Snap’s Sleeping Dragon

Checking out the nuts and bolts of Snap’s AI integration.

Photo via the U.S. Patent and Trademark Office.

Sign up to uncover the latest in emerging technology.

Snap wants to keep up with the AI heavyweights by giving itself some truth.

First up, Snap wants to patent tech for “generating ground truths,” or information that is known to be true, for machine learning models. Here’s how Snap’s tech works: First, a 3D model is developed by a “messaging system” (a.k.a., Snapchat), and the other is generated by a machine learning model based on images of the first one. The two models then are compared to see if the machine learning-generated one is accurate to the original. If changes to the image inputs are given to the AI model, rather than creating a new 3D model entirely, those changes are reflected in the AI-generated one. 

While this sounds esoteric, this tech essentially makes AR model generation a much smoother process with better results. The ability to generate a ground truth “enables application developers to develop robust image processing applications without spending a prohibitive amount of time developing the ground truth,” Snap noted in the filing. 

Snap also filed a patent for integrating machine learning into “augmented reality content items.” This essentially uses machine learning to generate customized “image augmentations,” using AR objects, like filters. 

However, what’s interesting to consider about this patent is that nothing described in it is novel or new, Glimpse’s Jake Maymar noted. While this filing is from September of last year, the technologies that are outlined in it have not only been worked on and integrated by Snap for years now but are fundamental in the development of AR. 

Snap hasn’t been entirely quiet on all the AI buzz either. Just last week, the company announced at its Snap Partner Summit that its AI chatbot would be free for all global users. Soon, the chatbot will have the ability to respond with “generative” visual Snapchats back, rather than just messages. However, much of Snap’s AI work, including the tech in these patents, plays into its larger bets on AR. 

The company has staked much of its future on AR-related tech, from product development within its Lens studio to its persistent work on wearables like its AR Spectacles

Maymar told me he sees Snap as a “sleeping dragon” in the AR and AI sectors: While the company isn’t always seen as a key player in either space the same way that Big Tech firms like Google or Meta are, the company has spent more than a decade building a loyal user base and now has more than 250 million that interact with its AR content on its platform regularly.

With its success in the consumer space, Snap branched into enterprise offerings with a new AR Enterprise Services division, which debuted in late March, to help businesses build their own AR tools. 

“It’s, in my opinion, the successful version of consumer AR,” Maymar said. “That’s why I say it’s a sleeping dragon. Even though it feels like a toy, it’s being used a lot, it’s generating revenue, and it doesn’t seem like it’d be too hard to flip that switch and make it enterprise.” 

Have any comments, tips or suggestions? Drop us a line! Email at admin@patentdrop.xyz or shoot us a DM on Twitter @patentdrop. If you want to get Patent Drop in your inbox, click here to subscribe.