Meta Unveils Cutting-Edge Datasets and Models for Science and Language

Meta's AI research division unveils groundbreaking datasets and models, including Open Molecules 2025 for molecular discovery, and insights into how language representations develop in the brain, mirroring large language models, potentially revolutionizing science and language processing.

• Meta FAIR is releasing new datasets and models for molecular property prediction, language processing, and neuroscience.
• The Open Molecules 2025 dataset and Universal Model for Atoms aim to accelerate molecular and materials discovery.
• Meta FAIR and the Rothschild Foundation Hospital conducted a study revealing how language representations emerge in the brain during development, with parallels to large language models.