Gotta map ‘em all: AI company uses Pokemon GO player data to chart the world

Millions of players who spent years scanning landmarks while chasing digital creatures in Pokemon Go have unknowingly helped build one of the largest real-world visual datasets in history, which is now being used to fuel an automated delivery service.

Released in 2016 by US-based Niantic, Pokemon Go was one of the first widely successful augmented reality games. It used players’ phone cameras and GPS to overlay digital creatures onto real-world locations, sending millions into parks, streets, and landmarks in search of the characters. 

Within weeks of launch, the game became a global phenomenon and was downloaded over 500 million times, amassing 232 million monthly players at its peak. 

Last month, Niantic Spatial, an AI-focused mapping company spun off from the game’s developer, announced a partnership with Coco Robotics to power its autonomous delivery fleets using over 30 billion images captured by Pokemon Go players over nearly a decade, creating a centimeter-precise map of urban environments.

“It turns out that getting Pikachu to realistically run around and getting Coco’s robot to safely and accurately move through the world is actually the same problem,” Niantic Spatial CEO John Hanke told MIT Technology Review last week, stating the company has over a million locations around the world “where we can locate you precisely.”