Interesting that they speculated that it was an algorithm that was trained by comparing a manually drawn map with imagery. US cities are all fairly uniform as t

Interesting that they speculated that it was an algorithm that was trained by comparing a manually drawn map with imagery. US cities are all fairly uniform as they are pretty much all based on a grid, but they vary in colour significantly from Oregon to Arizona. I wonder what level of granularity they would need to go to in Europe? Is London different enough from Paris that an algorithm trained on Paris would not work London? In Britain you would need to recognise different cultural structures that you wouldn't necessarily get in Paris, one example is many British parks have a raised up circular bandstand for a local brass band. So they probably need to recruit quite specific people for each locality.