AI Training Boom Raises Stakes for American Workers' Future

Robbie Hiser teaches artificial intelligence how to write music. He's a music producer who has spent months working with Mercor, a company that pays freelancers to train AI systems like ChatGPT and Claude on the specifics of their trades. Hiser has shown the machines the techniques he uses, the patterns he recognizes, the instincts he's developed over years of work. And yet he sleeps fine at night. "AI doesn't have any life experience to pull from," he said. "An AI never had a girlfriend break up with it." He believes that genuine creativity—the kind that moves people—will always require a human being on the other end.

Mercor has assembled more than 100,000 freelancers across dozens of industries to do exactly what Hiser does. They teach machines their expertise, and they're paid for it: anywhere from $20 to $200 an hour, depending on the complexity of the work. The company's CEO, Brendan Foody, sees this as the natural shape of the future. As AI handles the repetitive, grinding parts of work, he argues, humans will be freed to do the parts that matter—the parts that require judgment, intuition, creativity. AI training has become America's fourth-fastest-growing job category on LinkedIn. To Foody, it's not a threat. It's a career.

But the numbers tell a different story. A McKinsey Global Institute report released in September found that AI could force 11 million U.S. workers into new careers by 2035. Nearly three-quarters of Americans now fear that AI is coming for their jobs, according to Pew Research. And the evidence is already visible in the data: hiring for young people in AI-exposed fields like software development has declined. Wages and hiring among recent college graduates in AI-heavy majors have both fallen. The disruption isn't theoretical anymore. It's happening now, to people trying to start their careers.

Clara Shih watched this unfold from inside the machine. She oversaw major AI divisions at Salesforce and then Meta. She saw teams shrink. Tasks that used to require dozens of people—conceiving a product, building a prototype, delivering it to customers—now get done by a handful of people using AI agents. The work that typically went to junior employees, the entry-level assignments that taught people how to think and work, is now something AI does better and faster. Shih became so alarmed by what she was witnessing that she quit her job in January and started a nonprofit to help young workers navigate what comes next. "Traditionally, the tasks that you might assign to a young person, maybe doing some market research or creating the first draft of a memo, AI is just really good at doing those now," she said.

Some in the field push back against the darker predictions. Kim Herlihy is a trial lawyer and partner at Vorys, Sater, Seymour and Pease, a law firm in Columbus, Ohio. Her firm worked with Stanford University's Liftlab to build AI personas trained on the experience and philosophy of senior lawyers like her. These AI systems can offer feedback on documents, drawing on the unique perspective of individual partners. But Herlihy is careful about what she claims. "AI is never going to walk into the courtroom and give an opening statement to the jury," she said. "But it helps us have additional tools to use as we prepare that opening statement." The AI is designed for optimization, she insists, not replacement. "AI cannot take the place of what's a fundamentally human business. You call your lawyer because you want to look in their eye and say, 'What should I do about this issue?'"

Daron Acemoglu, a Nobel Prize-winning economist, sees the situation differently. He believes the current trajectory leads to mass displacement. "What we are living through with AI is unprecedented," he said. "The first phase of the Industrial Revolution took 80 years. Today it's taking place in one or two years and it's taking place across many different sectors at the same time." In a worst-case scenario, he warned, unemployment could triple in the next decade if nothing changes. He acknowledged that some people will find work training AI, but the math doesn't work in their favor. "The number of people they're employing in these positions is small relative to the number of people who will be replaced," he said. "I mean, that's the whole point of automation."

Shih draws a parallel to something she witnessed firsthand. Her family moved from Hong Kong to Ohio, and she watched factory work disappear. She sees the same pattern unfolding now, but this time it's white-collar work in the crosshairs. "I think what we can learn is that first we have to be honest with people about what the risks are," she said. "I think, you know, sugarcoating that this AI utopia, maybe we'll get there, but there's a lot of hurt that could happen between now and there." The cost isn't just economic. "It's never just about the jobs. It's people's sense of self-worth. It's the fracturing of communities."

Acemoglu doesn't argue for rejecting technology. Instead, he points to choices that still exist. Companies could design AI to make humans more capable rather than obsolete. Governments could use tax policy to incentivize hiring real people instead of replacing them. But he emphasizes that the window to act is narrowing. "It should move us into thinking about what we can do because there is a lot we can do," he said. "Because the future is so uncertain, because the future is so much subject to our choice and agency, it is all the more important that we have this discussion." The decisions made in the next months and years will shape whether AI becomes a tool that enhances human work or a technology that leaves millions of workers behind.