Giving AI a 'crystal clear' view of the materials world by mapping geometry to text
In principle, artificial intelligence could make finding the right material for an application as simple as ordering dinner: Scientists would tell a large language model what they want, and it would serve them the perfect material for the job.
Artificial intelligence (AI) has the potential to simplify the process of finding suitable materials for various applications. However, current methods struggle due to the use of 3D atomic structures rather than text streams, which are the basis of today's large language models. Researchers at MIT have developed a multimodal AI model that addresses these issues by combining 3D atomic structures with natural language descriptions.
This unified description includes chemical composition, structure, properties, processing considerations, and potential applications, enabling AI to gain a deeper understanding of materials. The model was trained using contrastive learning, which compares similar and dissimilar examples of structures and textual descriptions to achieve a unified understanding.
The researchers demonstrated that their model could identify better material designs than those trained on text streams alone, as it could "see" the atomic arrangements and bonding patterns. When asked to identify materials suitable for solid-state lithium-ion battery electrolytes, the model's top results included materials that were independently identified as promising candidates.
The researchers believe this multimodal approach can revolutionize materials science by providing a more comprehensive understanding of crystalline structures and their properties, potentially leading to breakthroughs in material design and discovery.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
