Transcription
The 1810s were a difficult time for England. War in Europe had dragged on for nearly two decades, and Napoleon's continental system, a trade embargo designed to strangle British commerce, was squeezing the economy. Britain was also fighting the United States in what became the War of 1812. Food prices were soaring, exports were collapsing, and working people were already under pressure.
Domestically, the effects of steam power were being felt across industry. The technology had been developing for decades, but was now diffusing at an increasing rate, displacing workers who had held jobs for life. Some 10% of the male workforce earned their keep as hand loom weavers. But high-pressure steam and machine-made textiles were putting them out of work. Against this came a movement named after a legendary figure called Ned Lud. They were machine smashers. They sought to wreck the new steam-powered economy and return it to its old ways. They were the Luddites.
I tell this story because it's like something we're seeing now, just over 200 years later. In January, YouGov published one of the most comprehensive international surveys of public opinion on artificial intelligence today. In the UK and the US, only around 25% of people held a positive view of AI. 44% of Britain's and 45% of Americans were negative. Polling from mid-2025 also showed anxiety over the future of work. 66% of people in the UK and 62% in the US believed AI would destroy more jobs than it creates. Only 6 or 7% thought AI would be a net creator of employment. And when asked whether AI's benefits outweigh its drawbacks, more people in both countries said no than said yes. Public opinion has turned against AI before any mass job losses have actually materialized. The backlash has arrived ahead of the thing people are afraid of. And the question that should concern those in government is what happens when the displacement really starts?
The backlash is showing up in three places at once. The first is politics. In March, Senator Bernie Sanders posted a video to YouTube in which he interviewed Anthropic's Claude Chatbot about the dangers of AI. It's set in a dark room with slightly sinister music, and its caption reads, "What an AI agent says about the dangers of AI is shocking and should wake us up." The video has 4 million views and rising at the time of filming. The tech press pointed out fairly that Sanders was asking leading questions to a system designed to be agreeable and that Claude was reflecting his assumptions back at him rather than providing independent analysis. But the video's popularity revealed something more important than these issues. Millions of people wanted to hear that AI was dangerous, and Sanders gave them a credible-looking vehicle for that belief.
Days later, Sanders and Representative Alexandria Ocasio-Cortez introduced the AI Data Center Moratorium Act, legislation that would impose an immediate federal moratorium on all new AI data center construction until Congress passes national safeguards for workers, consumers, and the environment. Sanders had previously spoken on CNN about AI risk. "If there are no jobs and humans won't be needed for most things, how do people get an income to feed their families, to get health care, or to pay the rent? There hasn't been one serious word of discussion in the Congress about that reality." The bill is unlikely to pass, but it represents something significant. Anti-AI sentiment has crossed from tech commentary into populist politics. According to Sanders and Ocasio-Cortez, over 100 local communities across 12 states have already enacted their own data center moratoriums.
The second strand is creative displacement. An analysis of 180 million job postings found that computer graphic artists experienced a 33% decline in listings in 2025, following a 12% drop in 2024. Photographers fell 28%. Writers also fell 28%. Research from Cornell University found that on freelance platforms, demand dropped 20 to 50% for skills that AI can substitute, like writing and translation. Meanwhile, copyright lawsuits are stacking up. The New York Times against OpenAI, Disney and Marvel against Midjourney, and a Stanford and Yale study found that Claude could reproduce copyrighted books near verbatim with accuracy rates above 94%. The framing of AI as theft has gained mainstream traction, and it feeds directly into the broader backlash.
The third strand is the bubble narrative. Gary Marcus, a former professor and long-standing AI skeptic, declared in late 2025 that the AI bubble was over. A report from MIT stated that 95% of organizations were getting zero return on their generative AI investment despite $30 to $40 billion in enterprise spending. A National Bureau of Economic Research study published in February found that 90% of firms reported no impact of AI on workplace productivity. Industry estimates suggest total AI spending could surpass $1.6 trillion over the next few years. And yet the returns remain unclear. For people already suspicious of AI, each of these data points reinforces the same story. The technology is overhyped. The companies are untrustworthy. And ordinary people are being asked to absorb the costs.
