Niantic, the developer behind the creature-catching game Pokemon GO, announced that it is building an AI map model using player data.
Its official website says it’s “pioneering the concept of a Large Geospatial Model that will use large-scale machine learning to understand a scene and connect it to millions of other scenes globally.”
READ: Google Maps now helps users beat traffic jams
This technology can make smaller, local mapping data more accurate.
As a result, navigation in these areas can become more reliable, and LGMs can become a new frontier for AI research.
‘Gotta map ‘em all’: Pokemon GO data for AI maps
It’s easy to imagine what the back of a familiar building may look like, even if you haven’t seen it from all sides.
That’s because humans have “spatial understanding,” meaning, we can fill in these details based on similar ones we’ve encountered.
On the other hand, even the most advanced AI models struggle to infer missing parts of a scene.
Fortunately, Niantic might change that with Spatial Intelligence.
The company has been training its Visual Positioning System (VPS) with over 50 million neural networks and 150 trillion parameters.
It uses a single image from a phone to determine its position and orientation using a 3D map.
Player data helped build that map by scanning interesting locations while playing Pokemon GO and other Niantic titles.
The company aims for these local networks to power a Large Geospatial Model, providing a shared understanding of geographic locations.
More importantly, it will map places that have not been scanned fully.
Anton Dahbura, the executive director of the Information Security Institute at Johns Hopkins University explained the technology further to USA Today.
He stated Niantic’s data is unique because it’s from a “pedestrian perspective.”
In contrast, most mapping systems rely on cars to capture images so their data may not show places inaccessible to vehicles.
Many Pokemon GO player did not expect Niantic to use their data for AI map training.
Consequently, several persons shared online complaints.
Some company privacy policies only explain the data they share with third-party vendors but not what the company does with it.
That is why Dahbura suggests players think carefully before sharing data with companies.
He also emphasizes the importance of location data.
Dahbura warns, “It can be used not only against you but also against national security.”