There are now more and more photographs of more and more things and places sitting on computer storage on phones, computers and servers all over the world.
This little bot would crawl through everyones photos, reading tags and timestamps, to collate photos of the same place or object from different
points of view. It would then use these groups to build 3D surface models of as much as possible of the entire world and the static objects in it.
Both for privacy and practicality the system would ignore anything moving or human. It should also probably ignore any written or printed text. The idea is for it to be sufficiently privacy-oriented that people will willingly allow it access to their photographs.
Presumably it should not be too difficult to combine geolocation and the resulting 3D models to restrict access to private places like inside your home, giving legal residents or persons with autorised access to a place the right to own or control access to the relevant part of the model.
The system could also put up flags requesting better imaging of places and objects in the world which it doesn't have enough images of, similar to how the crowdsourced Geograph website indicates places that are under-photographed. People could compete to complete the imaging of difficult-to-access or remote locations.