Multimodal AI and Self-driving Laboratories: Emerging Trends in Materials Chemistry

Abstract

This article highlights multimodal AI and self-driving laboratories as emerging approaches in data-driven research and development, particularly in materials chemistry. As data science becomes essential across disciplines, advances in machine learning and accessible robotics have lowered the barriers between cyber and physical domains. However, human researchers remain central in defining goals, designing strategies, and taking responsibility. Multimodal AI integrates diverse data types - such as spectra, images, compositions, and process conditions—enabling insights beyond single-modality analysis. The authors redefine this framework for materials science and demonstrate its effectiveness across various material systems. Self-driving laboratories, in contrast, employ closed-loop frameworks that update experimental conditions based on real-time data, allowing efficient exploration and exploitation, especially in complex systems with limited prior data. These approaches are not merely extensions of automation but require domain knowledge to design and interpret models. Future progress depends not only on technological advances but also on the development of expertise, collaborative environments, and research cultures that fully leverage data-driven methods.

graphical abstract