A decade after the ‘Godfather of AI’ stated radiologists were obsolete, their salaries are as high as $571K and demand is growing rapidly

(SeaPRwire) –   In 2016, the “Godfather of AI,” Geoffrey Hinton, delivered a stark warning on stage at a machine learning conference in Toronto, declaring that artificial intelligence would soon render radiologists obsolete.

At the time, he argued that training new radiologists should be halted altogether, asserting it was “completely obvious” that within five years—or at most ten—AI would surpass humans in performing the same clinical tasks.

“If you work as a radiologist, you’re like the coyote that’s already over the edge of the cliff but hasn’t yet looked down,” Hinton remarked.

For years, technology leaders such as Hinton predicted widespread job displacement by AI, partly because certain radiology duties appear routine and repetitive—such as interpreting medical images and drafting reports.

Yet, despite these dire forecasts, radiology may serve as a counterexample to claims that AI will inevitably replace human workers. Even Hinton has since moderated his stance; last year, he clarified that his comments were narrowly focused on image analysis alone, according to reporting by the New York Times. In future practice, he suggested, human radiologists will collaborate with AI systems to enhance both efficiency and accuracy.

Over the past decade, the number of practicing radiologists in the United States has increased by approximately 10%, stated Christoph Herpfer, an economist and professor of business administration at the University of Virginia’s Darden School of Business, who specializes in healthcare finance and physician labor markets.

“We actually face a significant shortage of radiologists—the exact opposite of what was predicted,” he told .

It is important to note that overall demand for healthcare jobs and services has risen steadily due to America’s aging population and expanded access to care under the Affordable Care Act.

Nevertheless, concerns persist about AI-driven job elimination across industries. Tech firms like Snap and Block have cited AI advancements as justification for mass layoffs, while Anthropic CEO Dario Amodei has warned that AI could displace half of all entry-level white-collar roles within five years.

As of March, there were roughly 4,333 open positions for radiologists, with companies taking an average of 130 days to fill each vacancy, according to data from radboard.io, a specialized job board for radiologists. This persistent shortage has driven the average annual salary for radiologists to $571,000 in 2025—a 9% increase compared to the previous year, per a Medscape study.

Once heralded as doomed by AI experts, radiology has become an example of how automation can coexist with—and even augment—human expertise.

Nvidia CEO Jensen Huang recently clarified on the Dwarkesh Podcast that many predictions of radiology’s demise conflate two distinct responsibilities: reading scans versus undertaking the full scope of a radiologist’s role. Similarly, Netflix cofounder Reed Hastings emphasized on the Possible podcast that radiology illustrates how even highly automated fields may retain human elements.

“We tend to focus on scenarios where AI completely eliminates entire professions,” Hastings noted. “But again, this hasn’t happened in radiology. Maybe it will eventually, but certainly not within the last five years.”

Structural barriers further limit AI’s capacity to fully supplant radiologists. For instance, Medicare and Medicaid only reimburse for radiology studies when performed by a licensed physician who provides the final interpretation. Additionally, legal and ethical uncertainties surround accountability if an AI system fails to detect a critical diagnosis.

Like Huang, Herpfer highlighted another key reason radiology remains resilient: image interpretation constitutes just one component of a radiologist’s daily workload. Radiologists also consult with referring physicians, monitor patients’ conditions, and—especially those specializing in interventional radiology—perform hands-on procedures. By automating scan review and report generation, AI simply frees up more time for these other essential activities, he explained.

“Complex professions like medicine consist of numerous subtasks. Even if AI automates one or two of them, it merely expands the time available for the remaining responsibilities,” Herpfer said. “As long as AI cannot execute every aspect of the job, the profession itself won’t disappear.”

Surging demand for diagnostic imaging—partly fueled by FDA-approved AI tools that make scans faster and more affordable to produce—has kept radiologists busier than ever. Between 2018 and early 2025, radiology caseloads surged by 25%, reported the Journal of the American College of Radiology.

“The volume of studies radiologists must read annually continues to climb, while reimbursement rates keep declining, resulting in extreme burnout,” said Dr. Jeff Chang, a former emergency room radiologist who cofounded RadAI. “They’re just overwhelmed.”

Chang knows this challenge intimately: before launching his company in 2018, he spent a decade reading between 150 and 200 imaging studies per night shift. His startup’s AI platform aims not to replace radiologists but to save them nearly an hour each shift by automatically generating conclusions sections in radiology reports.

Despite leading a company integrating AI into radiology workflows, Chang remains skeptical that technology can ever fully supplant the profession.

“That whole idea never really made sense,” he said.

In practice, working radiologists emphasize that AI lacks the emotional intelligence required for patient care. Dr. Tonie Reincke, an interventional radiologist based in Texas, pointed out that machines cannot offer compassion, empathy, or nonverbal cues during patient interactions.

“A computer can’t hold a patient’s hand when they’re crying,” she said. “It can’t hand them a tissue either.”

Reincke fears that exaggerated fears of AI-induced obsolescence might deter medical students from pursuing radiology—a field already among the most competitive specialties, requiring lengthy residencies lasting five or six years depending on subspecialty training. Worsening this issue, she added, could deepen the existing shortage of radiologists.

Herpfer drew a broader lesson from radiology’s evolution, noting a similar pattern emerged with accountants in the 1990s. Then, spreadsheet software was expected to eliminate their jobs entirely—but instead, it freed them from repetitive calculations so they could focus on higher-value advisory services.

“Unless AI achieves a breakthrough toward true Artificial General Intelligence (AGI), most occupations are likely to remain stable in the medium term,” Herpfer concluded. “That’s the key insight we should take from the experience of radiologists.”