In 2016, the “godfather of AI” Jeffrey Hinton took the stage at a machine learning conference in Toronto and declared that AI would soon kill radiology professionals.
At the time, he said we should stop even training new radiologists because it is “absolutely clear” that AI will be able to outperform humans at the same tasks within five years (or at most 10).
“Working as a radiologist, you’re like a coyote who’s already over the edge but hasn’t looked down yet,” Hinton says.
Technology experts like Hinton have been predicting AI replacement for years, in part because some radiologists’ jobs seem routine and repetitive, like reading scans and writing reports.
But despite doomsday predictions, radiology could be an example of how warnings about AI-based job replacement may be overstated. Even Hinton has backed away from his bold statements from last year to make it clear that he is talking purely about image analysis. new york times Reported. In the future, human radiologists will work with AI to become even more efficient and effective, he said.
Over the past decade, the number of practicing radiologists in the United States has increased by about 10%, said Christoph Helper, an economist and professor of business administration at the University of Virginia’s Darden School of Business who studies health care finance and the physician labor market.
“In fact, there is a huge shortage of radiologists, so the exact opposite of this prediction is happening,” he said. luck.
Indeed, as Americans age and more people enroll in health insurance under the Affordable Care Act, the demand for health care jobs and health services overall has steadily increased.
Much has been said about AI replacing human jobs. Some tech companies, including Snap and Block, have cited AI as a reason to lay off thousands of employees, while Anthropic CEO Dario Amodei, who previously warned that AI could eliminate half of entry-level white-collar jobs within five years, has recently argued that AI could transform and expand certain types of jobs rather than simply replacing them.
Still, as of May, there were 7,469 radiology job openings, of which 1,470 had been open for more than 60 days, according to an analysis by RadBoard.io, a recruiting platform specializing in radiology. This shortage will push the average salary for radiologists to $571,000 in 2025, up 9% from the previous year, according to a study by Medscape.
A few years ago, AI experts declared that radiology jobs would disappear, but some technology leaders have recently changed their tune.
Nvidia CEO Jensen Huang said: dwarkesh podcast Earlier this year, radiology doomers confused reading scans with the job as a whole. Netflix co-founder Reed Hastings said: Possible Radiologists, profiled in a podcast earlier this year, are an example of how even though their jobs may be affected by AI, they won’t be completely replaced.
“We’re attracted to these scenarios where AI wipes things out,” he says. “Again, this hasn’t happened in radiology. Maybe eventually it will, but not in the last five years.”
Structural issues may make it nearly impossible for AI to completely replace radiologists in the near future. Medicare and Medicaid will reimburse the cost of radiology tests only if the final exam is performed by a licensed physician. It is also unclear how AI will be held responsible for failure to diagnose.
Like Huang, Herpfer pointed out that another reason radiology has endured is because reading images is just one of its daily tasks. Radiologists consult with other doctors, observe patients, and some, like interventional radiologists, perform the actual procedures. Introducing AI automation for scan reading and reporting will simply free up doctors to spend more time on other tasks, argues Helpfer.
“A complex job like being a doctor is made up of many subtasks, and even if you can automate one or two of them, you’re just adding more time to other tasks,” he said. “Until AI can perform all tasks perfectly, jobs themselves will not disappear.”
Increased demand for scans, driven in part by AI tools already approved by the Food and Drug Administration that can make image processing cheaper and faster, is also keeping radiologists busy. From 2018 to early 2025, radiology caseloads jumped 25%, according to . Journal of the American College of Radiology.
“The number of studies that radiologists have to perform each year continues to increase, reimbursement continues to decrease, and it’s really, really exhausting,” said Dr. Jeff Chang, a former practicing ER radiologist and co-founder of AI healthcare startup Rad AI. luck.
Chan knows firsthand how burnout can occur among radiologists. He spent 10 years reading 150 to 200 imaging tests every night shift before starting his company in 2018. His company’s AI tools are designed not to replace radiologists, but to save radiologists nearly an hour per shift by automatically generating the conclusion section of radiology reports.
Despite being the co-founder of a company that brings AI into radiologist workflows, he doesn’t believe the technology can completely replace the profession.
“The whole concept didn’t really make sense to begin with,” Chan said.
Practicing radiologists in the field say AI cannot simulate the humanity involved in their work. Dr. Tony Reinke, a Texas-based interventional radiologist, said AI cannot provide sympathy, empathy, or the nonverbal cues that humans give when talking to patients.
“A computer can’t hold a patient’s hand when they’re crying,” she said. luck. “The computer can’t give them tissues.”
Reinke worries that the hype about AI completely replacing radiologists will discourage medical students. Radiology is one of the most competitive programs and requires one of the longest training periods in U.S. medicine, often lasting five to six years, depending on the specialty, she said. He added that concerns about AI, even if not well-founded, could worsen the shortage of radiologists.
Helpfer said the broad lessons from radiology go far beyond medicine. The same dynamic occurred with accountants, who were predicted to be wiped out by spreadsheet software in the 1990s. Instead, Excel eliminates routine number-crunching tasks, freeing accountants to focus on more complex advisory work.
“Unless AI takes the leap to becoming a type of AGI. [Artificial General Intelligence]”Unless this extreme scenario happens, most jobs will probably be pretty safe in the medium term,” Helper said. “I think this is a lesson we can learn from radiologists.”
A version of this article first appeared on Fortune.com on May 4, 2026..
Technology details
- Airbnb CEO Brian Chesky’s X account was hijacked in a malicious hack of AI promoting cryptocurrency tokenization
- Kevin O’Leary claimed that opposition to Utah’s data center was fueled by Chinese money. He and FOX News are currently being sued for defamation.
- As streaming competition drives content spending to $20 billion, Netflix leverages AI to produce 17-minute documentary ‘twice as fast and at half the cost’
