Deep learning turns mono recordings into immersive sound

We’ve had 3D images for decades, but effectively imitating 3D sound has always eluded researchers. Now a machine-learning algorithm can produce “2.5D” sound by watching a video.