SE155: Research competitiveness in the AI era: Synergistic innovation through scientific insight and intelligent tools

Artificial intelligence, machine learning, high-throughput computation, large databases, and laboratory automation are rapidly transforming materials and molecular research. However, advanced tools alone do not guarantee meaningful scientific discovery. Their effectiveness depends on the formulation of appropriate scientific questions, the selection of physically meaningful descriptors, and the integration of domain knowledge with data-driven methods.This presentation discusses how scientific insight and intelligent tools can work synergistically to accelerate materials discovery. A hierarchical research strategy is introduced, progressing from principle-based reasoning and descriptor-guided design to machine-learning screening and computational or experimental validation. Lithium superionic conductors are used as an illustrative example. Two simple, interpretable descriptors, the effective diameter of the Li-ion diffusion pathway and the dimensionality of the connected Li-ion network, capture the essential requirements for rapid ion transport. These descriptors successfully screen experimentally known conductors and identify promising candidates from a large materials database.The central message is that AI should complement, rather than replace, scientific reasoning. Combining physical intuition, interpretable models, and modern computational tools can improve efficiency, transparency, and research competitiveness in the AI era.