AI is already outperforming doctors on diagnostics and prescribing tasks, according to a new JAMA opinion piece from bioethicist Dr. Zeke Emanuel and health investors Vinod and Neal Khosla. The authors argue human intervention can make care worse and that fully autonomous healthcare could be deployed by 2030 - a claim they describe as "unsettling but seems profitable."
The position has drawn sharp responses across medicine and tech. Venture firm a16z questioned whether current benchmarks prove AI works in real-world clinical settings. The American Medical Association backed physician oversight. Bill Gates, who has previously voiced optimism about AI, walked back some of that stance and called for "Human Reserved" domains such as eldercare.
Who's pushing autonomous care
The authors align with positions held by Elon Musk and Demis Hassabis on AI's medical potential. Vinod Khosla backs OpenAI, and Neal Khosla runs Curai Health, a company building AI-powered primary care tools. Those financial ties have led critics to point out the authors are talking their book.
Still, the underlying trend is real. AI is already reshaping healthcare workflows, and a majority of doctors support its use in clinical settings. The debate is less about whether AI belongs in medicine and more about how much autonomy it should have.
Patients want control
Public sentiment lags behind professional enthusiasm. Two-thirds of Americans want more say in how AI is used in their care. That gap between physician readiness and patient comfort could shape how quickly autonomous systems move from pilot programs to standard practice.
For professionals working in healthcare, the skills conversation is shifting. Understanding how these systems make decisions - and where human judgment remains necessary - is becoming part of the job. Training in AI for Healthcare and AI Agents & Automation is increasingly relevant as clinical teams evaluate where automation fits and where it doesn't.
The authorization question
Whether framed as the end of doctors or proof of their necessity, the future of AI healthcare still requires human authorization. The systems may recommend, diagnose, or prescribe, but someone has to sign off - at least for now.
Why this matters for healthcare professionals
You will likely be asked to work alongside AI systems that make clinical recommendations before the decade is out. The debate over full autonomy is not settled, but the direction is clear: AI will handle more diagnostic and prescribing tasks. Knowing where these tools are reliable - and where they fail - will determine whether you use them effectively or get sidelined by them.
Your membership also unlocks:
