Nearly two-thirds of Americans fear that artificial intelligence will eliminate jobs over the next two decades. Their fears are well-founded. Dario Amodei, CEO of Anthropic, one of AI’s creators, recently warned that AI could wipe out half of all entry-level white-collar jobs, although it should be noted that Amodei has a vested interest in keeping AI front and center, as revenue for his and other AI companies has yet to come close to matching the venture capital invested.

Nevertheless, those likely to be affected as AI models get more sophisticated are junior software developers, paralegals, first-year law firm associates, among others. Their parents and high school classmates who went into blue-collar work, and lost their jobs thanks to globalization, are now the core of the MAGA populist rebellion that has transformed American politics. Unless something is done now to help white-collar workers navigate AI job displacement, their crushed dreams of a middle-class life may fuel a populism that will make the MAGA rebellion look like a skirmish in our political wars.

The challenge we face is to think outside the box about how we help people navigate the rapidly changing nature of work. A lack of imagination, daring and compassion was our failure in dealing with the job-shattering impact of globalization. For the sake of our democracy, we cannot afford to make that mistake again.

The first task is to throw some sand in the gears of the AI rollout.

Every effort should be made, both by companies and by the government, to slow job losses by finding ways to use AI to enable existing workers to improve their productivity and stay in the workforce, not be displaced by it.

An adaptation of the German Kurzarbeit (short-term work) program is an option. That plan involved the government subsidizing the lost or cut wages for employees in companies during distress, such as during the COVID-19 pandemic and the Great Recession. An adapted scheme could involve workers keeping their jobs, with a pay cut, while the government savings in unemployment benefits are used to help pay for retraining. 

The bigger task is to help those who have no future in white-collar employment to make the transition to something else. 

Colleges and universities will need to reorient their programs, mindful that many of the jobs for which they now train students — accounting, computer software development, law, some aspects of medicine — will no longer be needed to the same degree.

Students should be encouraged to pursue studies in fields less likely to be supplanted by AI. This may mean that more young people matriculate at community colleges, where they can learn skills to be plumbers, electricians, welders and other much-needed occupations not threatened by AI. 

And since AI won’t replace many human-facing social services, despite the proliferation of AI therapists and educational bots, retrained white-collar workers and students should be encouraged to move into nursing or physical therapy, environmental services or other human-centric work.

For older white-collar workers, the Tobacco Transition Payment Program, a government annuity created two decades ago to buy out tobacco farmers’ Depression-era production quotas, might be a model.

Such an annual payout — to replace not all, but some portion of, a worker’s previous income — could be offered to older AI-displaced workers. Such folks are unlikely candidates for retraining, will rapidly run out of unemployment benefits and will still need some financial cushion before they qualify for Social Security and Medicare. 

How to pay for these AI-displacement initiatives will be a major challenge.

The tobacco annuity was funded through assessments on tobacco product manufacturers and importers, not taxpayer money. 

The profits of AI companies could be taxed. Amodei has suggested that if someone uses an AI model and the company makes money, perhaps 3 percent of that revenue “goes to the government and is redistributed in some way.”

Since no AI company is currently making a profit, it might make more sense to require some percentage of their investments to be set aside to fund retraining or annuities for displaced workers.

The OpenAI Foundation, which has a $130 billion stake in OpenAI, could be required to devote some portion of its assets to helping those its technology has displaced, for instance.

All of these options are preferable to much-discussed proposals for Universal Basic Income (UBI). Experiments with UBI show mixed results, in part because money cannot replace the dignity of work.

Any financial support for AI-displaced workers — or UBI, if that becomes an option — should also include a community service component to give people a sense of self-worth while giving back to society, making it more politically acceptable to taxpayers. 

In the end, the biggest obstacle to preparing for AI’s assault on white-collar jobs may be Americans’ individualism.

Surveys have shown that most Americans feel it is the responsibility of individuals themselves to make sure they have the right skills and education to be successful in the future. But such sentiment didn’t serve their blue-collar parents and classmates well. And it may not prepare white-collar workers for the future.

We need to begin now to plan how society can help these Americans weather AI job losses. Our fellow citizens deserve no less. And U.S. democracy can ill afford a populist white-collar backlash in the years ahead.

Bruce Stokes is a visiting fellow at the German Marshall Fund.