wrong, and your program screams at you and dies. This is a universal experience.

obligatory plug of this piece @kara-williams.bsky.social and I wrote a few months ago explaining the stakes of policymaking with vague/improper ideas of what AI means consumerfed.org/specific-ter...

Providing false/biased information (including as a "summary") to a human decision-maker should receive similar scrutiny to algorithmic decision systems. This is usually lost in legislation, and governments are using such systems without adequate testing based on incorrect assumptions of what they do

Thanks for writing and re-upping this. I think terminology is important, but I don't think regulating based on technical or statistical properties (e.g., "generative" or "transformer") is appropriate. I think the mode of interaction and epistemic function is more important for regulation.

I hope the Canadian government is paying attention and stops hyping the AI industry and expecting tax payers to bail it out: "AI startup Cohere and CoreWeave plan multibillion-dollar data center in Canada; Government to help fund the project" So tired of the corporate welfare model...sigh

Experiencing a clear peak in trolls these last two days? I wonder why 🤔

Oh fun, I am being accused again of being a creationist 🫠 Sure …. onlinelibrary.wiley.com/doi/full/10....

Worst thing is that ppl who are experts on their field have to waste their time playing pidgeon chess with ppl who are not.

So sorry, my mistake: I thought your question was in good faith, but now it seems you just want to express discontent about a thread. The full paper of course includes information that does not fit in a thread (though even the thread made clear the context). link.springer.com/article/10.1...

Couple all that with the, sadly, sometimes successful approach of "fake it until you make it" (IMO usually driven by the sunk cost fallacy/magical thinking akin to gamblers, by C-suite) and this forces marketing folks to publicly talk that way to keep their jobs, with some as true believers. /end

Is it just me, or are Marketing experts more likely to be enthusiastic abt AI/LLMs? I guess if ones profession is to convince ppl to use/buy things they otherwise wouldn’t, it makes senses perhaps that AI hype is appealing? But perhaps I saw a non-representative slice? Happy to stand corrected!