Abstract Transition state search is the key to elucidating reaction mechanisms and exploring reaction networks in chemistry. However, due to the complexity of the potential energy surfaces, the search for accurate 3D transition state structures relies on a large number of quantum chemical calculations. For complex reactions, obtaining the transition state structure of a primitive reaction can take tens of hours or even days, which is relatively inefficient. MIT researchers have developed a generative AI model that is able to predict reaction transition state structures in almost 6 seconds with very high accuracy. This model provides another relatively convenient tool for non-computational chemistry faculty at universities to explore reaction mechanisms and design new chemical reactions. Meanwhile, with the deep integration of generative AI and chemistry education, generative AI is expected to become an important booster in the field of chemistry education.
DENG Song, WU Xiao-Chun. ChatGPT-Like Generative AI:Good Helper in Chemistry Research and Education[J]. Chinese Journal of Chemical Education, 2024, 45(18): 14-21.
