The tipping point for AI in medicine: how Korean doctors see the future of human judgment

The writing was on the wall long before the technology caught up. Korea’s aging population has stretched its healthcare system to the brink. Specialists are in short supply, and the costs of care continue to climb. For many, AI appeared to offer a lifeline—a chance to turn desperation into innovation.

But as the technology matures, a new question emerges: Can AI move beyond mere assistance to claim the mantle of independence—and should it?

In Sweden, the revolution has already begun. The country’s double-reading mammography system, once handled exclusively by two human radiologists, now includes an AI system, Lunit INSIGHT MMG, as one of the reviewers. Across the border in Germany, the shift has been even bolder: last year, AI was mandated as the primary diagnostic tool in specific scenarios.

For the cautious healthcare system, these developments seem almost radical. “Germany’s decision is a turning point for AI in medicine,” remarked an official from Coreline Soft, a Korean medical AI company. “It’s hard to imagine Korea adopting such a policy, but the global momentum is undeniable. It seems likely other nations will follow suit.”

For now, physicians remain the gatekeepers of clinical decisions—but the growing influence of AI is hard to ignore.

(Credit: Getty Images)

“In 10 years, AI will touch nearly every field, not just imaging,” predicts Professor Choi Joon-il of radiology at Seoul St. Mary’s Hospital. “In 20 years, AI will fully take over the roles of supporting physicians. Naturally, this will reduce the demand for doctors.”

Choi’s confidence is bold, though not universally shared. Professor Jeong Won-gi of radiology at Chonnam National University Hwasun Hospital tempers the excitement with pragmatism.

Jeong has worked extensively with Coreline Soft’s AVIEW LCS, an AI tool for detecting pulmonary nodules. “It’s transformed how we work,” Jeong acknowledges. “I can’t imagine going back to the old days of manually reviewing every scan.”

Yet, even as he sings AI’s praises, Jeong points to its flaws. “Measurement discrepancies remain an issue, and AI occasionally flags non-nodules as nodules. It’s not perfect,” he says, with the air of someone noting a quirk rather than a critical fault. A study published in the European Journal of Cancer found that while AI outperformed radiologists in detecting large pulmonary nodules—31 errors compared to 407 by humans—it was still far from infallible.

Still, Jeong is pragmatic. “Considering Korea’s aging population and limited resources, adopting AI isn’t just likely—it’s inevitable. There’s a strong possibility that AI will eventually function independently in certain capacities.”

The human factor

At Soon Chun Hyang University Hospital in Cheonan, Professor Kim Jin-young, a specialist in respiratory allergies and critical care, approaches AI with a careful, almost skeptical eye.

Kim has worked with VUNO Med-DeepCARS, an AI system designed to predict cardiac arrest. “AI is excellent at processing data,” he concedes. “But the system relies on manual input. Nurses still check vitals, measure temperatures, and feed this information into the machine. It’s helpful but clunky.”

“False alarms are common,” he adds. “These systems can overwhelm staff with alerts, many of which turn out to be nothing.”

For Kim, AI’s greatest limitation is more philosophical than practical. “AI can’t replicate human judgment. It doesn’t understand nuance. It can’t weigh the emotional or contextual factors that shape clinical decisions. And, frankly, it never will.”

Professor Choi Joon-il of Seoul St. Mary’s Hospital shared his forward-looking views on AI in medicine, reflecting themes raised by other medical AI experts Korea Biomedical Review spoke with.

This sentiment resonates with Lee Hae-min, professor of internal medicine and president of Incheon Christian Hospital. Lee’s team has embraced Lunit Insight CXR, an AI system for interpreting chest X-rays. “The speed is remarkable,” Lee says. “In emergencies, every second counts, and this technology helps us act faster.”

But even Lee, an AI advocate who envisions a future where AI integrates patient histories, symptoms, and clinical data to deliver comprehensive diagnoses—particularly for diseases like cancer—draws a clear boundary. “Patients don’t trust AI. They trust us—the doctors. No matter how advanced the technology becomes, it’s our responsibility to communicate, empathize, and take accountability. AI can’t do that. It never will.”

For all its potential, AI carries risks that extend beyond the technical flaws. Professor Choi warns of a future where over-reliance on AI erodes the very skills that make doctors indispensable. “In the past, we diagnosed with stethoscopes and intuition,” he reflects. “Now, many doctors can’t function without a CT scan. Imagine a future where trainees rely entirely on AI. If the system fails, they might not even notice it.”

Despite these concerns, most experts agree that AI’s role in medicine will only grow. Its ability to analyze vast datasets, identify patterns, and deliver rapid results makes it an invaluable tool. A study published in the European Journal of Radiology, for example, found that Lunit INSIGHT CXR reduced chest X-ray triage time by 77 percent.

But tools, no matter how advanced, are not replacements. “AI is a partner, not a competitor,” says Lee. “It can assist, it can augment, but it will never replace what makes us human.”