JAMA Paper Says Autonomous AI Could Surpass Physicians and AI-Assisted Care by 2030
The contested forecast recasts clinical AI from a physician tool into a possible replacement, while real-world patient use and medical training remain unsettled.
Listen to this story
The audio brief
Story brief
3 key pointsRather than treating clinicians as the default safety layer, a JAMA paper forecasts that autonomous systems could outperform physician-AI teams by 2030 across histories, diagnosis, testing, treatment, and chronic-disease management. Its evidence review covers studies published since January 1, 2024, but AMA CEO John Whyte disputes the foundation: many results come from simulations, not blind real-world trials....
- 01
The proposed advantage is not just automation: authors argue clinician involvement can introduce errors in some medical tasks.
- 02
A February 2026 Nature study found most real-world patients struggled to converse effectively with LLMs.
- 03
Curai Health, led by Neal Khosla, uses AI alongside doctors who prescribe and manage complex cases.
The clinical-AI debate is moving beyond whether software can help doctors. A Journal of the American Medical Association paper argues that autonomous AI could deliver better care than both physicians and physician-AI teams by 2030, putting the physician’s role in the care model itself at issue.
The paper, coauthored by Ezekiel Emanuel, Vinod Khosla, Neal Khosla and a researcher associated with Emanuel, answers yes to its own question: whether autonomous AI will exceed physician-AI teams as the best form of medical care.
The case for autonomy
The authors reviewed AI-in-medicine studies published from January 1, 2024 onward. They argue that a superior autonomous system will likely overtake people using AI in five core tasks: taking histories, diagnosing conditions, identifying needed tests, prescribing treatment and managing chronic disease.
The mechanism is not simply better assistance. The paper contends that humans in the loop can degrade AI performance in some medical tasks, making the clinician a potential source of error rather than an essential safeguard.
The real-world challenge
Whyte disputes the paper’s foundation, saying its survey includes simulations rather than blind real-world experiments and that its studies do not uniformly support the conclusion. His position is not that AI lacks promise; it is that the tools should remain inside a physician-governed care plan.
Patient interaction is a concrete constraint on an AI-only model. A February 2026 Nature study cited in the discussion found that most patients in real-life cases could not converse effectively with large language models well enough to access their expertise.
Training is part of the decision
The stakes also reach the training pipeline. The article identifies physician de-skilling as a risk if trainees rely on AI rather than learn history-taking, physical examination and clinical judgment. Whyte says medical schools and residency programs are debating whether trainees should use the tools.
The paper also discloses relevant interests: Vinod Khosla’s investments, Emanuel’s grants and consultancies, and Curai Health’s role. Neal Khosla runs Curai, an online medical company that uses AI alongside doctors who prescribe medication and handle more complicated cases.
Both camps see value in medical AI. The unresolved question is whether the evidence can support autonomous care while patient access to the systems and the effect on clinicians’ training remain contested.
Sources
- wired.comAI Has Human Doctors Asking: What’s Left for Us?
