Fascinated and alarmed, economists and policymakers are debating how AI — and especially much anticipated artificial general intelligence, or AGI — will reshape the workforce. Techno optimists argue that technology has historically been a powerful driver of economic growth, spurring new industries with novel jobs. That’s what happened with the advent of the automobile, after all: The ranks of carriage makers, horse breeders and stable owners melted away as jobs opened up in the emerging oil industry, and then in brand new sectors like motor hotels — motels — and drive-in theaters. Why couldn’t the same happen with AI?

But others hold that the changes wrought by AI are of a different scale. International Monetary Fund economists have estimated that AI may affect as many as 40 percent of all jobs as AI-driven machines replace work that was traditionally performed by people, much of it skilled. And even where jobs aren’t lost, work by human beings could become less valuable, causing wages to fall, says Anton Korinek, an expert on the economics of AI at the University of Virginia.

With such worries widely felt, it is perhaps no coincidence that two of the three recipients of the 2024 Nobel Prize for economic sciences have written extensively about artificial intelligence and its potential impact on jobs. MIT’s Daron Acemoglu and Simon Johnson argue that we must act deliberately to ensure that AI’s benefits are shared widely — through government intervention, bold new policies and reskilling programs to avoid deepening inequality and societal unrest in this age of growing automation.

The two economists advise that, as we navigate this moment, we heed lessons from the past — specifically, the early Industrial Revolution, another time of economic and social upheaval, and the flexible thinking of a key figure of that time: Englishman David Ricardo.

Acemoglu, Johnson and their colleagues say that if, like Ricardo, policymakers act with care and flexibility, AI might even help restore what the start of the tech boom put in jeopardy: decent-paying middle-class jobs.

Changes hit home

Ricardo, born in 1772, was a parliamentarian and noted economist of his time. In his younger years, he was a techno optimist of sorts. He believed that new spinning machines that converted raw cotton into yarn were going to increase worker productivity and prove to be beneficial for everyone — workers, entrepreneurs and the public.

The new machinery might initially displace some home-based spinners, he recognised, but eventually those people would find work elsewhere.

And that’s what happened — at first. Cotton textile manufacturing boomed during Ricardo's lifetime: The new spinning machines developed in the 1770s made producing yarn faster and cheaper. Workers who had spun yarn at home on spinning wheels were disrupted by these new machines, but many were able to transition into another growing cottage industry — weaving the now more abundant and cheaper yarn into cloth.

The first edition of Ricardo’s Principles of Political Economy and Taxation, published in 1817, makes no mention of the potential ill effects of machinery on workers. Indeed, in an 1819 speech before the English House of Commons, he declared that “machinery did not lessen the demand for labour.”

But a different reality was emerging with a second invention: power looms, introduced roughly a generation after the spinning machines. A single power loom could produce more cotton than 10 to 20 handweavers working from home, and the machines were so large that they had to be housed in factory buildings, taking cottage industry weaving off the table. As factory weaving eclipsed home weaving, this time the displaced workers had no place to go, because power looms created relatively few new jobs in the factories.

For home-based weavers, this was a disaster. Family earnings for handloom weavers in two Lancashire towns fell by half over a five-year period starting in 1814, Acemoglu and Johnson recount in a 2024 article in the Annual Review of Economics. Handloom workers in the English cotton industry overall averaged 240 pence per week in 1806, but by 1820 — around the time Ricardo was making his speech in the House of Commons — they were making less than 100 pence weekly.

Even the factory workers who had jobs tending the powerful new textile looms weren’t faring well. They experienced little real wage growth between 1806 and 1835. Growing wealth inequalities spawned social unrest, especially in the hard-hit industrial north. A major demonstration, with an estimated 60,000 people clamoring for political reform, was broken up by deadly force in Manchester in August 1819, in what is known as the Peterloo Massacre.

Ricardo, who had witnessed firsthand the consequences of power looms in the cotton industry, radically changed his mind. A more nuanced view of mechanization found its way into the third edition of Principles, published in 1821. He inserted a whole new chapter to discuss machinery’s impact, writing, in what amounted to a recantation: “The same cause which may increase the net revenue of the country, may at the same time render the population redundant, and deteriorate the condition of the labourer.”

'Middle skill' fortunes fall

Ricardo’s pivot holds important lessons for today’s labor economists and policymakers as they wrestle with the impact of AI’s impact on jobs and wages, Acemoglu and Johnson say.

But this is not a problem that began with artificial intelligence. Rather, it has been brewing since the 1980s with the proliferation of digital technologies that were soon to dominate so much of business, commerce and society — computers, the internet, smartphones, e-commerce, social media. Just as with power looms at the start of the Industrial Revolution, these failed to spread profits to ordinary workers. Instead, Johnson says, what they’ve done “is help the most skilled or the most educated.”

Since the 1980s, meanwhile, the wages of “middle skill” people — such as those who didn’t attend four-year colleges — stagnated or fell in real terms. The tech revolution “automated away a broad middle-skill stratum of jobs in administrative support, clerical and blue-collar production occupations,” writes David Autor, a labor economist at MIT, in an article in the magazine Noēma. Like the spinners and weavers of yore, those people saw their livelihoods melt away. The result was a hollowing out of the middle class, Autor writes, forcing many to turn to less-skilled, lower-paying work.

