Don’t be fooled by the question mark in the title of a paper published this month in the Journal of the American Medical Association. The paper asks, “Will autonomous AI surpass AI doctors and become the best medical care?” But the question is largely rhetorical. The authors—including medical expert Ezekiel Emanuel and venture capitalist Vinod Khosla—believe the answer is a resounding yes.
Many healthcare professionals are reluctantly accepting the idea that patients may receive the best care through a hybrid model in which doctors work alongside highly trained artificial intelligence systems. However, the paper goes further, arguing that autonomous AI alone could eventually deliver better results than doctors or doctor-AI teams.
“A review of all published papers on AI in healthcare since January 1, 2024, shows that healthcare is rapidly approaching a transition point where AI alone will exceed physicians and physician-AI hybrids in providing the best care in five fundamental medical tasks,” the authors wrote. These tasks include collecting a patient’s medical history, making a diagnosis, determining which tests are necessary, prescribing treatment, and managing chronic conditions—in other words, many of the core responsibilities traditionally associated with being a doctor.
The researchers argue that medical AI has advanced so quickly that clinicians may eventually need to limit their involvement because “human involvement degrades AI performance.” In this scenario, doctors would step back and allow autonomous AI systems to handle much of the work that once required human medical expertise.
Lead author Ezekiel Emanuel, a prominent oncologist and director of the Department of Medical Ethics and Health Policy at the University of Pennsylvania, told me that he once rejected the idea that AI could perform doctors’ core responsibilities. In the 2010s, he met Khosla at a conference where the venture capitalist predicted that AI would perform 85% of a doctor’s work by 2035. Khosla had promoted that idea for years and even published research on it.
In a 2012 TechCrunch article titled “Do We Need Doctors or Algorithms?”—another rhetorical question—Khosla explored whether algorithms could replace much of the medical profession. He later published a 101-page paper in 2016 that made a similar case. Emanuel now says he was wrong to dismiss the prediction. “I said some bullshit. That’s not true,” he says. “What doctors do is too complicated.”
About a year ago, however, Emanuel’s friend Robert Wachter sent him an advance copy of his book, The Big Leap. Wachter, dean of medicine at the University of California, San Francisco, described a future in which the highest-quality medical care would combine AI and physicians, while many patients could receive more affordable “economy class” care largely through AI.
That vision prompted Emanuel to reconsider Khosla’s argument. It also raised a troubling question: “When AI takes over, what will be left for doctors?”
Before reaching a conclusion, Emanuel wanted to test his assumptions. He asked Khosla to collaborate, and the venture capitalist suggested bringing in his son, Neil Khosla, along with Emanuel’s research team. Neil Khosla runs Curai Health, an online healthcare company that uses AI to support patient care while physicians prescribe medications and handle more complex cases.
The paper discloses this potential conflict of interest, as well as Vinod Khosla’s related investments and Emanuel’s grants and consulting relationships. After reviewing research on AI in medicine published from January 2024 through the present, the authors concluded that “superior autonomous AI” could surpass doctors and physician-AI teams in many medical tasks by 2030. Emanuel acknowledges that the prediction is “unsettling, but it seems likely.”
“I feel anxious” may be an understatement. One of the most vocal critics of the paper is John White, CEO of the American Medical Association, which represents physicians concerned about the future role of doctors in healthcare.
When I spoke with White, he pointed to several limitations in the research, including the fact that some studies relied on simulations rather than blinded, real-world trials. He also noted that not every study supported the paper’s conclusions. In particular, a February 2026 article in Nature Medicine found that many patients struggle to communicate effectively with large language models when seeking specialized medical knowledge.
White was understandably critical of the idea that healthcare could remove humans from the clinical decision-making process. “The AMA recognizes the potential of these tools, but they must be leveraged within a physician-managed care plan,” he says.
Source: www.wired.com