The Return of the Luddites

Erik J. Larson considers “The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want” by Emily M. Bender and Alex Hanna.

By Erik J. LarsonOctober 31, 2025

• Science & Technology

• Philosophy & Religion

BY THE TIME a frequently boozy, brilliant, and scandal-inclined Lord Byron delivered his impassioned speech to the British House of Lords on the February 27, 1812, the poet had already weighed in on the dangers of mechanization. In his poem “The Curse of Minerva,” he had lamented, “The starv’d mechanic breaks his rusting loom, / And desperate mans him ’gainst the common doom.”

Byron’s speech was aimed at the Tory-sponsored Frame Work Bill—colloquially known as the “Frame-Breaking Act”—which sought to make the destruction of power looms (or “wide frames,” which enabled multiple pieces of linen to be simultaneously knitted) a capital offense.

Byron rose not to defend property crimes but to defend the people accused of them. In so doing, he gave voice to what history would later call the Luddites—a brief, explosive uprising of unemployed and disenfranchised stocking-weavers around Nottingham. The Luddites were not anti-technology in any modern sense. They were breaking the frames—smashing the machines—because they saw, clearly, what those machines meant: the erasure of their livelihoods and the reduction of human craft to mechanical repetition. They weren’t rejecting technology; they were rejecting the social costs of automation.

Byron’s speech captured the essence of Romanticism, the movement he embodied alongside early visionaries like William Wordsworth and Samuel Taylor Coleridge. Romanticism, in its most urgent register, was not an aesthetic posture but a humanist revolt: an insistence that man not be remade in the image of the machine. That revolt would echo through Thomas Carlyle and John Ruskin, who would later warn that “we must not allow mankind to be sacrificed to improvements in Mechanism.” To which we postmoderns might now add: Nor to the seductive spread of digital automation under the name of AI.

Enter Bender and Hanna.

Emily M. Bender, an outspoken computational linguist at the University of Washington, and Alex Hanna, a sociologist and Silicon Valley transplant from Google, who now directs the Distributed AI Research Institute (DAIR), stand at the forefront of a growing movement critical of our unexamined embrace of generative AI. Their targets are the now-familiar large language models—ChatGPT, Google’s Gemini, Meta’s Llama series, Anthropic’s Claude, and others—that promise intelligence but often deliver automation in disguise.

Bender and Hanna’s mission? A forceful plea for socially centered technology, as well as a resistance movement against unaccountable systems, exploitative data practices, and the creeping displacement of human labor. In other words: Byron’s mission. In modern parlance, Bender and Hanna are hype-busters, dismantling the inflated promises that seem to trail artificial intelligence wherever it goes.

In their new book The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want, they take aim not merely at the technology itself but, rather, at the entire ecosystem that sustains its mythos—venture capital, platform monopolies, media complicity, and a growing cultural fascination with automation disguised as progress. If history moves in spirals, then we have arrived, once again, in Nottingham, 1812. But this time, the critics are on the inside.

Bender and Hanna’s seemingly modest task of de-hyping AI quickly becomes a sweeping critique of an entire technological and economic worldview. Their central move is deceptively simple: stop calling it “AI,” and start calling it what it is—automation. Reframing AI as automation rather than a kind of intelligence strips away the mystique and returns us to the real questions: What tasks are being automated? Who benefits from that automation? And what happens to the humans—writers, artists, drivers, translators, even coders—once they’ve been abstracted into a training dataset? By insisting on the term “automation,” Bender and Hanna reveal that what’s sold as innovation is often just labor displacement with better branding.

To be sure, de-hyping AI has been a full-time gig for critics since the field’s inception in the 1950s. No one seems to win this game. Boosters are fond of setting a moving window of 10 to 20 years for the arrival of AGI—artificial general intelligence equal to that of humans—or even ASI, superintelligence far beyond us.

The AI Con is not just another attempt to deflate AI hype through argument alone; much of its force depends on a tone that is both straightforward and sharply comical. Bender, after all, is the originator of the now-famous phrase “stochastic parrots” to describe large language models, and together the authors host Mystery AI Hype Theater 3000, a podcast that brings similarly satirical energy to the world of AI critique.

