Op-Ed
[Op-Ed] Toward Faithful AI That Understands Humans and Follows Reality
Jong Chul Ye, Endowed Chair Professor at KAIST Kim Jaechul Graduate School of AI, leads NAIRL’s research on high-dimensional generative foundation models. His column starts from a question medical imaging has come to face: as AI restores blurry MRI and CT scans with remarkable quality, is it actually understanding the human body, or just producing plausible pictures?
In medicine, plausibility is dangerous. What is needed, Professor Ye argues, is “faithful AI,” intelligence grounded in the structure of the body, the laws of physics, and biological principles. The same principle extends to other high-dimensional data, which is precisely the direction of the foundation models being developed at NAIRL.
Realizing this vision, he writes, requires interdisciplinary talent, an institutional foundation for safely using medical data, and the long-term support that foundation model research demands.
This column is part of NAIRL’s biweekly series “AI Talent Powerhouse,” published in ZDNet Korea, which examines the direction of Korea’s AI research ecosystem.