According to a recent report by the American Fun Project website, scientists from Tohoku University in Japan and Massachusetts Institute of Technology in the United States have successfully developed a new artificial intelligence (AI) model called GNNOpt. This model can predict the optical properties of materials with the same accuracy as quantum simulations, but at a speed one million times faster. The research team stated that this important progress is expected to accelerate the development of photovoltaic and quantum materials. The advancement of fields such as solar cells, photonic integrated circuits, and quantum computing relies on a deep understanding of the optical properties of materials. However, existing experimental methods, such as laser testing, are limited by the wavelength range of light waves. However, simulation calculations are expensive and require strict standards to be met. Therefore, scientists have been searching for alternative methods to quickly predict the optical properties of different materials. Previously, Graph Neural Network (GNN) machine learning models had already been developed. This model represents atoms as nodes in a graph and atomic bonds as edges, vividly representing molecules and materials. However, GNN faces difficulties in capturing subtle differences between complex crystal structures, which limits its widespread application in predicting material properties. The new AI model takes a different approach, using the crystal structure of materials as input, and can predict the optical properties of materials with astonishing accuracy in a wider range of light frequencies in a very short period of time. Once scientists master a certain optical property, they can use relevant formulas to derive other optical properties. The secret to the success of the new AI model lies in the "integrated embedding" technology. This technology endows AI with the ability to learn from multiple datasets, making it more precise and versatile. The research team claims that their new AI model can accurately predict the optical properties of crystal structures, opening the door to a wide range of applications, especially providing strong support for the screening of advanced solar cells and quantum materials. They plan to create a comprehensive database containing various material properties such as mechanics and magnetism to further expand the functionality of the AI model. (New Society)