Artificial Intelligence and Large Language Models Usher in a New Era in Chemical Synthesis
Researchers at the University of Washington are automating chemical synthesis and discovering production recipes for new materials using artificial intelligence and large language models.
University of Washington researchers are taking artificial intelligence and large language models a step further in chemistry and materials discovery, going beyond mere structure prediction to reveal actual production recipes.
Artificial Intelligence and Chemical Synthesis
While artificial intelligence does not replace human intelligence, it goes beyond human endurance to become a powerful tool for chemistry and materials discovery. Scientists now want not only to predict new structures, but also to produce these materials themselves.
Analysis of Scientific Works
Zhiling Zheng, Assistant Professor in the Department of Chemistry at the University of Washington, discusses in her award-winning article in the journal Science how AI systems can overcome this process by reading chemistry like a chemist.
Metal-Organic Frameworks
Zheng trained large language models on a literature-based dataset containing approximately 4,000 linker transformations. Through design agents and computational simulations, 10 new materials were discovered that showed stronger water-harvesting performance than existing aluminum-based adsorbents.
Data Management in Polymer Synthesis
Christopher Cooper from the McKelvey School of Engineering published an article in the journal Matter documenting how data for polymer synthesis should be curated. This process involves converting data into a form that can be easily digested by artificial intelligence models.
Dynamic Polymer Library
Cooper and Kathryn Miller from the National Institute of Standards and Technology developed an automated approach that created the Dynamic Polymer Annotated Library, which helps discover dynamic polymers.
The New Role of Scientists
These developed tools increase efficiency, allowing human scientists to act more like managers rather than spending hours working through trial and error.