Niantic is building a “geospatial” artificial intelligence model based on data from Pokémon Go players

Niantic has it announced that it is building a recent “large geospatial model” (LGM) that combines millions of scans taken from players’ smartphones Pokémon Go and other Niantic products. The company said in a blog post that this model of artificial intelligence could allow computers and robots to understand and interact with the world in recent ways noticed by Garbage day.

LGM’s “spatial intelligence” relies on neural networks developed as part of Niantic’s visual positioning system. The blog post explains that “Over the past five years, Niantic has focused on building our Visual Positioning System (VPS), which uses a single image from your phone to determine your position and orientation using a 3D map built from people scanning interesting locations in our games and Scaniverse” and “This data is unique because it comes from a pedestrian’s perspective and covers places inaccessible to cars.”

Niantic’s chief scientist, Victor Prisacariu, was more blunt: Questions and answers for 2022saying: “Using data submitted by our users when playing games such as Ingress AND Pokémon Gowe have built high-quality 3D maps of the world that cover both 3D geometry (or the shape of things) and an understanding of semantics (what is on the map, e.g. earth, sky, trees, etc.).

How 404 Media indicates that no one has downloaded Pokémon Go in 2016, they could predict that their data would “one day fuel these types of AI products.”