The AI Starfield

The AI Starfield

If Adder has one consistent mantra, it would likely be “we’ve been busy.” No, it’s not because we think we’re so popular. It’s because we have a great team with a lot of ambitions. Starry-eyed, you might call us.

If you’ve been keeping an eye on adder.io, you’ve no doubt noticed that we’ve recently taken a keen interest in the relationship between movement, space, and time. We focus primarily on our driver fleet, gig economy service, and car wrap advertisements, but there’s a massive backend that we’ve been building to make all this work.

Analytics for the real world is how we describe what the system provides.

In case you haven’t seen the video on our analytics page, GPS data points are laid out to show our outdoor advertising analytics capabilities — but before New York, Chicago, and Los Angeles become apparent, one could easily mistake the animation for an image of a star field.

Brandon Bush and I put this animation together at the suggestion of our designer, Mark Jackson. We felt that it was a beautiful way to represent our data and what we’re teaching our systems to process.

Cue the Interstellar theme.

Many people I’ve talked to don’t get the images right away. I’ve asked for guesses to tell me what the image could be. Most don’t really understand what we’re doing in the first few seconds, but if you take a moment, you’ll see patterns that begin to emerge.

It’s a representation of time and space and humanity.

This star field is anonymous, PII-less location sensor data. It’s not the only one of these animations we’ve created. In fact, we’ve mapped the entire continental United States for mobile device locations. With it, we can measure variables and KPI’s, we can gain insights — we can build maps.

25 billion data points, give or take, are indexed by our system every day. It’s quite a challenge to sort through all of them, let alone to understand all the insights and data that lies just under the surface. It’s a bit much for us mortal developers to do every day.

Which is why we’ve began training our Spatial Deep Learning System, Aldo.

A rendering of GPS data collected in Chicago, Illinois, in March of 2019
Shout out to our friends at Allstate and Arity. Oh, and any other Chicagoans out there reading this. How’s it going?

No, not yet. Not quite.

In a few days I’ll be sharing Adder’s story in developing these AI models, and eventually — I might even tell you what our big plan is all about. All I want to share now is that we’re working on spatial data projects that far surpass the field of out of home advertising, but will be very useful to OOH marketers as well.

In the meantime, our amazing development team will be showing some of the process that have led us to our latest advancements in spatial deep learning. There’s some really awesome stuff we’re doing with our neural networks and are very excited to share the progress when the time is right. It’s analytics for the real world(s).

For now, I’ve gotta run to go talk with George and Gracie. I’ll introduce them in a few days. But you don’t have to stop learning about our… learning! Head on over to the analytics page to read more!

-Ian Gerard

This is how you train a spatial AI.


Your email address will not be published. Required fields are marked *

fourteen − 13 =