MIT Researchers Develop AI Model to Predict Chemical Reactions
The new system, named FlowER, incorporates physical constraints like conservation of mass to generate more realistic and reliable scientific predictions.
Overcoming "Alchemy" in AI
Researchers at the Massachusetts Institute of Technology (MIT) on September 3, 2025, announced a new generative AI system designed to predict the outcomes of chemical reactions with greater accuracy. The model, named FlowER (Flow matching for Electron Redistribution), addresses a key limitation of previous AI approaches by grounding its predictions in the fundamental laws of physics.
Many earlier attempts to use large language models for this task failed to respect physical constraints. A researcher on the team noted that some models could generate outputs that were not physically possible, such as creating or deleting atoms …
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.