Rabbit R1 is an adorable AI-powered assistant co-designed by Teenage Engineering
This dedicated virtual assistant device fits in your pocket, but it's not a phone.
Yes, you probably already have a virtual assistant in your pocket on your phone. Heck, if you're reading Engadget, I'm willing to bet you've got at least one smart speaker floating around your home as well that you can ask to complete basic tasks. But a new startup called Rabbit seems to think these are less than ideal implementations of AI (if you can really call Siri and Alexa that). The company (which was on the ground at CES 2024 in Las Vegas but not an official exhibitor) envisions a world where you trade apps for conversation and, rather than a distracting device shoving icons in your face, you interact with what amounts to a walkie-talkie for an AI.
The Rabbit R1 is the first device to be launched by Rabbit and it's an objectively adorable little square in an endearingly bright shade of orange. Even if you're not sold on the necessity of a dedicated gadget for a virtual assistant, it's hard to deny the aesthetic appeal, which comes courtesy of the design gurus at Teenage Engineering. It features a small 2.88-inch touchscreen, an analog scroll wheel, two mics, a speaker and a "360 degree rotational eye," which is just a fancy name for a camera you can spin to face toward you or through the back of the handset.
The primary way you interact with the Rabbit AI, though, is by pressing and holding the "Push-to-Talk" button. This tells Rabbit OS to start listening. A heavily stylized and disembodied rabbit head bobs slowly as you ask your question or give it a task, and then it quickly gets to work. Want to book an Uber? Need a recipe to use up the leftovers in your fridge? Wondering who sampled The Isley Brothers "That Lady"? (The answer is Beastie Boys, Basement Jaxx and Kendrick Lamar, FTR.) The Rabbit R1 AI seems pretty capable of handling those tasks, at least in the controlled video demo.
Rabbit OS is able to tackle those tasks using what it calls the Large Action Model (LAM). This is what founder and CEO Jesse Lyu pitches as the company's major innovation. It's designed to take actions on interfaces rather than through APIs or apps. In short, it can be trained to carry out almost any task that can be accomplished through a user interface. It's sort of like a fancy version of a macro.
As a way of demonstrating its capabilities, Lyu teaches the Rabbit AI how to generate an image using Midjourney via Discord. Once Lyu walks and performs the process, with Rabbit OS recording his actions, it can repeat the task when asked.
The rotating camera faces up into the body by default, acting as a sort of privacy shutter. Only turning its sensor towards its target when summoned. It can do the usual tricks like identify people or thing in the real world (within reason, at least). But the way it interacts with the AI is sure to pique people's interest. In the demo Lyu points the R1 at a full refrigerator and asks it to suggest a recipe that's "low in calories" based on its contents.
Of course, there are still a lot of unanswered questions about the Rabbit R1. How is the battery life? The company claims it's "all day," but what does that really mean? And will the average user be able to train it easily. At least we know a few things, though. We know it costs $199 and is available for preorder now, with an expected ship date sometime in March or April.
We're reporting live from CES 2024 in Las Vegas from January 6-12. Keep up with all the latest news from the show here.
