In recent decades, and especially since the release of OpenAI’s large language model ChatGPT in 2022, artificial intelligence use has rapidly spread into almost every industry. And as AI proliferates, so do fears that a wave of mass unemployment will follow. Concerns are widespread that AI will be deployed to accomplish an ever-greater share of the labor needed throughout the economy, leaving fewer and fewer jobs available for human workers.
This fear led Dario Amodei, one of the world’s leading AI technologists, to sound the alarm earlier this year about an impending “white-collar bloodbath.” The Anthropic CEO told Axios that one very possible scenario within the next one to five years is that, “Cancer is cured, the economy grows at 10 percent a year, the budget is balanced — and 20 percent of people don’t have jobs.”
This was predictably latched onto by economic interventionists, such as US Senator Bernie Sanders. “We must demand that increased worker productivity from AI benefits working people, not just wealthy stockholders on Wall St,” Sanders posted on X in response to Amodei’s statement. Sanders later posted that, “With the explosion of AI, new technology and increased worker productivity, we should demand a shorter work week, increased life expectancy and a decent standard of living for all.” Other politicians have gone even further since then. US Senator Josh Hawley, like Tucker Carlson started advocating years ago, supports banning self-driving cars to protect the jobs of car and truck drivers.
Indeed, AI has already created countless efficiencies and new capabilities throughout the economy that have made specific jobs redundant. Consider the global herbicide industry. A Bloomberg article titled “AI-Powered Weed-Killing Robots Threaten a $37 Billion Market” explains:
After almost a century of deploying a more-is-more approach to chemical herbicides, the global agricultural sector is rapidly rolling out advancements that promise to curb the use of weed-control sprays by as much as 90 percent. Using artificial-intelligence powered cameras, the new sprayers can identify and target invasive plants while avoiding the cash crops. If even a fraction of growers adopt the new tools, it could mean a big shift for crop-chemical majors like Bayer AG and BASF SE.
This potential tenfold improvement in herbicide use efficiency may significantly reduce the agricultural demand for herbicide production, and thus put many people out of their jobs in that industry. And this is just one example. From facilitating new scientific research methods, to transforming solar cell production, to improving healthcare safety, to proliferating self-driving car use, AI offers new ways of doing things throughout every industry, which devalues some skillsets in the job market and rewards others.
However, there is an opposite fear as well. Last month, during a press conference the day he won the 2025 Nobel Prize in economics, Joel Mokyr gave a speech about the grand sweep of economic history. Toward the end of his lecture he weighed in on the AI job loss debate: “As I see it, the main concern about the labor market is not technological unemployment. It’s labor scarcity.”
Mokyr is an economic historian who is renowned for looking at the big picture. And when you do that, it is clear that technological unemployment is nowhere near as big a concern as the set of problems that AI will help solve if politicians like Bernie Sanders and Josh Hawley don’t succeed in debilitating it.
Specific jobs becoming obsolete is only one side of the coin of technology’s effect on employment. The other side of the coin reveals why artificial intelligence will lead to no overall reduction in employment opportunities. Like all prior technological revolutions, AI will create at least as many jobs as it eliminates, and most importantly, the new jobs will tend to be preferable to and easier to find than the old jobs.
Unemployment rates have remained relatively constant and quite low in recent decades, currently standing at about 4.3 percent in the US and 4.9 percent across the globe, according to recent estimates. This is approximately as low as they have ever been.
Meanwhile, ever since the industrial revolution, there has been an explosion of technological advancement and proliferation that has almost completely transformed all economic industries and everyday life across the globe.
Virtually every instance of technological progress reduces the need to employ some specific form of labor. Dish washing machines have meant that homes and restaurants need only employ a small fraction of the cleaning staff that was once required per meal served. Refrigeration has conserved countless labor hours preserving foods. Electric power has exponentiated the manufacturing capability of each factory worker by enabling a new class of machinery. And the list goes on and on.
The constantly low unemployment rate, in light of recent history’s technological progress, would amaze the Industrial-Revolution-era activist group known as the Luddites, a radical labor movement of anti-technologists who rioted and destroyed factory machinery to protect their jobs from automation in the early 19th-century. Their initial concern was for the jobs of craftsmen competing with mechanized looms and knitting equipment that allowed a single worker to produce the output of a hundred craftsmen. But the scope of Luddite animosity quickly widened to oppose industrialization generally.
