CES 2024 is hardly lacking for AI gadgets, but the most intriguing take on the concept to date is conspicuously absent from the show floor.
Developed by a startup called Rabbit, the Rabbit R1 is a handheld AI companion that's built around a charmingly colourful design developed in collaboration with Teenage Engineering. At a glance, the pitch here is familiar to recent LLM-based AI gadgets like the Humane AI Pin and older voice assistants like Amazon Alexa.
You can ask the R1 questions and it will read your answers via a speaker on the back. It's also got a rotating camera and it can use its microphone, so you could potentially ask it more contextual queries like the name of a nearby landmark or a potential meal you could make with whatever is left in your fridge on a Friday night.
So far, so familiar. But where it gets interesting is that the Rabbit R1 relies on what the company behind it are pitching as a "large action model" rather than a language-based one. The idea here is that rather than have to persuade countless developers and services to adopt yet another virtual assistant, the startup behind Rabbit has built infrastructure that allows it to teach its AI companion to operate the existing apps and web pages of a given service instead.
During this process, the 2.88-inch touchscreen on the Rabbit R1 will display information about a given query and ask for confirmation when it comes to making significant decisions like whether or not to spend money on a digital transaction.
Then, where it gets really interesting is that the R1 also comes with a dedicated training mode that allows you to teach it how to do new tasks and then take that new skill forward. Some examples that the company is throwing around already are as material as removing a watermark in Adobe Photoshop or as minute as interacting with a specific bot on a specific Discord server.
Writ large, Rabbit's approach to an AI assistant is a deceptively simple one with a huge canvas of potential use cases. That said, it's also kinda complicated to wrap your head around. I had to watch the video below twice before I could grok it.