You likely have a virtual assistant on your smartphone and perhaps a smart speaker at home, especially if you’re an Engadget reader. However, a new startup called Rabbit challenges the conventional AI implementations found in Siri and Alexa. Present at CES 2024 in Las Vegas, Rabbit envisions a future where conversations replace apps. Instead of dealing with distracting devices displaying icons, Rabbit proposes interacting with their AI device, the Rabbit R1, akin to a walkie-talkie. We had the opportunity to explore the Rabbit RI AI device firsthand at the event.
The Rabbit R1, Rabbit’s inaugural device, is undeniably charming with its small, square design in a vibrant shade of orange. Even if you’re skeptical about the necessity of a dedicated gadget for a virtual assistant, the aesthetic appeal, crafted by the design experts at Teenage Engineering, is hard to ignore. It boasts a compact 2.88-inch touchscreen, an analog scroll wheel, two microphones, a speaker, and a “360-degree rotational eye,” essentially a camera that can be swiveled to face either you or the back of the device.
The primary mode of interaction with the Rabbit AI involves pressing and holding the “Push-to-Talk” button, prompting the Rabbit OS to initiate listening mode. A stylized and detached rabbit head, moving at a deliberate pace, responds as you pose questions or assign tasks, swiftly commencing the required actions. Whether it’s booking an Uber, finding a recipe for leftovers, or identifying artists sampling The Isley Brothers’ “That Lady” (Beastie Boys, Basement Jaxx, and Kendrick Lamar, for the record), the Rabbit R1 AI demonstrates its proficiency in these tasks, as demonstrated in the controlled video demo.
What sets Rabbit OS apart is its utilization of the Large Action Model (LAM), touted by founder and CEO Jesse Lyu as the company’s key innovation. The LAM is engineered to execute actions on interfaces rather than relying on APIs or apps. In essence, it can be trained to perform virtually any task achievable through a user interface, resembling an advanced version of a macro.
To showcase its capabilities, Lyu instructs the Rabbit AI on generating an image using Midjourney through Discord. After Lyu guides and performs the process, the Rabbit OS records the steps, allowing it to replicate the task upon request.
By default, the rotating camera faces upward into the body, serving as a privacy shutter. It only redirects its sensor towards the target when explicitly summoned. The Rabbit AI excels at common tasks like identifying people or objects in the real world, within reasonable limits. However, what truly captivates attention is its unique interaction with the AI. In a demonstration, Lyu points the R1 at a fully stocked refrigerator, prompting it to suggest a “low-calorie” recipe based on its contents, showcasing the AI’s practical and personalized responsiveness.
Certainly, there remain several unanswered questions surrounding the Rabbit R1. One notable inquiry revolves around the battery life – while the company asserts it’s “all day,” the specifics of this claim remain unclear. Additionally, there’s curiosity about how user-friendly the training process will be for the average user.
Despite these uncertainties, some key details have been disclosed. The Rabbit R1 is priced at $199 and is currently open for preorders. The anticipated shipping date is slated for March or April, providing prospective users with a timeframe for the availability of this intriguing AI device.