Will Artificial Intelligence Turn Out to Be a Dream Killer?

Editor’s note: Rob Atkinson contributed the following perspective to The International Economy for a symposium of views on artificial intelligence featured in its Fall 2025 edition:

There is no issue more fraught with confusion than the impact of artificial intelligence on jobs. We see this in the statement we are addressing, “If productivity rates indeed soar to levels unimaginable as many AI proponents are suggesting, the part of America that owns stocks will enjoy an even greater bonanza while both blue and white-collar workers experience heavy job losses.”

Won’t happen.

First, despite what the apostles of artificial general intelligence warn, there is no reason to think AGI will get here anytime soon, if ever. To be sure, AI is a great tool for analyzing discrete information sets (for example, drafting legal briefs), but it can’t take care of toddlers in a daycare center.

And even with much better AI, there will still be massive amounts of work left. Does anyone think that self-driving school buses will not have an adult on them to watch the kids? What about police? AI robots will not be arresting criminals. What about game wardens, models, priests, stonemasons, plumbers, legislators, and flight attendants? Most of the economy involves working with people, things, or ideas so complex that AI shows no sign of being able to perform them.

Second, AI job-doomers commit the “lump-of-labor fallacy”: the idea that there is a limited amount of work to be done, and if a job is eliminated, it’s gone for good. But this fails to account for second-order effects whereby the savings from increased productivity are recycled into the economy in the form of higher wages, higher profits, and reduced prices, that all create new demand that in turn creates other jobs—some in new occupations (like “content creator assistant”), but most in existing ones that workers will now spend their savings on (such as personal trainers). This is why most scholarly studies find no net negative effect of productivity on employment.

And the idea that these savings will all go to profits is simply wrong. If employers could, they’d pay their workers ten cents an hour and charge outrageous prices. They don’t because of a thing called labor and product market competition. AI won’t repeal the laws of economics.

The idea that Americans will run out of things to buy is even more ludicrous. Even if by some miracle in fifty years AI boosts U.S. per capita income ten times to $600,000 per year, Americans would rightly still want more.

Given America’s chronically slow rates of productivity growth, the significant increase in retirees, and massive government budget deficits, America has to have higher rates of productivity. If we are really lucky, AI might boost productivity from around 1.5 percent per year to 3 percent.

But if AI leaders keep scaring the pants off of people, the response will not be to put the pedal to the metal of AI-driven productivity growth. It will be to slow it all down, and Americans will pay the price.

• Read the full AI symposium.
• Read The International Economy.