MIT and Google's Breakthru AI System

Creates Highly Realistic 3D Images
Source:  Google showing 2D & 3D images

Google & MIT Research Team
Researchers at MIT's Computer Science and Artificial Intelligence Lab and Google have created a next generation AI system capable of creating 3D shapes with realistic texture, lighting and reflections.

It's called VON or Visual Objects Network.  It not only generates images more realistic than most state of the art systems.  It also enables texture and shape editing, viewpoint shifts and other 3 dimensional tweaking.

3D & 2D
Most systems focus on generating 2D images, which limits their practical use in many fields.  VON synthesizes 3D shapes and 2D images.  The researchers next step is fine modeling for producing shapes and images at a higher resolution, disentangling texture into lighting and appearance and synthesizing natural scenes.  Their research was just presented and accepted at NeurlPS 2018 in Montreal.