Doctors and Lawyers Join AI’s $400-an-Hour Gig Economy

Summary

AI training work is drawing highly paid professionals alongside the lower-paid generalists who have long handled data annotation. Recruiters are seeking software engineers for up to $120 an hour, condensed-matter experts for $170 and civil-rights attorneys for $400, while Mercor says its average pay is $110 an hour without disclosing how many tasks receive that rate. Radiologists such as Anu Brixey label medical images on weekends, partly because they believe better training data could improve patient care. The shift reflects companies’ search for private, specialized data after relying heavily on information scraped from the public internet. It also raises concerns that models may absorb professional expertise and eventually reduce the value of entire occupations. Researchers disagree about the scale of the change: Antonio Casilli says experts remain a small supplement to a workforce still dominated by precarious, low-wage workers, while MIT’s Danielle Li says a small group of data workers could substantially devalue a profession. Platforms are increasingly treating human work as an input that can be rented ad hoc or converted into company-owned model capabilities, and some companies are exploring employee activity data as training material. The article concludes that human workers remain essential to current models, even as the market reorganizes around their specialized knowledge.