AI won’t replace radiologists, but it will dramatically change their jobs
Ars Technica Lola Butcher
Radiology didn’t get replaced, but AI is now in the exam room. That’s changing who does the work — and how well some scans get read.
Based on reporting by Ars Technica, Lola Butcher — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Geoffrey Hinton once said radiologists would be replaced by computers within five years. That didn’t happen. Radiology jobs are still growing, and the field is expected to add 26 percent or more over the next three decades.
But the bigger point in Hinton’s old warning holds up: doctors who read scans now have a silicon coworker. In radiology, AI is already doing real work, not the usual demo-room theater. It has become medicine’s clearest test case for how expert judgment systems get folded into daily practice.
By early 2026, about three-quarters of the 1,400 AI-enabled medical devices cleared by the Food and Drug Administration were aimed at radiology. Some of those tools help by drafting reports or flagging the images that need attention first. That kind of software doesn’t replace the physician. It changes the pace of the job, and probably the shape of it too.
And some systems go further. They can spot abnormalities that the human eye may miss, or interpret images as well as trained radiologists, sometimes better. A review of 43 clinical trials found that AI-assisted colonoscopies found more polyps than standard ones. That’s the real story here: not a robot takeover, but a steady transfer of certain calls from the person in the chair to the machine on the side.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI hype that deserves less shouting and more paperwork. Radiology shows the cleanest version of the deal: the machine doesn’t need to win a victory lap, it just needs to catch what humans miss and speed up the grind. The broader pattern is obvious enough — closed, clinical, and heavily checked systems are where AI earns its keep first, not in the grand “replace everyone” fantasy.
Read more about this at: Ars Technica
