# Will AI replace radiologists? What happened to the famous prediction
URL: https://offladder.com/answers/ai-and-radiologists
Last updated: 2026-11-18
By: Nivaan, Founder, OffLadder

No, and radiology is the best example of why confident predictions about AI replacing jobs can be wrong. In 2016 Geoffrey Hinton said we should stop training radiologists because AI would outperform them within years. Since then hundreds of AI imaging tools have been cleared by the US Food and Drug Administration, more than in any other medical specialty, yet radiologist shortages have grown in the US and UK. AI is good at narrow tasks such as flagging a possible bleed, measuring a nodule or prioritising scans. Radiologists do much more: combine images with the clinical picture, deal with unusual cases, talk to other doctors, perform procedures, and take responsibility. AI has added work as well as saved it.

## Key takeaways

- The 2016 prediction that radiologists would be replaced has not come true.
- Radiology has more FDA-cleared AI tools than any other specialty.
- AI does narrow tasks; radiologists integrate, decide and take responsibility.
- Shortages of radiologists have grown, not shrunk.

## The prediction and what happened

In 2016, Geoffrey Hinton, one of the pioneers of deep learning, said it was obvious AI would outperform radiologists within five years. Image recognition was advancing fast, and radiology looked like a job of looking at pictures. A decade later, radiologists are in short supply in many countries, imaging volumes keep rising, and AI tools are used alongside radiologists rather than instead of them.

## What AI actually does in radiology

| Task | What AI does | What the radiologist does |
| --- | --- | --- |
| Triage | Flags urgent findings like bleeds | Confirms and acts |
| Measurement | Measures nodules and volumes | Interprets change over time |
| Detection | Highlights possible lesions | Rules in or out with the full picture |
| Reporting | Drafts structured reports | Writes the conclusion and signs |
| Procedures | Little | Biopsies, drains and interventions |

*Imaging tasks and AI*

## Why the job wasn't replaced

- Radiologists do far more than detect one thing on one image.
- AI tools are narrow and often perform worse in new hospitals.
- Every AI flag still needs checking, which adds work.
- Regulation and liability keep a doctor responsible.
- Demand for imaging grew faster than any efficiency gain.

## What the story teaches about other jobs

When you read that AI will replace a job, ask which tasks it can do, how much of the job those tasks are, and what happens to demand if the work gets cheaper. Radiology shows that being good at one task is not the same as doing the job. It also shows that the job changes: radiologists now spend more time on hard cases and consultation.

## If you're considering medicine

Specialties that combine technology with patient contact and judgement, including radiology, look resilient. Medical students increasingly learn how AI tools are validated and where they fail. Clinical AI safety and informatics roles are also open to doctors who want to shape how tools are used.

## Illustrative example

Dr Okafor's hospital introduces an AI tool that flags possible lung nodules. In the first month it flags hundreds of tiny findings that need no action, adding to her workload. She works with the vendor to change the thresholds and now sits on the hospital's AI review group.

**One small thing to try: Break one job into tasks**

Pick a job you're worried about. List ten tasks it involves. Mark which ones AI could do today. Then ask: if those got cheaper, would demand for the whole job rise or fall? That is the radiology test, and it works for any role.

## Test a future job before you train for it

A forecast can tell you which jobs are growing. It can't tell you whether you'd enjoy the work. Laddie, our AI coach, starts from the experience you already have and works through four steps with you:

1. Explore: directions your experience carries into, including work that's only now taking shape as AI changes jobs.
2. Try: an experiment small enough to finish this week, a small piece of the work itself.
3. Learn: note which part you liked and which part you avoided, and keep it as a record of what you tried.
4. Apply: when a direction turns into a job post, OffLadder CV shapes your CV for it. What you can't show yet goes back to Laddie as a short lesson, a 15-minute task and a mock interview.

**Read next**

- [Will AI replace doctors?](https://offladder.com/answers/ai-and-doctors)
- [Is medicine right for me?](https://offladder.com/answers/is-medicine-right-for-me)
- [How exposed is my job to AI?](https://offladder.com/answers/how-exposed-is-my-job-to-ai)
- [Clinical AI safety reviewer](https://offladder.com/directions/clinical-ai-safety-reviewer)


## Questions people also ask

### What did Geoffrey Hinton say about radiologists?

In 2016 he said people should stop training radiologists because deep learning would do better within about five years. He later said he had been wrong about the timing and about how the job would change.

### How many AI tools are used in radiology?

Radiology accounts for the majority of AI-enabled medical devices on the FDA's list. Many are narrow tools for one condition or one type of scan.

### Is radiology a good specialty to choose now?

Demand is strong in both the US and UK, and many radiologists now work with AI daily. Expect the job to keep changing, with more focus on complex cases, procedures and consultation.

### Does AI make radiologists faster?

Sometimes. Tools that prioritise urgent scans or measure findings save time. Others add steps, because every flag must be checked. Results depend on how well the tool fits the workflow.

### Will AI replace radiographers?

Radiographers and technologists operate the scanners and care for patients during imaging. AI helps with image quality and positioning, but the hands-on patient work stays human.

### Can AI read scans better than doctors?

For some narrow tasks in research settings, AI matches or beats average performance. In real clinical use, performance often drops on different scanners, patients and hospitals.

### What is the lesson for other jobs?

A job is a bundle of tasks. AI can be excellent at one task and still leave the job intact, or even increase demand for it. Look at the whole job, not one task.

## Sources

- [Artificial intelligence-enabled medical devices](https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices)
- [Physicians and surgeons: Occupational Outlook Handbook](https://www.bls.gov/ooh/healthcare/physicians-and-surgeons.htm)
