If all you thought you were doing when playing Pokemon Go was enjoying a fun AR mobile title, think again. You were also helping developer Niantic feed an AI model to develop a 3-D world map to help machines navigate more freely. And it’s been happening for years.
As noted by 404 Media, the developer recently went full mask-off supervillain in a blog post touting its “large geospatial model” technology, which it claims will “enable computers not only to perceive and understand physical spaces, but also to interact with them in new ways, forming a critical component of AR glasses and fields beyond, including robotics, content creation and autonomous systems.”
The data fed into this AI model comes from the visual positioning system (VPS) that games like POGOÂ uses for its Pokemon Playgrounds feature that lets players toss Pokemon into real-world “landmark” locations. The company brags that it pulled “one million fresh scans each week” with “hundreds of discrete images” that have been harvested to build up its AI model to generate “a highly detailed understanding of the world.”
Application of these data is part of Niantic’s larger goal of creating spatial intelligence, which it claims can be used to help with AR gaming, logistics, and spatial planning and design – all of which are features that the company just so happens to offer with its newly announced spatial platform business. “Spatial intelligence is the next frontier of AI models,” the company’s blog heralds.
While this information appears to have caught many by surprise, the intention of Niantic’s use of data to create an AR map isn’t exactly news, as executives first outlined some of the studio’s plans in 2018 and its own privacy policy notes that such data is “[used] to build a 3D understanding of real-world places, with the goal of offering new types of AR experiences to [Niantic’s] users” – points that some apologists argue is reason enough to lay blame on players’ inattention, while many others are saying it’s the cost of business or not surprising. Some are even celebrating the tech.
Niantic further added an editor’s note at the top of the LGM blog post that tries to push back against wider assumptions of this collection and insist that it remains optional:
“This scanning feature is completely optional – people have to visit a specific publicly-accessible location and click to scan. This allows Niantic to deliver new types of AR experiences for people to enjoy. Merely walking around playing our games does not train an AI model.”
There are certainly people who are shrugging their shoulders or claiming that this data harvesting for Niantic to build an entirely all-new business is “old news” – which is partly fair, as our own Massively on the Go columnist Andrew has been shouting it into the wind for years ever since post-COVID design rollbacks made clear that Niantic’s data harvesting business was its core business all along. In fact, here’s a piece our own writing team did on the subject back in 2019. But there are more nefarious uses of this information that we should all be aware of.
“A lot of people really underestimate the importance of location data,” says Anton Dahbura, executive director of the Information Security Institute at Johns Hopkins University. “Our critical infrastructure is much broader than people realize, including transportation systems, pharmaceutical, financial, food manufacturing and so on. If people with bad intentions figure out that you have access to these kinds of facilities, it can be used not only against you but also against national security.”
