Niantic, the company behind the popular game Pokémon Go, announced the development of a new “Large Geospatial Model” (LGM), an artificial intelligence model that uses millions of real-world scans collected from Pokémon Go players’ smartphones and of other Niantic products. This model, as explained in a company blog post, will allow computers and robots to interact with the world in completely new ways. The LGM is based on Niantic’s “Visual Positioning System” (VPS), a system that uses images taken by users to determine the position and orientation of a device within a 3D map.
“For the past five years, Niantic has focused on building our Visual Positioning System (VPS), which uses a single image from a phone to determine its location and orientation using a 3D map built by people scanning interesting locations in our games and in Scaniverse,” the post reads. “This data is unique because it is taken from a pedestrian perspective and includes places that are inaccessible to cars.” Victor Prisacariu, Chief Scientist at Niantic, explained in a 2022 interview that data uploaded by users during their gaming sessions is used to build high-fidelity 3D maps of the world, complete with geometric and semantic information. In other words, AI not only “sees” the shape of objects, but also understands their nature (trees, buildings, roads, etc.).
The news inevitably raises questions about privacy and the use of user data. No one, when Pokémon Go launched in 2016, would have imagined that their personal information would contribute to the development of an artificial intelligence product. As Niantic highlights the potential of LGM for robotics and human-machine interaction, it is crucial that the company is transparent about how user data is used and ensures user privacy.