OpenGlass: Built Over 24 Hours At Meta Hackathon, These Open-Source AI Smart Glasses Might Give Ray-Ban Meta A Run For Its Money

Facebook owner Meta hosted a hackathon earlier this month to encourage developers to build open source tooling projects using the full collection of Llama models, including Meta Llama 3 and Meta Llama Guard 2.

As many as 51 projects were built in 24 hours. While there were many innovative projects on show, the show-stealer — and hackathon winner — was the team that built budget smart glasses. The wearable artificial intelligence (AI) gadget, built with the Meta Llama 3 open-source language model, can provide information about what the wearer is seeing.

So, the wearer can ask questions about anything they saw during the day or even whatever they are currently looking at.

For instance, in a live demonstration at the hackathon by Nik Shevchenko, the leader of the five-member team which built OpenGlass AI, the glasses were able to quickly tell, upon being asked, how many calories were packed into three different fruits — strawberries, blueberries, and raspberries — present in its view.

The promise of these AI-powered smart glasses is that its wearer will be able to record their life, remember people they meet, identify objects, translate text, and a lot more. And it apparently costs only $20 — approximately Rs 1,660 — to build the small hardware attachment that turns any ordinary pair of glasses into this powerful wearable AI.

In its subsequent versions, OpenGlass is expected to see a better form factor, the addition of audio recording using the existing microphone on the glasses, and the inclusion of subtitles when dealing with audio, among other things. The team plans to work closely with the open source community to further improve OpenGlass.

Interestingly, though it’s still rough around the edges, OpenGlass is 15 times cheaper than the Ray-Ban Meta smart glasses — and Meta hosted the hackathon where OpenGlass was built.

“We believe there is a massive gap between AI and the physical world around us; we wanted to bridge this gap by building OpenGlass,” co-creator Jatin Gupta wrote on Devpost, a platform which hosts hackathons with the mission to inspire developers to build great software.

Gupta reveals that the major challenge they overcame was the integration of the physical attachment piece onto the frame such that it works for any pair of glasses. “We've succeeded in making the attachment fit onto any pair of glasses in the most aesthetic manner possible,” Gupta wrote.

At the Mistral AI hackathon held in March this year, Shevchenko and team built over a 24-hour period an open-source wearable AI called Friend. Worn around the neck like a necklace, Friend can record the wearer’s conversations, provide AI-powered insights, and chat with the wearer via voice or text.

The device works longer than 24 hours on a single charge and, again, costs roughly $20 (approximately Rs 1,660). Two mobile applications serve as companions, one for the device and one standalone.

Most importantly, it is built as an open-source project — something Shevchenko is extremely passionate about. The open-source community helped Friend improve, leading to the launch of two subsequent iterations. The latest version of Friend is being sold at $39 (approximately Rs 3,244).

With OpenGlass, Shevchenko’s company Based Hardware is firmly marching along on the path of real open-source innovation.

Although several new AI devices have been launched over the last year or so, such as the Humane Ai Pin and the Rabbit r1, AI smart glasses stand apart as a use case and probably have the best shot at succeeding as an AI device. They are able to blend seamlessly into our living reality, as if they are an extension of our eyes, and accomplish routine tasks for us, like play music or search the web.

This ‘unseen’ quality of AI smart goggles might make them most suitable for common use in the future.

Also Read: Why You Might Want To Pick Up AI Smart Glasses