Chinese researchers have developed a new framework to train machine learning models with prior knowledge, such as the laws of physics, in order to create “real AI scientists” capable of improving experiments and solving scientific problems. This could potentially bridge the gap between deep learning models and their ability to accurately model the real world.
