Pokemon Go developer Niantic is using player data to train AI that could power map-roaming robots
A blog post on the Niantic website has gained the attention of numerous Pokemon Go fans and tech-savvy folks alike due to the company’s lengthy description of its work training up a machine learning powered geospacial model through its apps. That means you, yes you, might have played a part in building it up.
The blog post in question, titled “Building a Large Geospatial Model to Achieve Spatial Intelligence” states that through Niantic’s Visual Positioning System, it has gathered a lot of data and put its geospacial model through a lot of training (thanks, 404 Media).
“As part of Niantic’s Visual Positioning System (VPS), we have trained more than 50 million neural networks, with more than 150 trillion parameters, enabling operation in over a million locations. In our vision for a Large Geospatial Model (LGM), each of these local networks would contribute to a global large model, implementing a shared understanding of geographic locations, and comprehending places yet to be fully scanned.”
Okay, so there are probably a few big questions here. First, what’s this geospacial model actually do? Well, in simple terms, it’ll help computers percieve and navigate real-world environments. What this actually means for tech, as laid out in the aforementioned blog post, is that stuff like AR glasses, robots, and autonomous systems will be better able to walk across the street without struggling to identify the lamppost they’re walking towards.
New: the company behind Pokemon Go has announced it is using data collected by its millions of players to build an AI model that can navigate the real world, and could be used for robots. Doubt Pokemon Go players anticipated contributing data to this project www.404media.co/pokemon-go-p…
— Joseph Cox (@josephcox.bsky.social) 2024-11-19T15:36:06.741Z
Second question, what’s this Visual Positioning System? Well, Pokemon Go players might know it through the lens of the game’s AR mapping tool. These have ben part of field research tasks for a bit, and have the player scan parts of the world for in-game rewards. The VPS, as explained in another Niantic blog post, is a bit of tech that better maps AR content to the real world. So, again in simple terms, the way the game maps a poke stop to a local museum, rather than the pub toilets.
Niantic states it’s now got over 10 million locations scanned by its users across its many games, even providing a heat map to show the majority of this data has been gathered in Japan, the Coastal US, and Central Europe. AI and machine learning models have gotten a well-deserved bad wrap for their use in bunch of useless shite. But, if you’re on board with AI development, this does point to real technological improvements as far as AR goes. One just hopes you’ve not been using Pokemon Go to scan your entire house for Niantic, lest a real life Metacross breaks down your bathroom door like in The Shining.
What do you think about all this? Personally speaking, I’m largely against AI tech on an environmental level, at least until we’ve got a good solution to the energy crisis, but maybe you feel differently! Let us know below.