Indeed, while average household income in the United States increased by 95 percent after adjusting for inflation over the past four decades, according to the Peter G. Peterson Foundation, a nonprofit focused on the United States’ fiscal health, those gains varied widely among top earners and lower-income groups. In the highest income group, income rose 165 percent between 1981 and 2021, whereas among the lowest group, it grew only 38 percent: The wealthiest group did more than four times better than the lowest income group.

How AI can help

And now comes AI. Will it worsen the trend — or, if it’s handled right, might it present an opportunity? Johnson, Autor and Acemoglu express optimism. “AI — used well — can assist with restoring the middle-skill, middle-class heart of the US labour market that has been hollowed out by automation and globalization,” Autor writes. By training individuals in the application of AI software, middle skill workers might be able to take on many decision-making tasks previously assumed by doctors, lawyers, software engineers and even college professors. An experienced medical worker, for example, could master a new medical device, like the use of a new type of catheter, or could carry out an unfamiliar procedure during a medical emergency.

In the health care sector, “there are so many expert decisions to make, so much technology in use, and so many rigorously trained care providers who are not MDs but who could likely perform at a higher level — that is, do higher-stake tasks — with better support tools,” Autor says. Lung ultrasound, for example, is a promising technology for diagnosing patients with shortness of breath. Although it’s typically performed and interpreted by medical doctors, in a recent study health care clinicians not experienced in lung ultrasound “were successful in obtaining high-quality lung ultrasound clips using an AI tool,” Autor says.

Elsewhere, Autor adds, the HVAC industry (heating, ventilation and air conditioning) is drawing on AI tools that provide real-time guidance and remote diagnostics to enable HVAC technicians to solve more complex problems. Technicians who are already on-site in a home or building can use a ChatGPT-type AI tool for troubleshooting problems using a flowchart. Earlier, they might have had to reach an expert by phone for help.

A pre-AI analogy to these kinds of changes would be the rise of nurse practitioners, says Johnson. “There weren’t very many nurse practitioners 30 years ago. And now every pediatrician’s office has them. And parents are very grateful for the advice and access that they can get, because they have these highly skilled people who do not have MD degrees, but they are enabled — they have been empowered by their licenses.”

The idea isn’t to replace doctors, Johnson adds. “We’re just saying: Redress that balance. Boost the people who don’t have four years of college. Enable them to be more productive.”

Not all economists think this sort of solution will work with AI, though, especially once AGI, with human-level cognitive abilities, arrives on the scene. “The key distinction from past technological changes is that AGI would be capable of performing any cognitive task, potentially leaving few unique economic roles for human workers,” Korinek says. “This could lead to widespread labor displacement and significant wage declines” — unless governments intervene to avoid further widening of the wealth gap.

In the short term, Korinek says, this might mean upskilling workers in roles requiring still uniquely human abilities like “authentic human connection, emotional intelligence and ethical oversight,” where machines don’t excel or don’t satisfy a human need. Examples could include psychotherapists, childcare providers and religious counselors.

Further down the road, societies might need larger shifts, such as sharing prosperity through the implementation of a universal basic income: no-strings-attached regular cash payments from governments that citizens receive whether or not they work.

Guidance from the past

There are few convincing studies at this point on the impact that AI, including AGI, will have on global jobs and wages. Partly this is because researchers have to take into account jobs that don’t yet exist in industries not yet invented. As we wait for this new world to unfold, however, the life and times of David Ricardo might offer some food for thought.

The new machinery introduced to English textile factories in the early 1800s had a much worse impact on workers’ lives than Ricardo ever anticipated. It took him a long time to realize this. When he did, he was nearly 50 years of age, wealthy, well-connected and at the top of his profession. It would have been easy to ignore the untoward effects from these amazing machines, the power looms. Instead, Ricardo admitted his mistake and modified his theories. That kind of fluid thinking, and intellectual integrity, will be crucial in the age of AI, Johnson says. “We have much to learn from Ricardo’s openness to new ideas and new ways of thinking about economics,” he and Acemoglu write in their paper.

England did learn from the handloom weavers’ collapse, and the travails of its laboring class generally. It expanded the political voice of its industrial cities. Before 1832, cities like Manchester weren’t even represented in Parliament. England’s Factory Act, passed in 1833, introduced what was arguably one of the world’s first enforceable child-labor factory regulations. (It required factory inspections.) The government repealed its protectionist Corn Laws in the 1840s, which made food more plentiful and cheaper for the urban laboring class. Ricardo, a free trade advocate, was a powerful voice urging the Corn Laws’ repeal.

The new world is hurtling toward us. Daniel Kokotajlo, former governance researcher at ChatGPT developer OpenAI, who now heads the AI Futures Project, has posited a scenario in which AGI is fueling an economic boom by 2027 — but is also causing millions of job losses because its software is outperforming humans in coding, research and other cognitive tasks. (“It might take a few years longer than 2027,” he says now.)

Other predictions differ — they range from a world in which more and more wealth becomes sequestered with the few, to one in which AI helps to reduce inequalities. It depends a lot on how our governments elect to act. “All the smarts, all the talent around computer science is being drawn into the pursuit of AI, to a degree that I haven’t seen since the internet really boomed at the end of the 1990s,” Johnson says.

But, he adds, as a look at the industrial revolution and history more broadly makes clear, “just because you have new miracle machines does not mean most people will benefit.”

This article was first published on Knowable Magazine.