151. Battery Chemistry GPT

Engineering & simulationGPT-style AI model (University of Michigan)200,000 node hours on 34 petaflop Polaris supercomputer

The team trained a large language model on atomic structures of small organic molecules to predict chemical properties and identify feasible electrolyte-electrode pairings for batteries, reducing the need for manual hypothesis generation.

“they took that pretty much same approach where they gave this model that they built, the atomic structures of the, the molecules of interest. I think mostly hydrogen, carbon oxygen and nitrogen. And uh, all these different combinations of for the molecules and they didn't tell it any of the chemical properties.”

Engineering & simulationPolaris supercomputer34 petaflops; 200,000 node hours

The researchers utilized a 34 petaflop supercomputer to train the AI model on thousands of small organic molecules, enabling the brute-force computation required to predict missing atoms and infer compound properties.

“Well, the answer here is they're using a 34 petaflop supercomputer called Polaris, that I guess they have access to? Um, 34 pedophot means it can do 34 quadrillion calculations every single second. Um, This is as super as supercomputers can get.”