AI In Smart Glasses Is Missing the Point

I’ve used the Ray-Ban Meta Gen 1 and Gen 2 a lot over the last couple of years. One of my favorite things is the open-ear audio, which is a game-changer for hands-free calls and listening to music while you’re biking. And you know what? The cameras are okay, too, if you’re not looking for anything super hi-res. It’s good stuff! It’s actually the opposite stuff of one of the supposedly headline features being shoehorned into Meta’s description of the device—AI.

Unfortunately, the titular AI in Meta’s self-described AI glasses is still the worst part about using them. As useful as I find the ability to call and message hands-free (or on display glasses, navigation, and notifications), none of that ties into the one thing that Meta seems to think is the defining feature of these devices. Sure, voice assistants use AI, and those can be incredibly useful (if not critical) on a pair of smart glasses—especially those without a screen—but there are still a lot of misses there, too.

Voice assistants, in case you haven’t checked up on them lately, are about the same in terms of usefulness as they were 10 years ago, which is to say, moderately helpful. Google Assistant, Alexa, and Siri have all struggled mightily to improve over the years, and while two out of those three currently have next-gen voice assistants at least semi-available to the public, the word is out on whether they’re actually more helpful beyond turning our lights on and off or launching music. Meta AI, the voice assistant inside Meta’s Ray-Ban smart glasses, isn’t immune to those struggles either. It gets stuff right sometimes, and other times… not so much.

And the worst part is, the struggles of Meta AI in the voice assistant department pale in comparison to the computer vision side of things. If you’re not familiar with smart glasses, let me back up for a moment; in the Ray-Ban Meta Gen 1 and 2, you can use your smart glasses’ camera to ask Meta AI about your surroundings. That can be useful for stuff like translation or for low-vision people who need help reading things… if it works correctly. For everyone else and for most other scenarios, the feature can feel a bit obvious.

When I’m testing Meta AI and its computer vision abilities, I often struggle to come up with things to throw at it. Sometimes I’ll look at an object and ask, “Hey Meta… what’s that?…” Oh, it’s a scooter, thank god! You saved me, Meta. Arguably worse is when you do actually think to use Meta AI and it doesn’t get the job done, like the time it told me every shell I picked up at the beach was a shark’s tooth.

I’m singling Meta out here (in my defense, they put AI in the name of the smart glasses), but they’re not alone in their emphasis on AI. Prototypes of future smart glasses shown off by Google also lean heavily into computer vision, and even suggest a world where the camera in them is on literally all the time, watching everything you do. That sounds like the worst possible smart glasses future from a privacy standpoint. And based on Magic Leap’s recent presentation, it looks like other companies in the smart glasses game aren’t trailing far behind.

I’m not writing all of this just to rag on AI. I think smart glasses can be useful, fun, and maybe even augmentative, but it’s going to take attention to detail and focus in the right direction. And sure, maybe AI can be a part of that, but there has not been evidence yet to suggest that it should be the main part, even if Meta’s naming conventions would suggest otherwise.