In a recent blog post from Pokémon GO developer Niantic, the company has laid out its plans for what it calls a 'Large Geospatial Model', an AI that fills in the blanks for geographical locations that haven't been physically captured in person.
So for example, as Niantic highlights (thanks, IGN), if you only have a view of the front of a church, the AI can take what it knows from thousands of scanned churches on a global scale and essentially 'complete' the model.
Naturally, a shedload of data is required for this, but luckily for Niantic, thousands upon thousands of people have been scanning locations with their phones using apps owned by Niantic. This includes, of course, Pokémon GO.
The company says it now has around 10 million scanned locations around the world, about 1 million of which have been activated by a new service it called 'Visual Positioning System'. This, according to Niantic, "uses a single image from a phone to determine its position and orientation using a 3D map built from people scanning interesting locations in our games".
Niantic stresses that the data obtained from its games and VPS service is especially unique, as it largely comes from pedestrian locations not accessible by vehicle.
It's not known exactly what Niantic plans to go with its Large Geospatial Model in the long-term, but it states that the idea goes well beyond gaming, and could potentially be utilised for "spatial planning and design, logistics, audience engagement, and remote collaboration". Cool, cool.