So how bad could this actually get? In March, Claude developer Anthropic published a detailed survey of AI's real impact on the labor market. The researchers introduced a metric called "observed exposure," designed to measure what AI is actually doing in workplaces rather than what it could theoretically do. The gap turned out to be enormous. In fields like business, finance, computer science, and legal work, AI can theoretically cover more than 80% of tasks. But actual adoption remains a fraction of that capability. The study found no systematic increase in unemployment among workers in AI-exposed occupations since ChatGPT launched in late 2022. That sounds reassuring. But the researchers identified one early signal worth watching. Hiring rates for workers aged 22 to 25 in exposed occupations have dropped by around 14% compared to 2022 levels. The problem for young workers is not being pushed out. It is that the door is creaking shut. And because these workers never had jobs in the first place, the loss does not show up in unemployment statistics very easily.
We've covered the predictions from AI leaders like Dario Amodei and Mustafa Suleyman in earlier videos on this channel. Those predictions have not come true yet, but the Anthropic researchers named a scenario that anyone in the knowledge economy should be thinking about: a great recession for white-collar workers. During the 2007-2009 financial crisis, US unemployment doubled from 5% to 10%. A comparable doubling in AI-exposed occupations has not happened yet, but Anthropic's analysis suggests it could. The displacement of hand loom weavers in the 19th century was not sudden, but their wages fell from 21 shillings a week in 1802 to less than 9 shillings by 1817 – a 15-year wage compression that hollowed out a way of life. By 1815, Manchester's 40,000 hand loom weavers found it nearly impossible to compete with power looms. So they sold their labor at lower prices and depressed the market. If AI follows a similar pattern of gradual wage pressure rather than dramatic layoffs, the economic pain will accumulate long before it shows up in headline unemployment figures, and the political backlash will grow accordingly.
There is a concept from economic history that might describe what we are heading into better than any AI forecast. The economist Robert Allen coined the term "Engels' Pause" to describe the period during the Industrial Revolution, roughly 1790-1840, when productivity surged but wages flatlined. For 50 years, the gains went to capital owners while workers saw little return and even displacement. Allen named it after Friedrich Engels, who documented working conditions in England in 1845. Strategy consultant Neil Perkin drew a useful analogy on his Substack. When electricity arrived in the late 19th century, factory owners ripped out their steam engines and replaced them with electric dynamos. But they kept the same factory design, the same belts and pulleys, and the same workflow. Productivity barely improved. It took 30 years for someone to realize that instead of one big electric motor powering everything, you could put a small motor on each machine, redesign production around the flow of materials rather than flow of power, and invent mass production. The resistance to redesign was partly economic. Factory owners waited until existing assets depreciated before investing in new layouts, and the new paradigm required a generation of expertise – factory architects and electrical engineers who simply didn't exist yet. Most corporate AI adoption right now looks a lot like dropping a dynamo into a steam-powered factory: task automation within an existing paradigm rather than the kind of fundamental redesign that would unlock real value. Organizations are messy, political, and path-dependent. Jobs are not always allocated rationally. Status is tied to team size, and internal processes are full of exceptions that are opaque and hard to navigate. Re-architecting workflows at the system level rather than the task level is genuinely difficult, and it will take time.
This matters because it suggests the real threat is not sudden mass unemployment. It's something more insidious: productivity gains flowing to corporate profits while workers run faster on the treadmill for the benefit of someone else. As a Harvard Business Review study has suggested, instead of freeing up time for workers to refocus on higher-level tasks, AI risks intensifying work. You still have a job, but you are doing more of it, and the returns are going elsewhere. Another way of looking at it is that you have a two-tier economy: AI-native startups that undermine old processes and lead to job disruption, while the old order's slow processes and governance structures mean its existing defenses are its greatest flaw. If these patterns hold, an internet backlash will become more of a physical one: more frequent strikes as unions negotiate, political polarization, and potential civil unrest. In Indianapolis, a city councilman who voted in favor of data center development had shots fired at his home. A note was left at his door reading, "No data centers." And then, in the early hours of April 10th, someone threw a Molotov cocktail at Sam Altman's home in San Francisco. Less than two hours later, federal agents arrested the culprit while he allegedly tried to break into OpenAI's headquarters with a jug of kerosene, a lighter, and an anti-AI manifesto. In his response, posted at 3 a.m., Altman admitted he had underestimated the power of words and narratives. Perhaps the AI backlash isn't just coming; it's already here. When people feel the economic system is rigged against them and democratic institutions are not able to respond, history shows us what comes next.
