Godfather of AI wrongly predicted key job would be | Tech News

In 2016, AI pioneer Geoffrey Hinton—usually referred to as the “godfather of AI”—made a blunt prediction: medical faculties ought to “stop training radiologists right now” as a result of the specialty would be largely changed within 5 years.

That call turned out to be a unhealthy guess.

Instead of disappearing, radiology has grown, with demand and pay rising and the quantity of practising U.S. radiologists growing by roughly 10%, in response to University of Virginia Darden economist Christoph Herpfer, who research doctor labor markets.

Hinton has since walked the declare back. He later mentioned he’d been too broad and, in a 2025 interview with The New York Times, clarified that he was speaking particularly about image evaluation—not the total scope of a radiologist’s job.

He additionally conceded the timeline was incorrect, even when he nonetheless believes AI will keep enhancing. Hinton’s up to date view is that image interpretation will more and more be carried out by radiologists working alongside AI systems, boosting effectivity and accuracy somewhat than eradicating the doctor from the method.

One purpose his authentic prediction missed the mark is that radiologists do a lot more than “read scans.” They seek the advice of with surgeons and different medical doctors, talk with sufferers, review medical histories, and produce detailed experiences that apply imaging findings to a particular individual’s context—work that relies upon closely on judgment and expertise.

AI, in the meantime, is already altering how radiology is practiced, largely as an assistive software. Radiology has long used software program to reinforce photographs and spotlight areas of concern, however newer AI systems can prioritize pressing instances, counsel doable findings, and even draft components of experiences after being educated on huge datasets of medical photographs.

Not everybody agrees on how shortly the sector ought to undertake these instruments. NIH radiologist and AI researcher Dr. Ronald Summers has argued that some AI strategies are sturdy enough to make use of more broadly proper now.

At the identical time, many radiologists stay cautious, pointing to restricted real-world validation, “black box” decision-making, and uncertainty about whether or not coaching knowledge displays the affected person populations seen in on a regular basis follow—issues echoed by Stanford radiologist Dr. Curtis Langlotz.

Regulators are additionally signaling that AI is just not a substitute for medical doctors. The FDA has cleared more than 700 AI instruments to assist clinicians, with the bulk geared toward radiology, however the accepted systems nonetheless require a human within the loop. In different phrases, the long run Hinton as soon as framed as “radiologists gone soon” is wanting a lot more like “radiologists plus AI”—and his authentic career-ending prediction didn’t come close to matching what occurred.