The nineteenth-century Luddites made the same mistake, known to economists as the “lump of labor fallacy,” that antagonists of AI are making today. In a New York Times article titled “Trump, Immigration and the Lump of Labor Fallacy,” Nobel Prize winning economist Paul Krugman explains the fallacy: “This is the view that there is a fixed amount of work to be done and that if someone or something — some group of workers or some kind of machine — is doing some of that work, that means fewer jobs for everyone else.”
Krugman also explains why this “lump of labor” view is false:
When incomes rise, people will find something to spend their money on, creating jobs for workers displaced by technology or newcomers to the work force. Machines do, in fact, perform many tasks that used to require people; output per worker is more than four times what it was [in 1952], so we could produce 1952’s level of output with only a quarter as many workers. In fact, however, employment has tripled. … No, AI and automation, for all the changes they may bring, won’t ultimately take away jobs, and neither will immigrants.
The point about “incomes rising” is key, because it is a predictable consequence of technological automation and it is also the reason, along with new forms of work created by new technologies, why the newly created jobs will tend to be preferable to the old jobs.
Workers will only be displaced by technology if the technology is capable of producing more valuable output, meaning some net-positive combination of cheaper and higher quality, than human labor. Otherwise, mechanization won’t be profitable and employers will stick to employing humans. Therefore, we can safely bet that wherever AI is displacing human workers, consumers are benefitting from some net-positive combination of falling prices and rising quality. This means that consumers will have more income left to spend, and will therefore consume an ever-wider range of goods and services, thus funding more and more niche forms of employment that replace the jobs lost to technological improvements in the production process.
This is crystal clear if you look at the history of labor specialization. Since the mid 19th century, agriculture has gone from employing about 70 percent of US workers to employing only 2 percent. Meanwhile, the productivity of US agriculture has massively increased due to improved farming technology and food has become about 10 times more affordable for blue-collar workers in the last century.
If no new jobs had been created in America since 1840, this process would have left 68 percent of the population permanently out of work. But to the contrary, the modern economy is composed largely of jobs that people in any prior century probably never even imagined. Web designers, pet therapists, dietitians, travel agents, nail technicians, cosmetic surgeons, neuroscientists, and countless others are employed in now-common forms of labor that would have seemed either laughably frivolous or entirely unimaginable just a century ago. “Really, therapy for my dog while my children are starving?”
For a detailed description of one of the most recent and frivolous new industries imaginable, read the article titled “Does Your Plant Need a Nanny? A new crop of caretakers will spritz, polish and prune your houseplants — and even send photos while you’re away.” published by the New York Times in April. These new “plant nannies” offer virtual plant visits, and services such as speaking to, lighting, and photographing plants while their owners are away, in addition to traditional plant maintenance practices.
New and ever-more-niche and personalized forms of work are constantly made possible by the vast abundance of labor and consumption freed up by labor-saving technology. Monotonous or backbreaking tasks are left to machines while humans are freed to invent new tasks to pay each other for with all the wealth that technological efficiencies have saved or created for them.
As amazing as the present is compared to the past, the future can be an even more extreme improvement. If AI proves powerful enough to continuously transform the macroeconomy as the fearmongers predict, and so do I, this will mean such massive productivity gains that machines will eventually be running the farms and factories at almost no cost. This will render necessities more and more affordable until almost every worker can engage in more niche artforms, services, and research areas than employ almost anyone full-time in today’s economy. In addition, whole new realms of technological possibility can be unlocked, in outer space, the Metaverse, and elsewhere.
It is hard to imagine ever reaching a logical conclusion of this process, at which artificial intelligence is better than human labor for literally every purpose. That would even involve convincing those who prefer some services to have a “human touch” for sentimental, aesthetic, philosophical, or spiritual reasons that human labor is unhelpful even to them. If AI does get that powerful, that will be an end of material scarcity in which everything is free and worrying about “unemployment” will make no sense at all.
Instead of worrying about that hypothetical utopia of widespread godlike agency and infinite abundance, let’s focus on reducing human drudgery and increasing material wealth by proliferating artificial intelligence and other technological progress in the here and now.
