Dan McQuillan

that's good

image: book cover of 'Writings of the Luddites' edited by Kevin Binfield https://www.press.jhu.edu/books/title/3144/writings-luddites

One of the lessons the Luddites learned the hard way, and the critical AI movement could avoid relearning, is that when it comes to the harmful social impacts of tech the state is NOT a neutral arbiter between private capital and the common people.
💻🔨
Recommended text 👇'Writings of the Luddites'

Also unsurprising that contemporary AI is failure from start to finish even when it's trying hard to be nasty, as it's grasp on reality is so tenuous (aka correlational). The flip side is that it's error-prone fragility produces collateral damage at scale.

Do you have any views on the claim that fusion power is just around the corner and will provide for AI's energy needs? I am deeply sceptical.

Thanks, that looks very thorough.

Fission for Algorithms: The Undermining of Nuclear Regulation in Service of AI - AI Now Institute A report examining nuclear “fast-tracking” initiatives on their feasibility and their impact on nuclear safety, security, and safeguards.

It is very concerning. If you're interested in the details behind our comments on both 404 pieces, we wrote a report that breaks down these arguments (as two nuclear safety experts who work on AI):
ainowinstitute.org/publications...

Headline from linked 404media article which reads '‘Atoms for Algorithms:’ The Trump Administration’s Top Nuclear Scientists Think AI Can Replace Humans in Power Plants' followed by a photo of a nuclear power plant.

"we’re developing advanced software systems capable of accelerating nuclear reactor deployments by enabling AI to... generate 3D models, [and] conduct rigorous failure mode analyzes with minimal human intervention”
#ChernobylAI 🍄☁️☢️
www.404media.co/nuclear-rian...

Screenshot of Youtube livestream page with a photo of Dan McQuillan and the talk title, 'Resisting AI', and timing, 12 EST / 5 GMT,

"Our conversation with Dan McQuillan will examine the political stakes of AI, asking what can and should be done". Livestreamed Friday 12 EST/5 GMT on YouTube with audience Q&A. www.youtube.com/live/wi0u3Hs...

I hadn't heard of that film, thanks for posting

Yes, but is the earlier point ironic i.e. "Let the AI do the bullshit work and we students and teachers can get about the business of learning"? Where is the institution and politics in this Elysian picture?

A screenshot of a story from The Guardian on 20 Nov 2025 showing a powerpoint slide and with a headline reading ‘We could have asked ChatGPT’: students fight back over course taught by AI" and a subheading reading "Staffordshire students say signs material was AI-generated included suspicious file names and rogue voiceover accent"

"a course that's eliminated AI will appeal as an alternative pathway to universities concluding that, thanks to AI, they don't really need most lecturers, and students concluding that, for similar reasons, they don't really need universities."
danmcquillan.org/cpct_seminar...

What are those tasks doing if not modulating the token prediction? I'm not saying current approaches are simple, I'm saying they're harmful.

Palantir CEO Says Making War Crimes Constitutional Would Be Good for Business Alex Karp vows to use his

Palantir CEO Says Making War Crimes Constitutional Would Be Good for Business

gizmodo.com/palantir-ceo...

I didn't say bag of words. It is token prediction but with a massively expanded set of correlations.

I think using terminology like 'facts encoded in the model' is also misleading; I would say an LLM has encoded patterns which weight towards word associations that may be interpreted as factual by people reading the output.

I don't think an LLM has approximated a world model, although it might simulate the appearance of having done so. I agree on the language based consistency.

Form is all they got (the content is stochastic)

Screenshot of the mastodon post at https://ec.social-network.europa.eu/@EUCommission/115643458015221823 where the first bullet point reads "€2.6 billion invested in 19 AI factories, giving developers and researchers supercomputing and data for trustworthy AI innovation."

I wonder if the European Commission's insistence on calling data centres 'AI factories' is a deliberate attempt to encourage Luddism and machine breaking? 🔨
(Think of the alternative social benefits that could be delivered with €2.6 billion...).

This isn't a public utility it's a protection racket.

The concept of system prompts as natural language instructions seems to be confusing the idea of LLMs as content-addressable (not true) and simulcra-addressable (more accurate imho).

YouTube video by The Philosopher AI and the Digital: Resisting AI - Dan McQuillan in conversation with Andrés Saenz de Sicilia

In conversation with Andrés Saenz de Sicilia of The Philosopher journal as part of their 'AI and the Digital' series, covering reform, resistance, fascist resonances, (dis)passionate applications, system contamination and path dependency.
www.youtube.com/watch?v=nXEA...

If you think generative AI should be integrated into the classroom then you also think management consultancies should determine the future for higher education. Both are ethics-free forms of machinic optimisation based on reductive abstractions that strip out important context and relationships.

South Korea's Experiment in AI Textbooks Ends in Disaster The South Korean government's billion-dollar AI textbook initiative has been a complete failure from start to finish.

Who could have seen this coming except literally everyone

share.google/GFaMy3K2Jjlo...

Major AI conference flooded with peer reviews written fully by AI Controversy has erupted after 21% of manuscript reviews for an international AI conference were found to be generated by artificial intelligence.

Leopards, faces, etc.

"Yet 60 per cent of respondents also view AI as the single biggest threat to research integrity". Instead of reflecting on the fact that maybe they're right, JISC's recommendation is to assimilate generative AI more deeply at the core of the process 🤦

A screenshot of yesterday's Financial Times story with the headline 'Health minister says UK should cash in on NHS patient data - Zubir Ahmed says new central service for medical datasets should be leveraged for ‘benefit of Treasury coffers’ over a photo of a smiling Zubir Ahmed.

'Health minister says UK should cash in on NHS patient data - Zubir Ahmed says new central service for medical datasets should be leveraged for ‘benefit of Treasury coffers'' www.ft.com/content/0531... - the State emulating social media's surveillance-to-monetisation pipeline.

Full video of panel and discussion now up on PeerTube tubedu.org/w/teHCy6pkiK...

Office Hours 16-Haunting Futures: Big Tech and Resistance w/Syed Mustafa Ali YouTube video by Adnan Husain

Live now: Haunting Futures: Big Tech and Resistance w/ Syed Mustafa Ali www.youtube.com/live/GEZQXb8...

Left: washing machine with the words “Critical Washing” on it. Right: Quote: “When we critique AI, we should do so with intellectual honesty and in a principled way. Utmost care is needed to avoid ethics washing, greenwashing, and generally — what we dub — critical washing. (…) Researching and reflecting on the harms of AI is not itself harm reduction. It may even contribute to rationalizing, normalizing, and enabling harm. Critical reflection without appropriate action is thus quintessentially critical washing.” —Suarez, Müller, Guest, & van Rooij (2025) https://zenodo.org/records/15677840

Yes, hats off to the organisers! Glad it was relevant. I suppose the question is 'where next?'