Understanding the Potential of AI in Healthcare
Do not be misled by the question mark in the title of a recent article published in the Journal of the American Medical Association. When the authors, including prominent medical figure Ezekiel Emanuel and venture capitalist Vinod Khosla, inquired, “Will autonomous AI surpass doctors as the best option for healthcare?”, it was a rhetorical question. Their answer is a resounding YES.
Many within the medical community are reluctantly accepting the notion that patients may receive optimal treatment through a hybrid model where physicians collaborate with well-trained bots. However, the thesis of this article posits that AI on its own could yield superior outcomes.
“A review of studies on AI in medicine published since January 2024 suggests that the technology is nearing a critical point: in five fundamental medical tasks, AI alone could outperform both human doctors and systems that combine human and AI efforts,” the authors assert.
These tasks encompass gathering medical histories, establishing diagnoses, identifying necessary tests, prescribing treatments, and managing chronic diseases. AI has advanced to the point where human intervention may hinder its performance.
Shifting Perspectives on Medical AI
The lead author, Emanuel, a professor in the Department of Medical Ethics and Health Policy at the University of Pennsylvania and a renowned oncologist, shared that for years he dismissed the idea of AI performing basic medical functions. During the 2010s, he encountered Khosla at various conferences, where the venture capitalist insisted that AI would handle 85% of medical tasks by 2035.
Khosla has long been passionate about this vision, having penned an article for TechCrunch in 2012 questioning, “Do we need doctors or algorithms?” and drafting a comprehensive 101-page treatise on the subject in 2016. Emanuel initially rebuffed these claims, arguing that the complexities of medical practice were insurmountable for AI.
However, approximately a year ago, Robert Wachter, a friend of Emanuel’s, sent him his book, “A Giant Leap,” which outlined a future where top-tier medicine would be a collaborative process between AI and physicians, while the masses would have to rely solely on AI. This prompted Emanuel to reconsider Khosla’s assertions and led him to a troubling question: “If AI takes over, what is left for doctors to do?”
Research Collaboration and Findings
To validate this premise, Emanuel enlisted Khosla’s collaboration. Khosla suggested involving his son, Neal Khosla, who leads Curai Health, an online company leveraging AI to treat patients, with a medical team managing prescriptions and complex cases. The co-authors embarked on a thorough investigation of all AI-related medical studies published since January 2024 and concluded that, in many instances, a “superior autonomous AI” is likely to eclipse human practitioners using AI by 2030. They acknowledge that this prediction is “disturbing, yet plausible.”
Merely describing it as disturbing fails to capture the gravity of the situation. One of the most vocal dissenters is John Whyte, CEO of the American Medical Association (AMA), representing those physicians concerned about the implications of such advancements. In our conversation, Whyte highlighted the “flaws in the article,” noting that some studies examined are simulations rather than blind experiments and that not all support the article’s conclusions.
Notably, a February 2026 study published in Nature revealed that, in real-world scenarios, most patients struggled to communicate effectively with large language models to access their specialised knowledge. Unsurprisingly, Whyte expressed outrage at the notion of excluding humans from the clinical process: “The AMA acknowledges the potential of these tools, but they must be used within the context of a physician-supervised care plan.”
The Rapid Advancement of AI and its Implications
Emanuel countered that he and his co-authors had exercised caution in their conclusions: “It has been less than four years since ChatGPT was introduced. We are making a prediction for the next four years. Don’t you think AI will be far more advanced than a doctor?”
As AI continues to improve, it raises concerns among primary care physicians. Seeking a second opinion, I consulted Wachter. He noted in the article’s opening that he believes AI-only treatment could emerge as a more basic and inferior alternative. Surprisingly, Wachter conceded that the authors of the article are correct: “The argument they present is significant: combining AI with humans is beneficial. However, there will be instances where we will compromise the outcome.”
Nonetheless, Wachter insists that the unique human attributes of trained physicians are vital and irreplaceable. AI may never match a human’s ability to deliver a grim prognosis, and patients are likely to receive better guidance towards optimal treatment from a flesh-and-blood doctor.
The Unique Role of Human Physicians in a Changing Landscape
Wachter invokes what he terms the “doorkeeper fallacy,” named after concerns that those who opened doors for tenants in residential buildings would lose their jobs once doors could open automatically. In reality, doorkeepers perform numerous tasks and remain invaluable in luxury settings—receiving Amazon packages, caring for pets, and sometimes providing a sympathetic ear for tenants voicing complaints. “We will encounter a version of the doorkeeper fallacy applied to physicians. I can envision a world where a doctor accepts the AI diagnosis but also plays many other valuable roles.”
However, what is being replaced here is not as straightforward as opening a door. It involves the expertise gained through years of training. This process of “de-skilling” could accelerate AI’s takeover of medicine, particularly as new generations of doctors may not see the necessity of spending years learning skills that can be accessed through an AI expert. Whyte admits this is concerning: “Many medical schools and residency programmes are debating whether training doctors can utilise these tools, because if you have never learned to gather a medical history or perform a physical examination, how will you learn it?” On the flip side, some argue that neglecting such a useful tool is a form of professional malpractice.
The Future of Physicians in an AI-Dominated World
Vinod Khosla posits that we will still need doctors for surgical procedures and emergencies. Neal Khosla indicates that change is already underway and anticipates that in the coming years, we will grant regulatory approval for AI to prescribe medications.
Recall the television series “House,” where a diagnostic genius deciphered ailments while insulting patients and belittling his assistants. In a world dominated by AI, a Dr. House would find himself out of work, as AI would routinely and freely apply its vast knowledge base to obscure ailments. The value of human physicians will lie in those areas where House often faltered, such as empathy and communication grounded in high emotional intelligence. Of course, those performing practical procedures remain safe from AI encroachment… at least until robots catch on.
Healthcare is a vital aspect of our lives, but the debate extends beyond this sector. The question, “What is left for doctors to do?” could easily translate to, “What is left for humans to do?” Bill Gates suggests that society might reserve certain tasks for humans to prevent economic ruin and a wave of discontent. Convincing corporate shareholders to reassess funding for artificial jobs may prove challenging. At the very least, the plight of the masses facing unemployment may be alleviated by one advantage: first-rate medical advice just a click away.