Niantic recently dropped a fascinating post on their blog that’s caught the eyes of Pokémon Go enthusiasts and tech buffs alike. The reason? They’re diving deep into how they’re using your interactions with their apps to train a sophisticated machine learning-driven geospatial model. Yep, you read that right—you might be helping shape this tech just by playing.
The blog post, titled “Building a Large Geospatial Model to Achieve Spatial Intelligence,” walks readers through how Niantic’s Visual Positioning System (VPS) amasses vast amounts of data and puts their geospatial model through extensive training, as reported by 404 Media.
Let’s take a closer look at what they’re talking about. “With our VPS, over 50 million neural networks have been trained, incorporating more than 150 trillion parameters, and it now functions in over a million locations worldwide. Our aim with a Large Geospatial Model (LGM) is to have each of these local networks enrich a global model, fostering a shared understanding of geographical spots and envisioning yet-to-be-scanned areas.”
Now, I bet you’ve got some burning questions. For starters, what’s the whole purpose of this geospatial model? Put simply, it helps tech devices like AR glasses or robots get a grip on real-world surroundings, enabling them to navigate more intuitively. Imagine these devices crossing the road without misidentifying a lamppost as an obstacle to avoid. That’s the power of spatial intelligence introduced by Niantic’s efforts.
And what about this Visual Positioning System? Pokémon Go fans might recognize it as the game’s AR mapping component. Players often engage with it during field research tasks to earn rewards by scanning real-world locations. As detailed in another Niantic blog post, VPS fine-tunes the way AR elements are mapped onto our real environment. So, it’s kind of like how the game accurately matches a PokéStop to your local museum and not, say, the loos at the neighborhood pub.
With data from 10 million scanned locations courtesy of gamers across Niantic’s various platforms, the company has mostly seen mapping activity concentrated in places like Japan, coastal areas of the US, and Central Europe. Sure, some folks might roll their eyes at AI and machine learning tech, but in the context of actual AR innovations, this shows meaningful strides. Just ensure you haven’t inadvertently recreated scenes from The Shining by letting Pokémon Go map all corners of your home for Niantic.
So, what’s your take on all this? Personally, I’ve got reservations about AI, especially considering the environmental impacts unless we tackle the ongoing energy woes. But hey, maybe you see it differently! Share your thoughts below.
