Economic uncertainty is continuing to have devastating effects on the availability of jobs. Last year, the US labor market reeled from slowing wages, layoffs, and a notable lack of hiring, leading to the highest unemployment rate in the country in four years toward the end of 2025.
And while debate swirls about whether AI is actually replacing jobs in any serious numbers, many tech startups are trying to make it a reality. As the Wall Street Journal reports, a buzzy San Francisco-based AI company called Mercor is hiring desperate job-seekers for a particularly ghoulish task: training AI models to one day do the work they used to do.
It’s a depressing new reality as concerns over AI replacing jobs en masse continue to grow. Late last year, computer scientist and AI “godfather” Geoffrey Hinton predicted that AI would continue to “replace many, many jobs” in 2026 as the tech “gets even better.”
An MIT study also found last year that more than 20 million Americans’ work can be replaced with today’s AI, representing $1.2 trillion in wage value.
Paying those who are already struggling to find work in a disastrous job market to train their future replacements is a twisted new reality in the age of AI, leading to plenty of dark humor.
“I joked with my friends I’m training AI to take my job someday,” 30-year-old video editor Katie Williams, who has been captioning and rating video clips for Mercor for six months, told the WSJ.
Automotive journalist Peter Valdes-Dapena, who was laid off in 2024, has been critiquing AI-generated news articles for Mercor. The irony doesn’t escape him, but he rationalizes it away.
“I didn’t invent AI and I’m not going to uninvent it,” he told the newspaper. “If I were to stop doing this, would that stop it? The answer is no.”
Mercor hired tens of thousands of contractors last year after signing partnerships with AI industry stalwarts including OpenAI and Anthropic. Job security and a stable income appear to be hard to come by, with the company suddenly firing thousands of data labelers last year, for instance — only to hire them back on for a similar project, but for considerably less pay.
As a spokeswoman told the WSJ, contractors need to install time-tracking software on their computers to make sure they aren’t cutting corners. She said that some were even caught using AI to rate AI model outputs.
Some, however, remain skeptical of the tech’s ability to replace human workers wholesale. Lawyer Sara Kubik, who has been supplementing her income by working for Mercor, told the newspaper that the work has “taught me the limitations of AI.”
Indeed, researchers have already found that companies may be massively overestimating what AI can do. For instance, a Carnegie Mellon University study found that even the best AI models available at the time failed to complete real-world office tasks 70 percent of the time.
More on Mercor: AI Companies Are Treating Their Workers Like Human Garbage, Which May Be a Sign of Things to Come for the Rest of Us