Robbie Hiser spent months teaching AI to do his job. He trained Mercor, a company that builds AI models like ChatGPT and Claude, on how to handle everything from writing code to handling finance. Now he watches as his peers pick up the same tools and do the same work themselves.
Hiser is one of nearly 100,000 freelancers Mercor has gathered across many industries to train AI. His work is part of a broader shift: experts are training AI on how to do their jobs, and the demand for that training is growing fast.
The Numbers Behind the Shift
Nearly three-fourths of Americans fear AI will take their jobs, according to a Pew Research study. A September McKinsey Global Institute report found AI could force 11 million U.S. workers into new careers by 2035.
Mercor CEO Brendan Foody calls the changes a natural evolution that frees workers from dull tasks. Freelance AI training at Mercor pays $20 to $200 an hour.
The work is spreading beyond software. Columbus, Ohio law firm Vorys, Sater, Seymour and Pease built AI personas trained on the experience and philosophy of some of its senior lawyers with Stanford’s Liftlab. Partner Kim Herlihy said AI can’t replace lawyers’ human touch in court.
Who Is Winning, Who Is Losing
The picture is mixed. AI training ranks as America’s fourth-fastest-growing job category on LinkedIn. But the gains are narrow compared to the losses.
Nobel Prize-winning economist Daron Acemoglu warns AI is moving faster than past revolutions. The Industrial Revolution took 80 years, he says. AI is changing in one or two years across many sectors.
Acemoglu’s worst-case scenario: unemployment could triple in the next decade if nothing changes. He acknowledges AI training creates jobs but calls the number employed small compared to those displaced.
Research from Stanford shows hiring is down for young people in AI-exposed jobs like software development. U.S. Census data found hiring and wages among recent college grads in AI-exposed majors have fallen due to AI.
Clara Shih, formerly of Salesforce and Meta, said AI now lets a few people do work that once required dozens. She quit her tech job in January to start a nonprofit aiding young workers affected by AI.
Shih warns AI is very good at tasks traditionally assigned to young people, such as market research and memo drafts. She says AI affects more than jobs, citing threats to people’s sense of self-worth and community.
What Comes Next
Acemoglu suggests companies design AI to improve humans rather than replace them, or governments could use tax policy to favor hiring actual people. He urges action before the window closes, saying the future depends on choice and agency.
The stakes are personal. Shih’s exit from Salesforce and Meta was not about ideology. It was about seeing AI take over work in general.
Hiser’s path is different. He chose to train AI himself rather than watch it replace him. But the two are not the same. One is adaptation. The other is survival.
The McKinsey report’s 11 million figure is not a guarantee. It is a projection of forced change. The difference between those two numbers — adaptation and forced change — is the space Acemoglu is fighting to preserve.
The source’s position is clear. Here is how the potential paths rank:
- Companies design AI to make humans better, not expendable.
- Governments implement a tax code that incentivizes hiring real people.
The source’s position is clear. Concentration of power in government and corporations is dangerous. AI is moving too fast for the market to absorb it on its own. Government intervention, through tax policy or regulation, could protect workers.
The question is whether anyone moves before the window closes.
Where the paper stands
The paper backs the freelancers like Robbie Hiser who are training AI and the small businesses building these models, and is against any rules that would raise the cost of entry for them, whether through licensing regimes, compliance costs, or rules written with the biggest firms’ help. The danger is not the technology itself but the concentration of power around it.
Hiser’s work — training AI on coding, finance, and other specialized tasks — shows what small-scale adaptation looks like. He is not waiting for AI to take his job; he is using it to keep doing his work. That kind of adaptation is exactly what the paper wants to preserve.
What the paper is against is the kind of broad regulation that hands the market to the incumbents. Licensing regimes and compliance costs only giants can afford are a moat, not a safeguard. Rules written with the biggest firms’ help tend to protect the biggest players and raise the cost of entry for everyone else. The paper wants narrow rules against direct harm — forcing companies to disclose safety failures they hid — and nothing more.
See the video the story is built around at CBS News.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.