Garrison Lovely thinks many people on the American left have spent the past few years telling themselves a comforting story about AI: that it is fake, overhyped, a way for tech companies to raise money or lift share prices. Lovely, an author and journalist, says that story is wrong. The real danger, he argues, is that the companies might actually succeed.
Lovely set out his explanation, and his pitch to fellow progressives, in an episode of The Cognitive Revolution published on September 29, 2026. He was there to discuss his book Obsolete, which examines the leaders racing to build AGI (artificial general intelligence, AI able to do broadly what humans can do).
From techno-optimist to pessimist
The interview began with his broader view of technology. Lovely said he was a techno-optimist when he was younger, in a culture that shared the mood. "We were like told social media would connect us and Twitter would liberate people from authoritarian regimes," he recalled. Google was amazing, and Google Maps was very useful.
He still appreciates what technology can do, he said. But he is now "much more pessimistic" that people will get technology that is good for them under current conditions. By that he means shareholder capitalism, where companies maximize profits for shareholders with little concern for everyone else they affect.
To explain the shift, he borrowed a concept from the writer Cory Doctorow: enshittification. Lovely described it as a cycle. A tech company first makes its product as good as possible to win users. Once growth runs out, the goal becomes extracting as much profit from those users as possible, by cramming in ads and adding features that make money but may make the product worse. Lovely said he had not read Doctorow's book on the subject but found the concept "incredibly helpful."
Doctorow's original 2023 essay sets out three stages, not two. Platforms first treat users well, then shift value to the businesses that want to reach those users, and finally squeeze both groups for shareholders. Users put up with it because leaving would mean losing their contacts, audiences or customers.
Lovely's example was Google Maps. He said it no longer works as well as it used to. Paste in a full address, he said, and it may take only the first word and send you to the wrong place. He called this "crazy making" and "dispiriting": people have to use such products because there are no real alternatives, and they keep getting worse "in obvious ways."
AI, in his account, is the exception. "The only thing that's not" getting worse this way, he said, "is AI, where it just does get better", and faster and cheaper, with "amazing" steady progress. The catch is that this progress comes with what he called a "terrible risk and cost to society." He said he feels bleak about it, but remains "dispositionally optimistic" that people can change the structures and systems around technology so that it makes them "happier, healthier, wiser, more democratic."
Lovely also placed his book in relation to Doctorow's thinking about AI. He said Obsolete is in some ways a rebuttal to another Doctorow book. Both writers argue that AI is not inevitable, he said, but Doctorow thinks AGI is impossible.
Why the left looked away
The conversation then turned to the "stochastic parrot" meme, which describes language models as systems that merely stitch words together by probability, without understanding. The host asked why so much of the left had treated AI as hype and buried its head in the sand. Lovely said there was "no one explanation, but a few" and offered six.
Trusted experts took a hard line. The most prominent and influential AI experts who are also on the left have taken "this very hard line," Lovely said. He called the Stochastic Parrots paper "the quintessential example." People on the left, he argued, defer to experts who have PhDs and credibility and also share their values.
The paper, "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?", was presented at the FAccT conference in 2021. Its authors are Emily M. Bender, Timnit Gebru, Angelina McMillan-Major and Shmargaret Shmitchell. It warns about ever-larger language models: their environmental and financial costs, poorly documented training data that encodes dominant viewpoints and biases, and the risk that readers mistake fluent machine text for meaningful, accountable communication. It also acknowledges that bigger models have improved benchmark scores and have useful applications, and argues that these benefits should be weighed against the harms. It recommends careful dataset curation and research directions other than building ever-bigger models.
The left's base was in academia. Before Bernie Sanders ran for president in 2016, Lovely said, the left had real power mainly in academia. Academia has been hostile to the idea that AI is real "in a lot of ways," perhaps partly because of antagonism between the humanities and science and engineering fields.
Hype fatigue. Crypto, NFTs, the metaverse and social media were all promoted as transformative and positive. In Lovely's assessment they turned out to be "either not transformative and not positive or transformative and negative," and he put social media and crypto largely in the second group. Many people concluded that tech boosters always say their product will save the world, only for it to turn into "B2B SaaS or something" (ordinary business software sold by subscription), and got stuck in that frame.
Cope and denial. It is terrifying, Lovely said, to consider that companies could build machines that replace people. Even setting aside fears of human extinction, losing your job is "really, really bad for you," and unemployment across a whole society is very bad for that society. He pointed to Nazi Germany's rise as an example of what mass unemployment can lead to.
The bubble argument. The idea that AI is a financial bubble has been very popular on the left, Lovely said, and he named the writer Ed Zitron as its main promoter. Lovely credited Zitron with being persuasive and effective at reaching people, but in Lovely's own assessment Zitron says things that are "just not true" and easy to pick apart. He said Zitron "either doesn't know what he's saying or he's lying." Lovely devotes a chapter of his book to the bubble argument. Its appeal, he said, is that "it's a comforting story and it means you don't have to do anything."
Dislike of the messengers. Later in the conversation Lovely added one more reason. The most prominent people talking about existential risk from AI and superintelligence, he said, are figures such as Elon Musk, Sam Altman, Eliezer Yudkowsky and Nick Bostrom. The left does not like them, and the distaste is mutual. That "negatively polarized a lot of people," he said.
The message he wants the left to hear
Lovely's alternative is blunt: "These companies are trying to build machines that will replace us. They might succeed. And that would be disastrous for so many reasons. So we have to stop them." He admitted this is hard and requires people to organize, not just read and post. He called it "a message that could work."
He believes there is an "overhang" of quiet concern: many people worry about AI but don't want to get yelled at for saying so on Twitter or Bluesky. That mismatch is starting to resolve, he said, and he described seeing the dam begin to break, which he found encouraging.
He hopes that laying out the full case, from someone who shares their perspective, will persuade people on the left to take AI more seriously. Once they do, he argued, the stakes become familiar left-wing territory: "capitalists are trying to fulfill the lifelong dream of turning capital into labor without the intermediary of workers." If that succeeds, he said, capital's share of returns would go to 100% and labor's to zero, an outcome he called bad "from almost any perspective." Maybe some libertarians are into it, he said, but by and large they are still concerned too.
That makes it, in his words, "a pretty easy one for the left": the conclusion should be that "this is a real thing" and that the left should get on board and "not let them build those machines."