The Ministry of Science and ICT is advancing the development of artificial intelligence (AI) technology that accurately interprets and predicts scientific phenomena by leveraging not only data but also physical and mathematical principles. It is anticipated that AI will become a core technology for accelerating scientific discoveries and research innovation, beyond being a mere research assistance tool.
On the 24th, the Ministry held a kick-off meeting for the “Next-Generation AI + Science and Technology (S&T) Foundation Technology Development Project” and announced four principal researchers: Professor Hao Su (KAIST), Professor Jae-Seok Yu (DGIST), Professor Min Hwa Oh (Seoul National University), and Professor Cheol-Hee Yoon (KAIST). Professors Hao Su and Jae-Seok Yu will lead the development of physics- and mathematics-based AI interpretation and prediction models, while Professors Min Hwa Oh and Cheol-Hee Yoon will focus on next-generation AI architecture research.
The Ministry highlighted that while AI is rapidly being utilized in scientific research and industrial fields, most models operate by learning vast datasets to predict outcomes. However, in science and engineering domains where physical laws and mathematical principles are critical, AI often struggles to explain its results and faces accuracy limitations under new, untrained conditions.
To address this, the Ministry aims to develop a general-purpose AI model that significantly enhances the efficiency and accuracy of scientific research across disciplines. The project will invest a total of 20 billion KRW over six years, from 2024 to 2031. The Ministry will provide computing infrastructure, such as GPUs, for selected tasks, and the research data and AI models developed will be made publicly available.
In the physics- and mathematics-based AI interpretation and prediction model sector, Professor Hao Su's team will develop a mathematical and physics-based causal AI model that infers causal structures and governing equations embedded in data, even when conditions and environments change. Professor Jae-Seok Yu's team will implement an “AI physicist” that visualizes which physical laws dominantly govern various phenomena, enabling the AI to autonomously judge and select interpretation strategies.
These two projects are expected to reduce the cost and time required for design, interpretation, and verification, eliminating the need for repeated validation whenever research conditions change.
In the next-generation AI architecture research sector, Professor Min Hwa Oh's team will mathematically identify the structural limitations of current AI models, where computational demands surge with longer contexts, and develop a new AI architecture and learning method capable of efficiently processing extended contexts. Professor Cheol-Hee Yoon's team will mathematically prove the principles by which generative AI models enhance performance during learning and derive a next-generation AI development methodology applicable throughout the entire process—from structural design and learning to reliability verification.
These projects are also expected to lower dependency on vast datasets and computational resources while designing AI models that operate stably, thereby enhancing the self-reliance and competitiveness of domestic AI technology.
Yoon Kyung-sook, Director-General of the Ministry's Basic and Source Technology Research Policy Division, stated, “Securing AI models based on physical and mathematical laws will enhance the rigor and accuracy of scientific research, providing practical benefits not only to research fields but also to industries such as semiconductors and batteries. Additionally, the development of new AI model architectures will contribute to the advancement of AI technology itself.”
