The rise of artificial intelligence is rapidly reshaping numerous industries and a surprising new sector is emerging: AI training. Increasingly, experts in specialized fields are being employed not to *do* their jobs, but to *teach* AI systems how to perform them. This burgeoning field, estimated to be worth at least $17 billion, involves a process called reinforcement learning, where human experts evaluate and refine AI responses, guiding models toward greater accuracy and effectiveness.
Dr. Alice Chiao, formerly an emergency medicine instructor at Stanford University’s medical school, exemplifies this trend. She now dedicates her expertise to training AI-powered chatbots to diagnose and prescribe with a level of clinical reasoning. “AI is going to be the new Doctor Google, the new WebMD that people will go to, to seek out medical information,” Dr. Chiao explained. “I knew that I needed to be a part of that to make sure that the information is accurate, that it’s safe, and that it makes sense to the person using it.”
Companies like Mercor are central to this process, acting as intermediaries between AI developers and a vast network of subject matter experts. Mercor contracts with professionals spanning medicine, law, finance, and even fields like comedy and wine tasting. These experts can earn substantial income – potentially hundreds of dollars per hour – for their contributions. The core principle is that while AI models are trained on massive datasets, they require human guidance to discern between good and bad responses, a process that large AI companies rely on “large armies of people” to accomplish.
The recent volatility in the tech sector, including a dip in software stocks following the release of a new AI tool from Anthropic and a viral essay highlighting potential job disruption, underscores the anxieties surrounding AI’s impact on the workforce. However, Dr. Chiao views her role not as contributing to job displacement, but as enhancing the capabilities of healthcare professionals. She envisions AI assisting doctors with tasks like scan interpretation, chart completion, and note-taking, freeing them to focus on direct patient care.
“Physicians were selected because we really want to help people. We want to heal. We want to spend time talking to people — listening, engaging,” Dr. Chiao said. “I don’t want to see it as AI taking over our jobs. I want to see it as AI taking over the aspects of our jobs that prevent us from being good doctors, good healers and good listeners.”
The training process itself involves presenting AI models with real-world scenarios, prompting them to respond from both a patient’s and a physician’s perspective. Dr. Chiao uses her decades of clinical experience to evaluate these responses, identifying areas where the AI’s reasoning is flawed, potentially misleading, or unsafe. A crucial aspect of this work is crafting responses that are not only accurate but also accessible and reassuring to the end user.
Mercor employs a rubric, developed in consultation with experts in each field, to standardize the evaluation process. This feedback is then used to refine the AI model, encouraging it to generate responses that consistently receive high ratings. The company’s CEO, Brendan Foody, notes that while some tasks are easily taught to AI, those requiring subjective judgment – like humor – present a greater challenge.
Mercor’s foray into training AI in comedy, utilizing comedians from Harvard’s Lampoon, revealed the inherent difficulty in defining what constitutes “funny.” The variability of humor across cultures and individual preferences necessitates a more nuanced approach, requiring experts who can understand and adapt to these diverse perspectives.
The origins of Mercor lie in the recruitment industry. Founded just three years ago by a group of young entrepreneurs, the company initially focused on connecting job seekers with employers. However, recognizing the growing demand for skilled professionals to train AI models, they pivoted their focus, leveraging their existing network of resumes to identify and recruit experts. This strategic shift has proven remarkably successful, with Mercor now processing over $1 million in daily payouts to its network of trainers and achieving a revenue run rate exceeding $500 million. The company is currently valued at over $10 billion, reflecting investor confidence in the long-term viability of this emerging market.
The rapid growth of Mercor and similar companies has attracted significant investment, with Meta’s $14 billion investment in Scale AI last year being a notable example. These investments signal a belief that human feedback and expert testing will remain essential components of AI development for the foreseeable future.
Despite the potential for disruption, Dr. Chiao emphasizes that AI should be viewed as a tool to augment, not replace, human expertise. She cautions against relying solely on AI for medical advice, stressing the importance of a physician’s clinical judgment and the human connection that is integral to effective healthcare. “There is a gut feeling that comes with experience, that comes with sitting with a patient, looking them in the eye, and seeing something that is beyond their history, their lab values, the words that are coming out of their mouth,” she explained. “So, this is where it’s really important to know that the AI is not a doctor, it’s not a human being.”
the success of AI hinges on its ability to address complex global challenges. As Brendan Foody articulated, “We need to cure cancer. We need to solve climate change. And making everyone 10 times more productive so that they’re able to better work on those key problems is going to be a huge, huge benefit to how we make progress as a society.”