But I'm also skeptical about the more extreme predictions. The Anthropic study itself is the most important counterargument. Despite all the anxiety, there has been no systematic increase in unemployment in AI-exposed occupations. Not yet, anyway. And the broader research landscape is full of contradictions. A Washington Post analysis in March noted that one Stanford study found AI was bleeding jobs from young workers in exposed occupations, while research from the Economic Innovation Group concluded the opposite: that young workers in those same fields were faring better than peers in less exposed roles. The experts, put simply, cannot agree.
There's also a reasonable case that some of the current disruption is self-inflicted. Harvard Business Review reported in early 2026 that, based on a survey of over a thousand global executives, companies were laying off workers in anticipation of AI's impact rather than because AI was actually performing the work. The job losses were real, but driven by expectation rather than demonstrated capability. Set against a backdrop of geopolitical instability and a shakeup of world trade that certainly hasn't helped, the 2020s has been a historically weak decade economically. If that is the case, some of the backlash we're seeing is a response to corporate panic rather than genuine technological substitution. History suggests that job disruption from new technology is real, but technology alone has never produced the kind of sustained mass unemployment that critics fear. The working week has shortened over time. Younger people spend more years in education, and entirely new industries emerge, which would have been unimaginable a generation earlier. The more likely outcome is a painful but manageable transition. Roles change, wages compress in some sectors, new kinds of work appear, and the economy absorbs the shock over a decade or two rather than collapsing in a few years. But if I'm wrong about that, the political consequences will be severe, and governments will need to act fast.
Even Sam Altman, the CEO of OpenAI, seems to recognize this. OpenAI recently published a policy paper that explicitly acknowledged the risks of jobs being disrupted, industries being reshaped, and wealth concentrating among a small number of firms. Altman told Axios that AI superintelligence is so close and so potentially disruptive that America needs a new social contract on the scale of the 1930s Progressive Era and the New Deal. His company proposed a public wealth fund to give every citizen a stake in AI-driven growth, robot taxes, 4-day work week pilots, and adaptive safety nets that trigger automatically when displacement metrics hit certain thresholds. To me, it's not particularly reassuring. He's basically saying this new safety net is down to the state. Big tech has become immensely powerful in the last 20 years, and OpenAI is really the icing on the cake. If more people are out of work, then tax receipts decline, which weakens the state. The private enterprise responsible for the situation should be required to do more. This is an unprecedented situation in modern capitalism should it unfold.
So where does that leave us? The Luddites were right that machines would destroy their specific livelihoods. They were wrong that stopping the machines was possible or desirable. The living standards that industrialization eventually produced were genuinely transformative. In 1800, 30% of babies born in Britain died before the age of 5. 200 years later, it was 0.4%. But "eventually" was cold comfort to the hand loom weavers whose wages halved in their own lifetimes. The question for our generation is whether we repeat the Engels' Pause – decades of rising productivity captured by capital while workers absorb the disruption – or whether this time we manage the transition before it reaches breaking point. The backlash is already here. If the new system cannot be contained, there really will be a rage against the machine.
You've been watching Absolutely Agentic. We're a media consultancy and startup focused on AI, and we create weekly YouTube videos and a newsletter we send a few times a week. In a world of AI hype, we're trying to strike a balance, looking at a range of opinions to cut through an increasingly complex world. The best thing you can do to support us is to sign up to the newsletter in the description or subscribe to the channel. But also, don't forget to like and leave us a friendly comment. See you next time. It'll be absolutely agentic.