You have probably seen at least one of those videos where some high-tech crime unit obtains a blurry, tiny image of some bad guy from a surveillance camera, and then effortlessly "zoom in" until they got a perfect portrait image out of a few pixels.
Well, seems that Google just claimed to have invented such technology.
But don't let the images fool you: "Deep neural networks" have already been shown to be easily fooled when classifying pictures, and I dont's see why a tool like the great Ostrichinator could not just as easily fool Google's "guessing zoom" to derive the picture of some arbitrary person from just any seemingly innocent 8x8 pixel sample.
It looks like you're new here. If you want to get involved, click one of these buttons!