The recent publication in the Journal of the American Medical Association poses a question that seems to invite debate, but the authors, a heavyweight team including medical ethicist Ezekiel Emanuel and venture capitalist Vinod Khosla, are not really asking—they are declaring. The question, “Will Autonomous AI Exceed AI-Physicians as the Best Medical Care?” is purely rhetorical, and their answer is a resounding and unequivocal yes. This stance is forcing the medical world to confront a profound and uncomfortable shift. While many are grudgingly accepting a future where the best care comes from a hybrid team of human doctors working in tandem with sophisticated algorithms, this new thesis goes much further. It contends that the pinnacle of medical care will be achieved by AI operating entirely on its own, without the “interference” of human clinicians, suggesting that the future of medicine is not about collaboration, but about a complete handover of responsibility to the machines.
The article’s argument is rooted in a sweeping review of published research on AI in medicine since the start of 2024, which the authors claim demonstrates a rapid approach to a critical tipping point. They assert that autonomous AI is on the verge of surpassing both human physicians and physician-AI hybrids in five fundamental tasks that encompass the very essence of doctoring: taking a patient’s medical history, establishing a diagnosis, determining the necessary tests, prescribing treatment, and managing chronic diseases. The bold, even radical, conclusion they draw is that clinicians should actively refrain from meddling in these processes, because according to their reading of the data, human involvement can actually degrade AI’s performance. This is a direct challenge to the very definition of the physician’s role, essentially suggesting that the best thing a doctor can do is to stand back and let the AI do what was once their job, a statement that sends shockwaves through the profession.
The personal journey of the lead author, Ezekiel Emanuel, a renowned oncologist and chair of Medical Ethics and Health Policy at the University of Pennsylvania, adds a layer of human drama to this technological forecast. For years, Emanuel was a prominent skeptic, dismissing Khosla’s predictions as the fanciful talk of a tech investor who didn’t understand the complexities of the medical craft. He vividly recalls encounters with Khosla in the 2010s, where the VC would enthusiastically champion the idea that AI would soon handle the vast majority of a doctor’s work. Emanuel’s internal response was one of blunt dismissal: “I said bullshit, there’s no way. What doctors do is too complicated.” His conversion began about a year ago when his friend Robert Wachter, the head of medicine at UCSF, shared the galleys of his book, A Giant Leap, which outlined a future of collaborative care but also suggested that a tiered system would leave the masses reliant on AI alone. This sparked a terrifying question in Emanuel’s mind: if AI truly takes over, what is left for doctors to do?
To answer this, Emanuel decided to test the very premise that had so challenged his worldview. He enlisted Khosla as a collaborator, who brought in his son Neal, the CEO of Curai Health, an online company that uses AI to treat patients with physician oversight. Together, they launched an extensive review of the most recent research, and the evidence that they gathered was compelling enough to convince the once-skeptical Emanuel that AI is already approaching, and will soon exceed, human performance in core medical tasks. The speed and confidence of his conversion are remarkable, and the paper he co-authored openly acknowledges the unsettling nature of their prediction, stating that by 2030, superior autonomous AI will likely surpass humans using AI, a conclusion they admit is “unsettling but seems probable.” The paper’s authors are not just speculating; they are placing a stake in the ground that the future of medicine will be dominated by machines, and that the era of the human doctor as the primary diagnostician and treatment planner is drawing to a close.
This bold assertion has, unsurprisingly, drawn sharp criticism, most notably from John Whyte, the CEO of the American Medical Association, the powerful professional organization representing physicians across the United States. Whyte is quick to point out what he sees as the paper’s methodological flaws, arguing that many of the studies it relies on are simulations rather than real-world blind experiments, and that they don’t all support the article’s sweeping conclusions. He cites a recent Nature article that found patients were often unable to effectively converse with large language models to access their expertise in real-life scenarios. More fundamentally, Whyte bristles at the core idea of removing humans from the medical loop entirely. For the AMA, AI is a powerful tool, but it must be used within the context of a care plan governed by a physician, who can provide the judgment, empathy, and accountability that a machine simply cannot replicate. The debate is now a fundamental clash of worldviews: one that values the promise of pure algorithmic efficiency versus one that insists on the necessity of human connection and oversight.
The controversy goes to the very heart of what we value in medicine and what we fear about the future. Emanuel’s personal reflection on his own impending obsolescence—a fear of being an “old male doctor” with nothing to do—highlights the existential crisis that this technology poses for millions of healthcare professionals. On a patient level, the prospect of interacting with an autonomous AI is filled with both promise and profound unease. While AI could potentially diagnose a rare disease from a scan with perfect accuracy or identify subtle patterns in symptoms that a busy doctor might miss, it also raises unsettling questions about the therapeutic value of human touch. Can a machine offer comfort, read the emotional state of a patient in a way that builds trust, or navigate the complex ethical and social nuances of a difficult diagnosis? The authors of the JAMA paper would argue that for the best medical outcome, at least for the core diagnostic and treatment tasks, the answer is to trust the data. For the skeptics, this cedes a fundamental part of our humanity in the name of efficiency, transforming the deeply personal experience of being cared for into a cold transaction with a machine. The argument is no longer about whether AI will be a part of medicine, but about the very soul of the profession and the essence of care itself.
