New AI algorithm is able to invent more than 31 million previously non-existent materials

Scientists at the University of California San Diego have developed an artificial intelligence algorithm that can almost instantly determine the structure and dynamic properties of any materials, both existing and new. The algorithm, which they call M3GNet, has been used to predict 31 million yet-to-be-synthesized materials, the scientists say, and could help create materials “with exceptional properties.” This study tells magazine Nature Computational Science.

The M3GNet algorithm is currently being used to find safer and more energy efficient electrodes and electrolytes for rechargeable lithium-ion batteries. Scientists compare it to the AlphaFold algorithm, which was developed by Google’s DeepMind to predict protein structures and became a breakthrough technology. “Similar to how they do it with proteins, we need to know the structure of the material in order to predict its properties,” explains Shayu Ping On, professor at the University of California at San Diego. Now scientists intend to significantly expand the number of materials that make up the database of materials not yet synthesized, and will continue to work to identify the most promising new materials.

Alena Miklashevskaya