Medical Elite Argues AI Alone Will Soon Surpass Doctor-AI Collaboration for Patient Care

A recent paper in the Journal of the American Medical Association argues that standalone AI systems will outperform both physicians and human-AI teams across core medical functions—a claim that has ignited fierce disagreement among medical leaders about the profession's future.

Authored by renowned cancer specialist Ezekiel Emanuel and technology investor Vinod Khosla, the piece contends that reviewing published AI research since early 2024 reveals a clear trajectory: autonomous AI is racing toward superiority in five fundamental clinical tasks—gathering patient histories, making diagnoses, ordering appropriate tests, choosing treatments, and managing ongoing illnesses.

From Skepticism to Conversion#

Emanuel spent years dismissing Khosla's predictions that algorithms would handle 85 percent of physician work by 2035. After encountering Robert Wachter's book A Giant Leap—which imagined a system where wealthy patients receive AI-human collaboration while others rely mainly on AI alone—Emanuel's thinking shifted. The turning point came when he wondered aloud: if AI truly takes over medicine, what purpose remains for doctors?

To investigate this troubling possibility, Emanuel recruited Khosla to co-author a systematic review. Their team, which included Khosla's son Neal (who runs Curai Health, an AI-enabled primary care platform), examined available research and reached an unsettling conclusion: "superior autonomous AI" will likely eclipse humans by 2030. The paper notes this prediction is "unsettling but seems probable."

The Pushback#

The American Medical Association's CEO, John Whyte, mounted several objections. He flagged methodological concerns: many studies reviewed were simulations rather than blinded trials, and not all support the authors' thesis. A February 2026 Nature study, for instance, showed that patients struggled to effectively communicate with large language models in real-world scenarios. Whyte rejected the premise of removing human oversight, arguing instead that tools must remain embedded within physician-directed care plans.

Wachter, whom the paper cites, surprised observers by conceding the authors raise an important point. He acknowledged that AI alone now rivals or exceeds human-AI partnerships in certain contexts. Yet he cautioned that "humans in the loop" will sometimes "muck up performance"—a paradox the field cannot dismiss with certainty. Beyond diagnostic accuracy, Wachter emphasized irreplaceable human strengths: delivering difficult prognoses with empathy and guiding patients toward optimal paths through personal rapport and expertise that machines cannot replicate.