JAMA Study Predicts Autonomous AI Will Outperform Doctors

A newly published paper in the Journal of the American Medical Association argues that autonomous artificial intelligence will surpass both human physicians and human-AI hybrids by 2030. Coauthored by oncologist Ezekiel Emanuel, venture capitalist Vinod Khosla, and Curai Health founder Neal Khosla, the paper analyzed medical AI literature published since January 1, 2024. The authors conclude that AI alone will soon deliver superior outcomes in five core clinical areas: taking medical histories, diagnosing ailments, ordering tests, prescribing treatments, and managing chronic diseases.

The paper represents a major shift for Emanuel, who previously dismissed Vinod Khosla's long-standing prediction that AI would perform 85 percent of clinical duties by 2035. The authors now argue that keeping humans in the loop actually degrades AI performance. This challenges the prevailing industry consensus, championed by UCSF head of medicine Robert Wachter in his book A Giant Leap, which envisioned AI-only care as a lower-tier economy class option while wealthy patients enjoyed collaborative human-AI care. Wachter now concedes that human intervention can sometimes disrupt optimal AI performance.

The medical establishment has pushed back strongly against this vision of fully autonomous care. John Whyte, CEO of the American Medical Association, criticized the paper's reliance on simulation studies rather than blind experiments. Whyte also pointed to a February 2026 Nature article showing that real-world patients often struggle to communicate effectively with large language models. For practitioners, this shift raises serious concerns about de-skilling, as future doctors who rely on pocket-sized AI assistants may never master fundamental diagnostic skills. However, proponents like Vinod Khosla argue that while humans will still be needed for surgery and emergency care, their primary value in general practice will shift toward empathy, communication, and refining AI systems.