As artificial intelligence in healthcare advances, many physicians assume that they will continue to serve as the trusted decision-makers guiding AI-enabled care.
A provocative piece published in the medical journal JAMA suggests a more disruptive possibility: that autonomous AI systems could eventually outperform both physicians and physician-AI teams on key cognitive tasks.
Ezekiel J. Emanuel, M.D., Ph.D., an oncologist, bioethicist and a former White House policy advisor, along with Neal Khosla, the co-founder and CEO of Curai Health, Vinod Khosla, founder of Khosla Ventures, and researcher Abe Baker-Butler, authored the Perspective article published August 17 that argues that AI-alone medical care is likely to be better than physician-only or physician-AI hybrid care.
Large language models (LLMs) were only publicly introduced in November 2022, and already generative AI is matching or outperforming physicians on key cognitive medical tasks, including gathering patient information, diagnosing conditions, selecting tests, recommending treatments and managing chronic diseases, the authors argue.
"[A] review of all published articles on AI in medicine since January 1, 2024, shows that medicine is rapidly approaching the transition point at which AI alone will exceed physicians and physician-AI hybrids in providing the best care at 5 fundamental cognitive medical tasks," Emanuel and Vinod and Neal Khosla wrote.
They contend that as AI continues to improve, autonomous AI systems may eventually provide better care than either physicians alone or physician-AI teams because human oversight can introduce errors and reduce the performance of superior AI systems.
"Physicians will not disappear. Surgery, deliveries, and other physical procedures still need human hands. But 'doctor plus AI' as the automatic gold standard of care no longer holds up against the evidence," Emanuel wrote in a LinkedIn post. "AI will likely be ready to be deployed for real-world cognitive medical tasks in some, maybe many, workflows by 2030. Now is the time for leadership to develop workflows and regulation."
But the opinion piece authors acknowledge more assessments of AI in real-life clinical encounters are essential. Liability, regulatory, and reimbursement barriers limit studies examining AI alone in real clinical interactions, they noted.
Even if AI proves better than physicians at certain cognitive medical tasks, that alone won't guarantee widespread adoption. Many areas of medicine still require physical exams, procedures and hands-on interventions that AI and robotics cannot yet perform independently, meaning physicians would still be needed in many workflows. They also note that technical, operational and financial hurdles could slow deployment of autonomous AI in clinical settings.
"Nonetheless, superior autonomous AI will likely be ready to be deployed for real-world cognitive medical tasks in some, maybe many, workflows by 2030. Consequently, physicians, policymakers, and others need to urgently devise approaches to workflow, liability, regulation, reimbursement, and medical education," Emanuel and Khosla wrote.
The opinion piece prompted intense debate and discussion both on- and offline.
In a LinkedIn post, John Whyte, the CEO of the American Medical Association, wrote, "I respect and appreciate Zeke’s views on a variety of issues, but on this one, we are going to have to disagree. Zeke may be right about the tasks. But medicine is not a collection of tasks. Getting the diagnosis right matters. Choosing the right test matters. Recommending the right treatment matters."
Medicine still requires human judgment, empathy and an understanding of patients' individual circumstances, Whyte argued.
"It’s why, even with autopilot, most of us would not board a plane without a pilot in the cockpit. Clinical performance is only part of the equation. Performing medical tasks well is not the same as practicing medicine well. Even the JAMA piece acknowledges that much of the evidence comes from simulated tasks, not real-world clinical encounters with real-world patients. That matters when we’re talking about taking the physician out of the equation," Whyte wrote.
Many medical professionals likened the debate to the use of autopilot in an airplane cockpit—it didn't eliminate the need for pilots but changed their work.
This week, the AMA and the Digital Medical Society released a framework about the role of physicians in the age of AI. The AMA said the framework was the result of a collaboration of leaders in medicine, health systems, government, technology and the patient community. The AMA asserts that physicians will continue to be central to patient care and play a critical role in shaping how AI is used in healthcare.
"Individual tasks will evolve as technology advances, but the physician's enduring responsibilities will remain remarkably stable," the AMA and DiME wrote in the framework's intro.
Those responsibilities, including ensuring empathy, communication and shared decision-making, remain at the heart of medicine, applying clinical expertise, judgment and accountability and shaping technology-enabled models of care that improve quality, continuity and access while preserving the physician’s highest-value contributions.
Physicians also are charged with stewarding the responsible use of technology to guide the safe, effective and equitable integration of digital health and AI into clinical practice, the AMA asserts, while also preserving and strengthening clinical expertise and ethical decision-making.
AMA and DiME argue that AI and digital health can improve healthcare access, quality and sustainability. The framework (PDF) presents AI adoption in healthcare as a shared, phased effort in which patients, clinicians, technology companies, policymakers and payers each play distinct roles to create a future of accessible, high-quality care supported by trustworthy AI.
Under this framework, patients will define desired outcomes and experiences, and physicians and clinical professionals will lead the evolution of care delivery and clinical practice. Technology developers and investors will build the tools and capabilities while policymakers and payers create the incentives, regulations and reimbursement structures needed to scale innovation.
The roadmap also outlines a progression from short-term outcomes to align priorities and incentives, while medium-term outcomes focus on implementation, where new models of care become visible in practice. Long-term outcomes describe a healthcare system in which these models are routinely and sustainably delivered, according to the framework.
The AMA framework and the JAMA opinion piece are aligned on some points, including that the healthcare industry will need to rethink workflows, medical education, reimbursement, regulation and liability as AI capabilities advance. But they diverge sharply on the future role of physicians.
The AMA-DiME framework establishes an initial direction for how the practice of medicine must evolve in the era of AI, the AMA wrote.
"The next phase will require continued collaboration across the healthcare ecosystem to translate these principles into practice: defining evolving roles and responsibilities, developing the education and competencies physicians need, designing and evaluating new models of care, and addressing the policy, payment, technology and infrastructure required to support them," the AMA wrote.