Those warning of AI-driven mass unemployment are predicting something that has never happened before.
In an era when the architects of artificial intelligence speak openly of rendering human labor obsolete, history offers a quieter counsel: no technology has ever spread fast enough to outpace humanity's capacity to adapt. From the steam engine to the shipping container, disruption has arrived gradually, creating more work than it destroyed, even when the suffering along the way was real. The labor markets of wealthy nations today remain near record strength, and economists warn against mistaking the loudness of a warning for its accuracy. Whether AI will finally break this pattern remains an open question — but it is a question history has heard before, and answered differently than the fearful expected.
- The most powerful voices in Silicon Valley — Amodei, Gates, Altman — are predicting unemployment levels not seen since the Great Depression, yet the actual job market refuses to cooperate with their forecasts.
- Economists push back hard, arguing that the 'lump of labor' fallacy haunts these warnings: technology has always destroyed some jobs while generating the wealth that funds entirely new ones.
- A deep dive into centuries of economic history reveals that even the most celebrated disruptions — the steam engine, mechanized farming, early computing — unfolded across generations, not quarters, giving labor markets time to breathe.
- The infamous 'Engels' Pause,' when Industrial Revolution wages stagnated, turns out to have been caused less by machines than by war, grain tariffs, and slow productivity growth — not a template for AI-driven collapse.
- The real warning signs to watch are specific and measurable: per-capita GDP growth cracking the historic 2.5% ceiling, wages decoupling from productivity, and mass job losses cascading across sectors — none of which are visible yet.
- History's final lesson is that if AI disruption comes, it will likely announce itself during a recession — making the next economic downturn a genuine diagnostic moment for whether this time is truly different.
The people building AI are also its loudest alarmists. Dario Amodei of Anthropic warns of unemployment reaching 10 to 20 percent. Bill Gates has suggested humans could become largely unnecessary. Sam Altman, sensing the backlash, has softened his language — but still speaks of significant disruption. And yet the labor markets of wealthy nations tell a different story: OECD employment at record highs, the United States adding jobs even in fields supposedly most exposed to automation, and the Bureau of Labor Statistics projecting 5.2 million new American jobs by 2034. Economists, far less alarmed than the technologists, point to the 'lump of labor fallacy' — the mistaken belief that the job market is a fixed pie that technology can simply shrink.
The deeper question is whether AI is genuinely unprecedented. Economic historian Robert Gordon found that no leading economy has ever sustained per-capita GDP growth above roughly 2.5 percent annually. That ceiling reflects how slowly transformative technologies actually diffuse through an economy. Agricultural mechanization took generations. The tractor, invented in early 20th-century America, displaced farm labor across decades, not years. Even the mid-20th century boom — early computers, containerized shipping, rapid industrial change — produced labor disruption at double today's rate while delivering rising wages and expanding opportunity. People remember it fondly.
The Industrial Revolution's 'Engels' Pause,' when British wages stagnated between 1790 and 1840, is the cautionary tale most often invoked by AI pessimists. But recent scholarship complicates the analogy. Employment in Britain nearly tripled between 1760 and 1860. Wage growth was slow, but no slower than in the fifty years before the steam engine arrived. The real culprits were war and grain tariffs that spiked food prices — not the machines themselves. As historian Nicholas Crafts concluded, the Industrial Revolution is simply not a model for technology that boosts productivity while shrinking labor's share of income. That combination has never actually occurred.
This does not mean it cannot. The signals would be hard to miss: productivity surging past Gordon's historical ceiling while real wages stagnate, corporate profits soaring as workers fall behind, and job losses spreading across multiple sectors at once. History also suggests that if disruption comes, it will arrive during a recession — the moment economies shed unproductive work most rapidly. The jobs that vanish in the next downturn will be the clearest evidence yet of whether AI is rewriting the rules. Until that evidence arrives, even the most prominent voices warning of catastrophe are, by their own admission, speculating.
The leaders of the very companies building artificial intelligence are now among the most vocal about its dangers. Dario Amodei of Anthropic has warned that AI could push unemployment to between 10 and 20 percent. Bill Gates, Microsoft's cofounder, suggested that in an AI-dominated world, people would become largely unnecessary. Sam Altman of OpenAI, sensing the backlash his own rhetoric has provoked, has begun reframing the technology as a tool to amplify human capability rather than replace it—though he still speaks of "significant disruption" and "transitions to new work." Yet despite these warnings from the people closest to the technology, the actual labor market shows no signs of cracking.
The unemployment rate across wealthy nations sits at roughly 5 percent. The OECD's working-age employment figures keep hitting new highs. The United States employs more people than ever in fields supposedly most vulnerable to AI, including law. The U.S. Bureau of Labor Statistics projects the country will add 5.2 million jobs between 2024 and 2034, a 3 percent increase in total employment. Economists, as a group, remain far less alarmed than the tech executives. They reject what they call the "lump of labor fallacy"—the idea that the job market is a fixed pie. When technology displaces workers from one field, the argument goes, it enriches others, who then spend their gains on goods and services that create new employment elsewhere.
But the real question is whether this time is different. If AI truly did throw millions out of work and keep them there, it would be unprecedented in human history. No previous technology has spread fast enough to create sustained mass unemployment. Understanding why requires looking back at how previous revolutions actually unfolded.
