Google have a neural network that takes a photo and returns a location

Google have trained a neural network that can determine with better accuracy than humans the geographical location given an arbitrary image; from the MIT Technology Review:

That’s impressive work that shows deep neural nets flexing their muscles once again. Perhaps more impressive still is that the model uses a relatively small amount of memory unlike other approaches that use gigabytes of the stuff. “Our model uses only 377 MB, which even fits into the memory of a smartphone,” say Weyand and co.

That’s a tantalizing idea—the power of a superhuman neural network on a smartphone. It surely won’t be long now!

That is indeed tantalizing, and while this trick is in some ways a gimmick, it’s a great part of an intelligent system. The hard part about this is to compose these skills into a system than can use the right skill at the right time, analyzing an image for location when needed, finding the shortest path when needed, seeing the trend in data structures when needed, etc.