Niantic Uses Pokémon Go Data to Build AI Geospatial Navigation System

The company has leveraged millions of player-generated scans to create a 3D mapping model aimed at advancing AR, robotics, and autonomous systems.

Niantic announced its Large Geospatial Model (LGM), trained on data from Pokémon Go and other apps, to help AI navigate and interact with physical spaces.
The model utilizes over 10 million scanned locations worldwide, with 1 million new scans contributed weekly by users capturing images from pedestrian perspectives.
Niantic's Visual Positioning System (VPS) underpins the LGM, allowing precise mapping of areas inaccessible to cars or traditional mapping tools.
The technology aims to support applications in augmented reality glasses, robotics, and autonomous systems, though concerns about potential misuse, including military applications, have been raised.
While Niantic states the data collection is opt-in, critics highlight transparency and ethical concerns, as many players were unaware their gameplay contributed to this AI model.