Top labor economist says don’t believe the AI doom narrative—and his reason why is the ‘underpopulation crisis’ Elon Musk talks about

Last summer, Tesla CEO Elon Musk said declining birth rates were “the biggest danger civilization faces by far.” In separate comments, Musk said at the time that he was “doing [his] best to help curb the underpopulation crisis,” in response to a Business Insider report that he had secretly fathered twins with a top executive at his brain-implant technology company, Neuralink.

But it’s very unlikely that AI will put everyone in the U.S. out of a job, argues MIT labor economist David Autor, and although he didn’t touch on Musk specifically, he was describing the same undeniable reality. “Barring a massive change in immigration policy, the U.S. and other rich countries will run out of workers before we run out of jobs,” Autor writes in the paper, “Applying AI to Rebuild Middle Class Jobs,” recently issued by the National Bureau of Economic Research. 

It’s simple demographics, Autor writes. With birth rates across the industrialized world and China plummeting well below the roughly 2.1 children per woman needed to keep a population steady, large parts of the world are facing a severe worker shortage, he argues. 

“The industrialized world is awash in jobs, and it’s going to stay that way,” Autor writes. “This is not a prediction, it’s a demographic fact. All the people who will turn 30 in the year 2053 have already been born, and we cannot make more of them,” he writes.

Employed, but is it good?

But simple employment isn’t a guarantee of economic well-being, Autor notes (as anyone observing the downward mobility of the American worker since the 1970s can attest.) That’s where policy comes in. AI, Autor claims, has the potential to reverse the impact of the past 40 years of automation, in which computers have devalued manual labor and increasingly rewarded knowledge—and he claims that AI can even grow middle-class, well-paid jobs. 

The key is to leverage AI’s ability to increase human “expertise,” he argues. Pre-AI, computers made information cheap and easily accessible, which boosted the value of expertise provided by highly paid professionals (e.g., doctors, lawyers, educators). They had “faultless and nearly costless execution of routine, procedural tasks” along with an “inability to master nonroutine tasks requiring tacit knowledge,” Autor writes. AI, while in its infancy, does precisely the opposite. Generative AI can riff on existing pictures without being specifically trained, and follow instructions without knowing all the rules. “If a traditional computer program is akin to a classical performer playing only the notes on the sheet music, AI is more like a jazz musician—riffing on existing melodies, taking improvisational solos, and humming new tunes.”

This capability means AI can be used to create better, fairer jobs—by disrupting the very top of the elite workforce and giving everyone else a leg up, Autor argues. 

In Autor’s telling, today’s “modern elite experts such as doctors, architects, pilots, electricians, and educators” are the modern version of the artisans who were put out of work by the mass-mechanization during the Industrial Revolution. Like their 18th-century predecessors, today’s experts spend years learning their craft in a type of apprenticeship, and then “combine procedural knowledge with expert judgment and, frequently, creativity to tackle specific, high-stakes, and often uncertain cases.”  

The growing importance of “experts” is one reason that the cost of education and health care have ballooned some 600% and 200% over the past 40 years, Autor says. And he’s advocating for AI to take a chunk out of those elites’ judgment and help lower the cost of living for everyone else. 

“By providing decision support in the form of real-time guidance and guardrails, AI could enable a larger set of workers possessing complementary knowledge to perform some of the higher-stakes decision-making tasks currently arrogated to elite experts like doctors, lawyers, coders, and educators,” Autor writes. “This would improve the quality of jobs for workers without college degrees, moderate earnings inequality, and—akin to what the Industrial Revolution did for consumer goods—lower the cost of key services such as health care, education, and legal expertise.”

Consider the nurse practitioner

To show how that could work, Autor used the example of nurse practitioners. The job of NP was basically invented in the 1960s to stave off a looming physician shortage; NPs receive additional training on top of a nursing degree to be allowed to run and interpret medical tests, diagnose patients, and issue prescriptions—tasks that were once exclusive to doctors. Aided by the growth of technology, including electronic medical records and communications tools, this job has increased 40% in the past 20 years, with the median NP in 2022 earning $125,000, or 50% more than the median household income. 

It’s not likely that AI will make experts superfluous, Autor argues, since it is only a tool, like a chain saw or calculator, and “tools generally aren’t substitutes for expertise but rather levers for its application,” he writes. (Take a pneumatic nail gun, for instance—it’s indispensable for a professional roofer and a looming injury for an amateur.)

But AI can give a trained worker a leg up to do their best work while minimizing the drudgery, Autor says. “AI used well can assist with restoring the middle-skill, middle-class heart of the U.S. labor market that has been hollowed out by automation and globalization,” he writes. 

To be sure, the technology is still in its early stages, and governments will need to make policy to protect existing workers from unscrupulous applications of the technology (and perhaps even overeager cost-cutters in company leadership).

But as proof that it’s not just a pipe dream, Autor offers three recent studies that compare workers with and without AI assistance. Programmers using GitHub Copilot, an AI coding assistant, are more than 50% faster than those without, according to one study. Another NBER working paper found that AI helped customer-service agents be more productive and attain experience faster (and helped them stay on the job much longer than previously). And a study in Science that compared professional writers (marketers, grant writers, consultants, and others) using ChatGPT found that AI helped writers at all levels. “While the best writers remained at the top of the heap using either set of tools, ChatGPT enabled the most capable to write faster and the less capable to write both faster and better,” Autor writes. 

Governments should embrace the AI-assisted future to allow more workers to regain “stature, quality, and agency” in their work, which has eroded over the past 40 years, he argues. If AI instead accelerates the race to the bottom, the results could be devastating, creating a world where everyone has a job but no one has agency.  

“A future in which human labor has no economic value is, in my view, an ungovernable nightmare,” he writes.