Think back to summer 2016: Were you one of the 200m+ people wandering around with your phone looking for a Pikachu. If so, your “catch ‘em all” fever contributed to Pokemon Go creator Niantic’s location data.
Per Niantic:
- The company has 10m scanned locations worldwide.
- Over 1m are available for use in its Lightship Visual Positioning System (VPS), which lets AR developers anchor virtual content to real-world locations with “centimeter-level accuracy.”
- Each week, it receives ~1m more from its games and apps.
Now, Niantic is using this data to produce a “large geospatial model,” per 404 Media.
What’s that mean?
ChatGPT is a large language model (LLM), which means it uses language to produce language. It trains on text scraped from the internet, then produces text, like when you ask it for a five-paragraph essay on why you should get a cat.
Niantic’s idea for a large geospatial model is to use all this location data to understand and navigate the physical world.
As Niantic explained:
- Humans can usually imagine what a whole building looks like even if we’ve only seen one side because we have spatial intelligence.
- Bots don’t. But a model could understand what the rest of the building probably looks like based on common characteristics of similar buildings and places, even if that side hadn’t been scanned.
I guess that’s cool, but what’s the point?
What’s interesting about Niantic’s data is that it comes via pedestrians, who can go into parks and down narrow alleys and other places cars cannot.
And though Niantic didn’t specify, there are myriad potential applications in the kinds of AR games and apps Niantic makes. It’d also be useful for developers of AR wearables, like the Ray-Ban Meta smart glasses.
It would also benefit companies that rely on robot navigation — self-driving taxis, delivery bots or drones, etc.
Naturally, some people noted this could pertain to governments interested in robot-based warfare. So when you’re hiding underground from Skynet, just remember it was all because you had to have that Snorlax.
