The system is described in a SIGGRAPH technical paper, “Neural Supersampling for Real-time Rendering.” It is based on a machine learning approach that turns low-resolution input images into high-resolution outputs for real-time rendering. "This upsampling process uses neural networks, training on the scene statistics, to restore sharp details while saving the computational overhead of rendering these details directly in real-time applications," wrote the team.
The team states that this new approach achieves |16x supersampling of rendered content with high spatial and temporal fidelity."
You can learn more here.
