Niantic is constructing a ‘geospatial’ AI mannequin primarily based on Pokémon Go participant knowledge

Niantic has introduced that it’s constructing a brand new “Massive Geospatial Mannequin” (LGM) that mixes hundreds of thousands of scans taken from the smartphones of gamers of Pokémon Go and different Niantic merchandise. This AI mannequin might permit computer systems and robots to grasp and work together with the world in new methods, the corporate mentioned in a weblog put up noticed by Rubbish Day.

The LGM’s “spatial intelligence” is constructed on the neural networks developed as a part of Niantic’s Visible Positioning System. The weblog put up explains that “Over the previous 5 years, Niantic has centered on constructing our Visible Positioning System (VPS), which makes use of a single picture from a telephone to find out its place and orientation utilizing a 3D map constructed from folks scanning fascinating areas in our video games and Scaniverse,” and “This knowledge is exclusive as a result of it’s taken from a pedestrian perspective and consists of locations inaccessible to automobiles.”

Niantic Chief Scientist Victor Prisacariu was extra express in a 2022 Q&A, saying, “Utilizing the info our customers add when enjoying video games like Ingress and Pokémon Go, we constructed high-fidelity 3D maps of the world, which embrace each 3D geometry (or the form of issues) and semantic understanding (what stuff within the map is, comparable to the bottom, sky, bushes, and so forth).”

As 404 Media factors out, no one who downloaded Pokémon Go in 2016 might have predicted their knowledge would “at some point gasoline the sort of AI product.”