TL;DR

Despite AI “godfather” Geoffrey Hinton’s 2016 prediction that radiologists were on “borrowed time,” UK NHS radiologist numbers have grown 40% since then. AI tools are being used by radiologists rather than instead of them, with 69% of UK departments now using AI in clinical practice.

Hinton’s Prediction Revisited

In 2016, Hinton told audiences that “people should stop training radiologists now” because “within five years deep learning is going to do better than radiologists.” By 2017, specialist algorithms were outperforming experienced radiologists at detecting pneumonia. Last year, the NHS adopted AI diagnostic tools that helped detect lung cancers earlier.

Yet the predicted job losses haven’t materialised. In the UK, radiologist numbers grew 40%. In Canada, new radiologist residencies reached record highs in 2025. In the US, new recruits increased 20% whilst pay outpaced most other specialties.

Why Two Pairs of Eyes Beat One

“Almost all of the AI tools in use by healthcare providers today are being used by radiologists, not instead of them,” the Financial Times reports. Both human and AI remain fallible, making paired assessment more accurate than either alone. In high-stakes medical settings, the risk of autonomous AI errors is simply too great.

Royal College of Radiologists surveys show 69% of departments now use AI clinically, up from 54% in 2023. Yet only 6% of clinical directors report reduced workloads—37% actually saw workloads increase due to new responsibilities like “post-deployment monitoring” and bottlenecks created when AI speeds up one part of the workflow.

The Irreplaceable Human Element

Professor Amaka Offiah, a paediatric radiologist, explains that AI tools are typically trained for “simple and single tasks.” A tool looking for lung nodules might miss bone metastasis visible on the same scan. Radiologists also interpret images within the context of patient history, discuss cases with multidisciplinary teams, and decide what imaging is needed.

Looking Forward

“AI will assist radiologists, but will not replace them,” Offiah predicts. “I could even dare to say: will never replace them.” The case demonstrates why even when AI excels at high-value tasks, displacement isn’t inevitable—and why sweeping predictions about professions require deep domain understanding.

Source: Financial Times