American Medical Association Fires Back Against Paper Finding AI Is Now Outperforming Doctors

A recent study published in the Journal of the American Medical Association has sparked controversy by claiming that artificial intelligence (AI) is already outperforming human doctors in various cognitive medical functions and will soon care for patients independently better than a physician could. The authors, renowned bioethicist Ezekiel Emanuel and venture capitalist Vinod Khosla, state that "in cognitive medical functions, AI-alone medical care is likely to be better than physician-only or physician-AI hybrid care."

However, the American Medical Association (AMA), led by CEO John Whyte, disagrees with the paper's findings. Whyte argues that some of the studies surveyed are simulations, not blinded experiments, which are the gold standard of medical research. Additionally, he points out that some studies contradict the authors' conclusions, such as a February Nature article finding that most patients struggle communicating with AI chatbots for adequate care.

The AMA still acknowledges the potential of AI tools but insists they must be utilized within a physician-governed care plan. The paper cites the rapid advancement of large language models (LLMs) since their public introduction in November 2022 with ChatGPT, citing their ability to already surpass doctors in five fundamental tasks, including gathering relevant medical information from patients, diagnosing them, conducting diagnostic testing, and managing chronic diseases.

The authors predict this gap between physicians and AI will widen as AI continues to improve, while human doctors may become "deskilled" by relying too heavily on AI models. According to the authors, "human-AI hybrids actually degrade performance compared with AI alone," suggesting that human physicians are better off not hindering AI progress.

However, high-profile AI advocate and UCSF's Robert Wachter cautions that while AI and humans working together are better for now, the future may see scenarios where humans make mistakes. He likens the fear of AI replacing doctors to the doorman fallacy, arguing that AI will not completely replace physicians, who will always be irreplaceable due to their years of training and unique human qualities.