A decade ago, machine learning scientist and Nobel laureate Geoffrey Hinton made a proclamation that still puts radiologists on edge. He said radiologists were like coyotes that had already walked off a cliff but hadn’t yet looked down. Deep learning was getting so good, so fast, that “people should stop training radiologists now.” In five years — ten, max — AI would do better than radiologists, he predicted.
That prediction has already come true. Yet rather than watching from the sidelines while technology transforms the profession, a rising number of radiology practices are moving forward with artificial intelligence. They are building and buying their own technology, putting it into use within their own operations, and promoting their “AI-native” skills to both their radiologist staff and their hospital clients.
Hinton’s Warning
The father of deep learning spoke with authority on the future of medicine. Hinton, whose work has transformed fields ranging from image recognition to speech processing, made a pointed forecast about radiology. He urged that training for the specialty be halted, arguing that machines would soon surpass human practitioners.
The reasoning was straightforward. Radiology depends on sight. It calls for recognizing patterns, something deep learning handles with skill. As processing speed and precision rose, the issue shifted from whether AI could take over radiologists’ work to when it would happen.
What Radiologists Are Doing Now
STAT reports that radiologists aren’t just waiting for artificial intelligence to arrive. Instead, they are acting on it. Practices are building their own AI tools and putting them into use within their facilities. At the same time, they are promoting these capabilities directly to their customers, presenting themselves as “AI-native” providers.
A notable change has occurred: instead of remaining mere spectators of technological evolution, radiologists are now taking an active role. They are constructing the instruments they will employ, and they are deploying those instruments to secure contracts.
The Line Between Developer and Practitioner
What stands out most about this shift is how it erases the boundary between building technology and putting it to use in medicine. Radiologists are moving beyond simply using AI to actually making it themselves. This represents a deep alteration in the connection between a practitioner and the instruments they employ for their work.
The study zeroes in on outpatient and teleradiology practices, which points to these smaller, more adaptable groups taking the lead. The promotional approach says a great deal. By presenting themselves as “AI-native,” these groups are signaling that they are ahead of the curve. They are betting that hospitals and patients will pay a premium for providers who have their own proprietary technology.
What Comes Next
It is plain where things stand. AI did not arrive at radiology; it has already arrived. The real question concerns not whether radiologists will employ AI, but rather how they will put it to use.
The report’s framing — that AI is blurring the line between technology development and clinical practice — captures the core tension. Rather than being mere spectators to technological progress, radiologists have become its drivers.
The Verdict on Hinton’s Prediction
Time has passed since Hinton made his prediction, yet the practice of radiology has not stayed still. The story instead concerns how radiologists themselves have become their own tech companies. Rather than wait for AI to replace them, they have seized the technology and turned it into something they control.
A new breed of medical professional is emerging: the radiologist who constructs the software that interprets the images. This is a striking shift, and it is taking place at this very moment.
Key Facts Box
1. Geoffrey Hinton, Nobel laureate and AI pioneer, predicted in 2017 that AI would surpass radiologists within five to ten years.
2. STAT reports that outpatient and teleradiology groups are leading the charge on AI adoption in radiology.
3. Practices are developing and acquiring their own AI tools, deploying them in-house, and marketing “AI-native” capabilities.
4. The prediction has run out, but radiologists have not been replaced en masse; they have adapted by becoming their own tech companies.
Source material: “STAT+: In radiology, AI is blurring the line between technology development and clinical practice,” STAT.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.