MIT CSAIL researchers have created an AI-based method for low-variance sampling that evenly distributes data points, improving the accuracy of simulations.