An opinion piece published in the medical journal JAMA argues that autonomous AI systems will soon outperform human doctors in medical reasoning, prompting calls for regulators to avoid mandating human oversight. The authors suggest that AI's growing capabilities in tasks like diagnosis and treatment planning could render human involvement unnecessary in certain workflows. The piece highlights that AI systems have already demonstrated higher accuracy in simulated patient interactions compared to human doctors, raising questions about the future role of physicians in healthcare decision-making.
The JAMA authors cite multiple studies showing AI models outperforming doctors in core medical reasoning tasks, including diagnosis and treatment planning. For instance, Google's AMIE system scored higher than primary care doctors in simulated patient conversations, while ChatGPT o3 achieved a 60% accuracy rate in diagnosing complex cases, compared to 15.9% for a group of internists. Microsoft's diagnostic orchestrator also outperformed doctors by finding correct diagnoses under budget constraints at a lower cost. The authors argue that many studies showing the opposite result are outdated or methodologically flawed, as they often exclude the most advanced AI models.
The authors of the JAMA piece include Ezekiel Emanuel, a bioethicist at the University of Pennsylvania, and Neal Khosla, CEO of Curai Health. They acknowledge limitations in their argument, noting that most evidence comes from simulations rather than real-world patient care. They also admit that physical procedures like surgery and childbirth will remain in human hands for the foreseeable future due to current technological constraints. The piece concludes by urging policymakers to rethink medical training, liability, and regulation as AI's role in healthcare continues to evolve.
Source: thedecoder