Bender and Hanna introduce a host of alternative terms and vivid metaphors designed to reorient readers away from the mystique of “artificial intelligence” and back toward the more familiar—and far less magical—domain of automation. Neural networks, the core architecture behind most modern AI, become “mathy maths.” Large language models are recast as “synthetic media machines” or “text extruding machines,” a nod to both their reliance on scraped human content and their true function: probabilistically generating sequences of words based on prior patterns. In place of Big Tech’s endless exhortations about “the coming wave” of revolutionary intelligence—or breathless warnings about rogue robots and synthetic consciousness—we get what we might call the thoughtful critics’ Newspeak: a demystifying, often hilarious lexicon that cuts through the fog of hype.

The result is a full-throated, informative, and often funny self-described takedown—not of AI as technology but of the hype-tastic (one of their neologisms) and mendacious ecosphere that sustains it, or the illusion of it as a new kind of intelligence we must bend the knee to. What Silicon Valley frames as a revolution in thinking, Bender and Hanna reframe as more of the same: an intensification of automated systems that call for the kind of humanist pushback Byron once gave the loom.

The effect is edifying. Though I’ve been chipping away at AI hype for most of my career, I still found much of interest in The AI Con. Bender and Hanna brim with fresh ways of looking at what many of us now consider old-hat problems in AI, like the infamous 1960s chatbot ELIZA, created by Joseph Weizenbaum. ELIZA, named after Eliza Doolittle, the working-class flower girl who feigns upper-class manners in Pygmalion (1913), was one of the earliest attempts at conversational AI. It had no access to the internet or vast textual corpora. Instead, it used simple pattern-matching grammar rules, succeeding only because the context was so cleverly constrained.

Weizenbaum modeled ELIZA’s behavior after a Rogerian psychotherapist, the kind who simply repeats a patient’s words back in the form of a question:

“Men are all alike.”

“In what way are men all alike?”

The structure mimicked the cadence of a real conversation. Students of AI know ELIZA well, as it’s a staple of the field’s lore, and typically appears alongside discussion of the Turing test, that other canonical benchmark that asks a machine to convince a human interlocutor it is a person.

But Bender and Hanna aren’t interested in ELIZA’s technical limitations. They’re interested in the gorilla in the room: ELIZA created a powerful illusion of mind while having none. So what exactly is the difference between ELIZA and today’s chatbots, besides more text and more processing power? This is a question that points to what many of us critics have long suspected: the emperor still has no clothes.

Bender and Hanna are quick to draw out these perennial traps in our thinking about AI, from ELIZA’s parlor tricks in the 1960s to today’s breathless media coverage of ChatGPT. Their explanation for the success of today’s systems is both disarming and straightforward. Unlike ELIZA, which never left the lab, modern AI has been deployed widely, but not because of a conceptual breakthrough. Rather, the key ingredients were Moore’s law (ever cheaper computing) and the rise of the internet (ever more data to train on). The authors don’t dismiss the engineering, but they rightly note that many of the advances in the web era are infrastructural and economic, not intellectual.

This kind of informed deflation is especially welcome. I’ve long argued that today’s AI is more powerful largely because computers are faster, and there's simply more data to crunch. That looks impressive, but it’s not really a scientific revolution. In place of the mystical narrative of progress, Bender and Hanna offer something more grounded: hype. They define it succinctly early on: “Hype is the aggrandizement of some person, artifact, technology, or technique that you, the consumer, absolutely need to buy or invest in as early as possible, lest you miss out on entertainment or pleasure, monetary reward, return on investment, or market share.” That about covers it.

As The AI Con deepens, the Luddite theme emerges more clearly as the book’s larger through line. Its aim goes beyond just debunking AI hype; Bender and Hanna also hope to restore the good name of the Luddites—those early 19th-century stocking-weavers who were against the dehumanizing effects of mass automation, rather than just opposing technology as a concept. In the authors’ telling, “Luddite” isn’t an insult—it’s a warning and a word of caution. One we must return to, repeatedly, if we’re to resist the constant drumbeat of consumerism dressed up as science and progress.

Yes, these are well-trodden themes. But Bender and Hanna bring a fresh, updated vision of what it means to be a “Luddite” today. In the end, The AI Con reminds us that the fight over AI is not only about technology. It’s also about language, labor, and the stories we choose to tell about the future, which we can humanize rather than surrender to the myriad ways mass automation dehumanizes. Byron might’ve recognized the stakes.