AI Enters the Real World

Vision models are giving machines sight and the ability to judge your form, find your LEGO bricks, and more.

ChatGPT brought artificial intelligence to the collective consciousness with the chat interface for interacting with LLMs that became the de facto lightbulb moment for many people discovering the opportunities of AI tools. But we've also been interacting with AI tools in different ways since well before we were oversharing with a server farm (Clippy died so ChatGPT could soar).

ChatGPT is throwing resources into every interface for integrating AI, with Sora (social app), Atlas (web browser), and hardware (Jony Ive collab) alongside their core chat and API features. Meta has been throwing money at gaining a foothold in the hardware game with their VR headsets for years, trying to break free of their dependence on Apple. Apple, meanwhile, has seemingly fumbled their hardware advantage in terms of rolling out AI features. Though I feel that's more narrative than fact.

One of the most powerful features of AI is the ability to get things done without being tethered to typing on a physical device. Meta's Ray-Bans seem like they're gaining a foothold here, but time will tell if they are a slightly less nerdy Google Glass. Not to mention the privacy concerns around people potentially always recording with their glasses. But you know what people are already pretty used to seeing people wear? AirPods.

There's a decent chance someone in your general vicinity is rocking a pair right now. With Apple rolling out AI translation features into AirPods recently, they've already established it as a device capable of hardware-accelerated AI features in an already ubiquitous device that can listen and pair with a less privacy-intrusive camera for visual models. I guess that is to say, don't count out Apple.

🎥 Watch
Wait... Smart Glasses are Suddenly Good?
Marques Brownlee
I understand Meta Ray-Bans are selling well. I mean...I'd try them on. But of course as you can see (points at newsletter) I'm a bit of a tech nerd. The latest version of Meta's Ray-Bans adds a digital display, which does seem like the logical next step.

I see a couple hurdles to adoption. While they're not necessarily bad looking glasses, they're pretty obviously thicker. Then we have to consider how people are going to feel with someone walking around with glasses potentially always recording or feel about being the person that is walking around with glasses that are potentially always recording.

And of course its ads all the way down at Meta.

📰 Article
First impressions of Alexa+, Amazon’s upgraded, AI-powered digital assistant
TechCrunch
Not to be outdone, Amazon released an upgrade to Alexa to integrate AI features into the smart speaker you have that was mainly for setting timers, checking the weather, and Jeff Bezos listening in on your dinner conversations. We'll see if Alexa+ (not a streaming service) entices people to bring Amazon's devices out of storage.

📰 Article
Peloton Is Having Its Biggest Hardware Launch in Years
Wired
Speaking of expensive tech that tends to be repurposed for things like storage and hanging clothes, Peloton is getting into the AI game with its latest release. Fitness feels like a good space for AI, and I like what Peloton is aiming to do with vision models to help coach users on their form and exercise. Pair it with other health information from wearable devices and there's a lot of potential for unlocking health benefits with AI in the fitness space.

📰 Article
Google’s Gemini-powered smart home revamp is here with a new app and cameras
Ars Technica
Listen, we're either going to unlock some cool use cases or we're just giving the eventual robot overlords more data into the inner workings of our homes and lives at this point. I'm hoping for the former with Google's updates to their home app and devices. Particularly interesting is the combination of vision models and LLMs to let users do natural language search for events in their camera history. As someone who's spent way too much time scrubbing through camera footage to see the last whereabouts of a favorite stuffed animal, it would be nice to let the machines point me in the right direction.

👀 Nifty
Sort A Brick
This turned into more of a vision model issue, so why not end it with the coolest use case I've come across recently? Sort A Brick, a Western European company, is using vision models to help identify, sort, and repackage your LEGO bricks. You send in your assortment of bricks, and they sort, clean, see what sets you can build, and send them back. Hopefully they come to the US so I can use that to convince myself and my wife that the LEGO Death Star is an investment opportunity.

Thanks for reading,
Jason