Jensen Huang runs Nvidia, the company whose chips power most AI training. Dario Amodei runs Anthropic. Neither leads a radiology department. Yet both have publicly stated that AI is positioned to replace radiologists - and now physicians are pushing back, calling those statements factually wrong.
The CEO of America's largest public hospital system has made the same claim, adding institutional weight to the argument. That combination - tech executives and hospital administration aligned against the medical profession's own assessment of what its practitioners do - has prompted a direct response from doctors and medical associations calling the statements misleading.
What Radiologists Actually Do
The replacement argument rests on a real finding: AI image analysis tools can match or exceed experienced radiologists on certain narrow tasks. Detecting specific types of tumors in chest CT scans. Flagging potential fractures in X-rays. Screening mammograms for suspicious tissue. On those specific benchmarks, the results are genuine.
What those benchmarks don't test is the rest of the job.
Radiology includes correlating imaging findings with a patient's full clinical picture - their symptoms, lab results, medication history, and the referring physician's specific question. It includes real-time communication with clinical teams, often during procedures. And interventional radiology - where the radiologist performs procedures like biopsies, drain placements, and arterial catheterizations guided by live imaging - involves the radiologist as the physician doing the procedure, not observing it.
AI tools today handle the pattern-recognition layer. They don't handle clinical context, physician communication, or the procedural work. The gap between "performs well on a benchmark" and "can replace a physician" is large, and doctors argue that tech executives are treating it as smaller than it is.
Why the "Replacement" Framing Persists
Radiology is a roughly $15 billion industry in the United States. Globally, there's a significant shortage of radiologists - in low- and middle-income countries especially, the shortage is severe enough that patients wait weeks for diagnostic reads. AI that addresses that shortage has real value.
That legitimate use case - AI as a force multiplier for understaffed systems - gets compressed in public statements into replacement language. It's a more attention-grabbing claim, and it maps better onto narratives about automation-driven efficiency.
Physicians aren't arguing AI has no role in radiology. Many already use AI tools to flag priority cases, check for missed findings, and handle high-volume screening work. The specific objection is to the claim that physicians are no longer necessary - a claim that, doctors say, deters medical students from entering a 13-year training pipeline, shapes patient expectations in misleading ways, and reflects a fundamental misunderstanding of what diagnostic medicine requires.