Robert Gordon, an economic historian at Northwestern University, examined growth rates across centuries. From 1300 onward, the most advanced economy of any given era never sustained per-capita GDP growth above roughly 2.5 percent annually. When other nations grew faster, they were simply catching up to a wealthier leader that had already pioneered the technology. This slow pace of innovation at the frontier also meant slow destruction of jobs. Take agriculture. Despite revolutionary changes in farming over a thousand years, the share of English workers in agriculture declined gradually from the 1500s onward—not in a sudden collapse. The modern tractor, invented in early 20th-century America, took generations to displace farm labor, not years.
Even rapid disruption need not harm workers. In the mid-20th century, early computers, shipping containers, and other innovations prompted British Prime Minister Harold Wilson to speak of the "white heat of technology" spreading through Western economies. American per-capita GDP grew at 2.5 percent annually—the fastest rate ever recorded for a leading economic power. Labor disruption, measured by the share of jobs shifting between sectors and occupations, sometimes exceeded double today's rate. Yet many remember that era with nostalgia: rising wages, expanding opportunity, less political division.
The Industrial Revolution in 19th-century Britain offers the most famous case of technological upheaval. James Watt's steam engine innovations in the 1760s-1780s made factories viable. What followed was explosive economic growth that seemed to coincide with stagnant real wages. Between 1790 and 1840, inflation-adjusted wages barely moved, even as capitalists accumulated vast fortunes. This period became known as "Engels' Pause," named after Friedrich Engels, a communist-turned-radical who documented the misery of Manchester's slums in the 1840s. Modern Silicon Valley leaders often invoke this pause as a warning. But recent scholarship challenges whether it's a useful model for what AI might bring.
The composition of British employment barely shifted until the 1850s, and then only at today's pace. More importantly, if technology destroyed jobs, it created far more. Between 1760 and 1860, the number of employed Britons jumped from 4.5 million to 12 million. Unemployment remained modest. Wage growth was indeed slow during Engels' Pause, but no slower than in the previous fifty years. The real culprit was slow productivity growth in the early Industrial Revolution—a direct result of how gradually Watt's innovations actually spread. In 1830, all of Britain used only 160,000 horsepower of steam energy, equivalent to about 1,000 modern cars. Given the rapid population growth of the era, it was, as one British demographer noted, "truly remarkable" that workers' purchasing power grew at all. When wages are measured not by consumer prices but by the average price of all goods produced nationally, the picture becomes even clearer: workers were paid fairly for what they produced. The real enemy was not the machines but politicians—food prices spiked due to war and high tariffs on grain imports, squeezing workers' budgets.
This reframes the industrial unrest of the era. Textile workers smashed mechanical looms; laborers destroyed threshing machines across southern England. Historians link this to technological disruption, but strikes and sabotage are as old as work itself. England actually saw fewer riots in the early 1800s—during Engels' Pause—than in the late 1800s, when real wages were rising sharply. The Chartists, who won voting rights and other protections for workers, gained momentum only after wage growth resumed in the 1840s. As economic historian Nicholas Crafts wrote plainly, the Industrial Revolution "is not a model" for technological change that boosts productivity while shrinking labor's share of national income. In other words, those warning of AI-driven mass unemployment are predicting something that has never happened before.
That does not mean it cannot happen. The warning signs would be unmistakable: productivity surging while real wages stagnate in the United States, the world's leading economy. This would show up as per-capita GDP growth exceeding Gordon's 2.5 percent ceiling, paired with soaring corporate profits—evidence that gains from increased output flow to capital, not workers. Another signal would be large-scale job losses spreading across many sectors simultaneously. History offers one final lesson: if such disruption comes, it will likely arrive during a recession. Recessions purge economies of unproductive work. Weak firms fail; capital and labor shift to stronger ones. Nearly every job that was once common in America has vanished during past recessions. The jobs that disappear in the next one will tell us whether this time is truly different. Until then, even Amodei, Gates, and Altman are simply guessing.
Citações Notáveis
The Industrial Revolution is not a model for technological change that boosts productivity while shrinking labor's share of national income.— Nicholas Crafts, economic historian
It was truly remarkable that workers' purchasing power grew at all, given rapid population growth and slow productivity gains in early industrialization.— Tony Wrigley, British demographer (paraphrased)
A Conversa do Hearth Outra perspectiva sobre a história
Why do you think the tech leaders themselves are so worried if history suggests we should be calm?
Because they're looking at the raw capability of the technology, not at how it spreads through an economy. They see what the AI can do in a lab and assume it scales instantly. History says it doesn't.
But couldn't AI be different? Couldn't it spread faster than steam engines or tractors?
It could. But if it does, we'd see it in the data first—productivity exploding while wages flatline. We're not seeing that yet. The labor market is still creating jobs.
What about all those people whose jobs disappear? Don't they suffer even if new jobs eventually appear?
Absolutely. The transition is real and painful for individuals. But the question the article is asking is whether we get mass unemployment that persists, like the warnings suggest. History says that's rare.
So you're saying Engels was wrong about the Industrial Revolution?
Not wrong exactly. He saw real suffering. But he misdiagnosed the cause. Workers weren't being exploited by machines—they were being squeezed by food prices and tariffs. The machines were spreading slowly enough that employment actually grew.
What would actually prove the doomsayers right?
Productivity growth above 2.5 percent a year combined with stagnant wages and job losses across multiple sectors at once. That would be genuinely unprecedented. We should watch for that in the next recession.