Researchers at NVIDIA recently showcased an interactive world created and rendered using an AI model. They achieved this by combining machine learning and computer graphics to perform image generation using deep learning based neural networks. The model was trained using real footage from driving sequences in a number of different cities, the UE4 engine was then used to generate a high-level layout of objects such as other cars, buildings and scenery which is converted into an image by the trained network. They hope this new advancement will make creating a virtual world both faster and significantly cheaper.
You can read more here!