AI “Godfather” Geoffrey Hinton warns that his invention will create massive unemployment and send profits soaring — “that is the capitalist system.” Within five years, Dario Amodei, CEO of AI company Anthropic, predicts that half of “administrative, managerial and tech jobs for people under 30” could vanish.

Yet throughout history, automation has not translated into a permanent rise in unemployment or into a lasting shift of income from labor to capital. Producers capture only a small share of the total returns from technological change. Disruption raises, not lowers, overall living standards.

Look at several Western European countries over the period of 1870-1998. Massive industrialization generated an 18-fold gain in productivity per hour worked. Incomes soared. The work week is shortened, increasing leisure. Employment stayed steady. Higher productivity raised real incomes and demand, absorbing labor into new tasks and sectors even as old ones vanished.

Nor did the rich owners profit at the expense of workers. During the Industrial Revolution from 1770 to 1860, British economist Nicholas Crafts found that the labor share of income versus the capital share remained almost unchanged, rising from 61% to 62%. During the second half of the 20th century, Nobel Prize-winning economist William Nordhaus concluded that producers captured “only a minuscule fraction of the social returns from technological innovation.”

Some may argue that this time may be different — that AI really is going to take all the jobs. But a decade ago, Geoffrey Hinton predicted the demise of radiologists, arguing that “deep learning is going to do better than radiologists.” Today, demand for radiologists is rising, not falling. Hospitals report shortages. A radiologist’s job consists of many tasks. AI may complement or potentially automate some, driving the value of, and demand for, radiology up.

Another concern is the weak job prospects for new college graduates. A much-quoted paper, “Canaries in the Coal Mine?”,  blamed AI. But other research disagreed, saying a tight labor market was not confined to graduates. Little evidence existed to support “large-scale AI-driven displacement of early-career knowledge-sector jobs,” says John Burn-Murdoch, a data economist.

No smoking gun exists. Even if we accept that entry-level jobs could come under pressure from AI, demand for “AI natives” is bound to rise. Tech giant Shopify, for example, is “hiring more interns because they’re the ones who are using AI in the most interesting ways.”

AI has emerged against the backdrop of declining economic dynamism. Despite the digital revolution, productivity growth and the creation and destruction of firms and jobs have declined in advanced Western economies over the past two decades.

AI may help return us to past rates of progress (and disruption) rather than usher in an unusually disruptive era. A bottleneck bind will require human oversight. These “bottlenecks” include tasks subject to occupational licensing, which stipulate human credentials to perform certain tasks, such as prescribing a medicine. Although some requirements should and will be modified to permit AI alternatives, occupational licensing applies to a wide range of tasks, and efforts at reform will meet resistance.

Deeper reasons than licensing also suggest that AI might not substitute for all desk-based work. First, human preferences ultimately matter, and we will continue to hold humans accountable when things go wrong.

Second, work often involves bargaining, and its counterparts’ withholding information and deception. If we insist that AI not deceive humans, it would not be able to successfully bargain on our behalf. During a recent Cheeky Pint podcast, Elon Musk argued that AI should be “truth-seeking” and not say things it doesn’t believe.

Third, to pursue socially beneficial goals, particularly when exploring new ways of doing things, people may bend or break the rules. Sometimes we turn a blind eye, sometimes we change the rules, and sometimes we punish transgression. AI is unlikely to be afforded such discretion.

Humans can productively deceive others, bend or break the rules when the payoff from doing so is large (and not simply zero-sum), and can be punished if they err. It is a delicate dance. Progress depends on it, and it seems most unlikely we will permit AI to perform such tasks. To err is human.

I expect economically productive work for humans to persist. What we need now is more innovation and growth — not less — and AI offers that prospect. We should embrace the new technology and accept that it will both create jobs and destroy them.

Brian Williamson is a partner at the London-based Communications Chambers consultancy. He works at the intersection of technology, economics, and policy. Clients have included governments, regulators, telcos, and tech companies. This article is adapted from a just-published paper, AI, Copyright and the Public Good.

Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict intellectual independence policy across all its projects and publications.

The post Will AI Destroy Every Job? appeared first on CEPA.