DeepMind AI tackles one of chemistry’s most valuable techniques

The artificial-intelligence company DeepMind has developed a new machine-learning model to predict the density of a molecule’s electrons — a key step to calculating its physical properties. The algorithm relies on a technique called density functional theory (DFT), which has been hugely successful in chemistry, biology and materials science. But DFT goes a bit wonky in some cases, because it’s not a perfect reflection of the complex quantum mechanics that govern matter. Researchers trained an artificial neural network on data from hundreds of accurate solutions derived from quantum theory, plus some handy laws of physics.