Millions Of Pokémon Go Players Unknowingly Trained Niantic’s AI That Helps Robots Navigate

Remember Pokémon Go? The mobile game that got everyone on the streets hunting for those virtual Pokémons? If yes and you were someone who played it as well, humming "Gotta catch 'em all", then you might have unwillingly trained an AI system, reports suggest. The game by Niantic might have used that data to train its AI systems and robots to equip them with the knowledge of navigating through the city; basically, it was reportedly used to build a highly accurate navigation system, which is now under use by delivery robots in many cities.

ALSO READ: MacBook Neo Touchscreen Support Unlikely As Apple May Reserve It For MacBook Pro: Report Suggests

According to reports, in order to make the game's augmented reality work seamlessly, the company utilised its Virtual Positioning System (VPS). Only recently Niantic's (developers behind Pokémon Go) AI division, Niantic Spatial, entered a partnership with a firm specialising in robotics, called Coco, which will employ the use of this technology for delivery robots.

This project that is built using Niantic's VPS was trained utilising around 30 billion images that the players captured while scanning for PokeStops, Gyms and other game-related landmarks that occurred. Over the course of time a large database was created with the advantage of having real-world locations with accurate geographical positioning and data entry. The system has now become capable enough to accurately identify any location with high precision. It does it by comparing the camera images and its stored database of images. This gives the machines an upper hand in navigating even in parts where traditional navigation tools fail.

The company says this technology helps fix a problem robotics companies have faced for years. Most delivery robots depend on GPS, but in dense cities GPS signals can become unreliable because tall buildings often block or disrupt them.

Instead, Niantic uses a visual positioning system (VPS) that relies on visual cues. Robots can recognise surroundings like buildings, street signs and landmarks to figure out their exact location. This allows them to move through sidewalks, intersections and pickup spots much more accurately than relying on GPS alone, making last-mile deliveries in busy cities more practical.

Niantic Spatial's CEO believes that the gap between augmented reality and robotics is getting narrower and the overlap between the two technologies is closer than many might think. The systems that make any virtual character navigate through roads and other landmarks realistically in the real world actually employ the same spatial awareness system that a robot might utilise for its navigation through streets, sidewalks and even parks.

John Hanke, CEO of Niantic Spatial, expressed in a statement to MIT Technology Review, “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."

The company tells us that the database does not only have images but also metadata like GPS coordinates, camera angles, time of day, and weather conditions as well as motion data. This is due to the fact that players had captured the same place from many different angles in various conditions; this helped the system to learn about a specific place under different circumstances. This makes the robots more capable and the model more reliable when it comes to navigating through various terrains.

ALSO READ: WhatsApp Update: New 'Guest Chat' Feature In Beta Lets Users Chat Without A WhatsApp Account

The company might not stop just here, as it further has plans to build what it calls a "living map", which will be a digital representation of the world that gets constantly updated and has a use case for both machines and humans. The company has expressed that as these delivery robots carry on with their tasks, the devices will collect new data while operating in order to help the system improve by creating a model that showcases even better details of its surroundings, like buildings, streets, etc.