AI Advances in Materials Prediction and Design: A Review

A recent publication on arXiv (2607.21660) examines the advancements, obstacles, and outlook regarding generative and multimodal AI in the context of materials prediction and design. The authors contend that AI facilitates the efficient investigation of chemical and structural domains for discovering new materials; however, substantiating claims of novelty is challenging due to the necessity for chemical plausibility, structural uniqueness, property relevance, and the ability to realize experiments. They propose a hierarchy of materials properties, ranging from intrinsic, composition-based characteristics to extrinsic, processing-influenced performance, to clarify constraints on deployment and differentiate types of novelty. The analysis highlights the need for diverse data types, as current evidence primarily focuses on composition and idealized structures. This paper was released on arXiv on July 26, 2025.