TLDR: During a press briefing in Seoul on July 8, 2025, Professor Su-In Lee of the University of Washington emphasized the critical need for Explainable AI (XAI) in South Korea’s technology development, particularly in sensitive fields like medicine. She advocated for understanding AI’s decision-making processes to build trust and ensure reliability, while also cautioning against an exclusive focus on sovereign AI development, suggesting that adopting superior foreign technologies and investing in foundational research like XAI can enhance domestic efforts.
Seoul, South Korea – On July 8, 2025, Professor Su-In Lee, a distinguished expert in artificial intelligence from the University of Washington, delivered a pivotal press briefing in Seoul, underscoring the urgent necessity for Explainable AI (XAI) in South Korea’s burgeoning technology landscape. Professor Lee, renowned for developing the widely adopted SHAP (Shapley Additive Explanations) framework and the first female recipient of the prestigious Samsung Ho-Am Prize in Engineering, highlighted that understanding how AI reaches its predictions and decisions is paramount for building trust and ensuring the reliability of AI systems, especially in life-critical sectors such as medicine.
Professor Lee articulated that the ‘black box’ problem, where AI’s reasoning remains opaque, poses significant risks as AI integrates into fields with profound social impacts, including healthcare, autonomous driving, and law. She stressed, ‘When AI makes mistakes, humans must be able to understand and correct those mistakes,’ emphasizing that explaining complex AI processes in an understandable manner is the core of XAI. Her SHAP framework quantifies the influence of specific information on AI model judgments, a technology currently utilized across diverse sectors like bio and semiconductor industries, boasting over 35,000 citations.
Citing practical examples, Professor Lee explained the limitations of current AI without XAI. In the context of Alzheimer’s disease, she noted, ‘The disease involves complex interactions among multiple genetic variables. It’s not enough for AI to merely point out one gene as a culprit. To develop effective treatments, we need AI that can explain why a gene matters.’ She also criticized dermatology apps designed to detect skin cancer, pointing out their lack of transparency in analysis, which undermines their reliability and has led to misdiagnosis cases in her research.
Addressing South Korea’s strategic direction in AI, Professor Lee acknowledged the government’s plans to develop its own foundation models and expand into domain-specific AI, including healthcare applications. However, she cautioned against an insular approach. ‘It’s essential to selectively adopt superior foreign AI if it better suits the task,’ she advised, advocating for flexibility rather than an exclusive focus on domestic development. She believes that deeper investment in foundational research, such as XAI, can significantly enhance the quality and robustness of sovereign AI efforts. ‘SHAP and other frameworks are grounded in mathematics and game theory,’ she added, emphasizing that ‘This kind of theoretical strength is only possible through long-term funding and commitment.’
Also Read:
- Google Cloud Enhances Sovereign Cloud Capabilities in Korea, Expanding Local Data and AI Processing
- Japan’s Generative AI Adoption Trails Global Leaders, Ministry Report Reveals
Professor Lee’s visit to Korea also included delivering a keynote speech at the ‘2025 World Korean Science and Technology Conference’ on July 9, further solidifying her message on the importance of transparent and trustworthy AI for future technological advancements.
