Niantic Utilizes Player Data from Pokemon Go to Develop AI for Navigating Robots

A recent blog post from Niantic has stirred quite a buzz among Pokemon Go fans and tech enthusiasts. It delves into the company’s impressive efforts in developing a machine learning-powered geospatial model, and you might be surprised to learn that you’ve likely contributed to its development through their apps.

The blog, titled “Building a Large Geospatial Model to Achieve Spatial Intelligence,” reveals how Niantic has been using its Visual Positioning System to gather extensive data. Thanks to this system, they’ve been rigorously training their geospatial model, and hats off to 404 Media for highlighting this intriguing advancement.

According to the blog, “As part of Niantic’s Visual Positioning System (VPS), we have trained more than 50 million neural networks, with over 150 trillion parameters.” This technology is now operational in over a million locations. Their vision is to create a Large Geospatial Model where these local networks collaboratively contribute to a global model, enhancing our understanding of geographic spaces, including those not yet fully mapped.

Now, you might be wondering what exactly this geospatial model does. Simply put, it helps computers perceive and navigate real-world settings. The practical implications, as explained in Niantic’s post, are significant for tech. Devices such as AR glasses, robots, and autonomous systems could seamlessly navigate environments, like effortlessly recognizing a lamppost instead of crashing into it.

Another question arises: What is this Visual Positioning System? For Pokemon Go players, it’s recognizable through the game’s AR mapping tool, which appears in field research tasks. Players are prompted to scan real-world areas for in-game rewards. In a separate blog, Niantic describes the VPS as technology that aligns AR content more accurately with actual locations—think of mapping a Pokestop to a museum rather than a restroom.

Niantic reports that users have scanned over 10 million locations across their various games, with a heat map showing data concentration in Japan, Coastal US, and Central Europe. While AI and machine learning models have often been criticized for their frivolous applications, this development hints at genuine advancements in AR technology. Of course, it could be unsettling if you’ve unknowingly been mapping out your home for Niantic!

So, what’s your stance on this? Personally, while I can appreciate the tech progress, I have reservations about AI on environmental grounds until we find sustainable energy solutions. But you might have a different perspective! Feel free to share your thoughts below.