Rabbit is an AI firm, and its CEO and creator, Jesse Lyu, says he doesn’t want to replace your smartphone. Not immediately, anyhow. The R1, a $199 standalone AI gadget from his startup, is so incredibly ambitious that Lyu seems to believe he will eventually have to replace your phone. But not quite yet.
The R1 bears some resemblance to a Playdate console or perhaps a more contemporary take on the handheld TVs from the 1990s. It’s a stand-alone device that’s roughly half the size of an iPhone with a 2.88-inch touchscreen, a rotating camera for pictures and movies, and a button or scroll wheel that you can push to interact with the assistant that comes built in. Along with 4GB of RAM and 128GB of storage, it sports a 2.3GHz MediaTek processor housed in a rounded body that was created in partnership with Teenage Engineering. Regarding the battery, Rabbit only mentions that it lasts “all day.”
After Rabbit’s launch event, I had a brief opportunity to use the R1, and I must say that it’s an amazing piece of hardware. There was only one working device (Lyu’s), and even that one was limited in its capabilities by the sporadic hotel Wi-Fi. However, the R1 feels more nicer in person than it does in photos and is surprisingly light. Not surprisingly for Teenage Engineering, the buttons are clicky and gratifying, and it all fits neatly in my hand. Still, it’s undoubtedly a fingerprint magnet.
The true news is in the software that runs on the R1: Rabbit OS, the operating system, and the AI technology that powers it. Rabbit claims that Rabbit OS is built on a “Large Action Model,” which is best described as a kind of universal controller for apps, as opposed to a huge language model akin to ChatGPT. “Like large language models, we wanted to find a universal solution,” he explains. “It would be great if there was a way to activate our services on all platforms, including desktop and mobile, without having to use an app or website.”
It’s kind of like Google Assistant or Alexa in concept. With only one interface, Rabbit OS can buy groceries, order a car, exchange messages, manage your music, and much more. Ask the device to fulfil your request and let it do the rest—no need to balance apps and logins. Lyu explains that the main purpose of the R1’s on-screen interface, which consists of a series of category-based cards for video chats, music, and transportation, is to allow you to independently evaluate the model’s results.
Instead of creating a multitude of APIs and attempting to persuade developers to embrace the R1, Rabbit trained its model to utilise pre-existing apps for its own purposes. Humans interacting with apps such as Spotify and Uber effectively taught the large action model, or LAM, how these apps operate. The LAM gained knowledge about search menu locations, how to determine when an order was confirmed, and what a Settings icon looked like. Lyu claims that all of it is applicable to any app, anyplace.
It is also possible to teach the R1 how to perform an action through its dedicated training mode, after which it should be able to repeat the task on its own moving forward. Lyu provides an example: “You’ll be like, ‘Hey, go to a software called Photoshop first.'” Crack it open. Take your pictures right here. On the watermark, draw a lasso and then click, click, click. Here’s how to get rid of watermarks. According to Lyu, Rabbit OS may automatically delete all of your watermarks moving forward after processing for 30 seconds.
Here, Rabbit has taken a really shrewd stance. Even for the largest IT companies, it can be difficult to convince people to accept a new operating system. The LAM approach gets around this problem by simply teaching the model how to use apps. More generally, we’re witnessing a deluge of new AI-powered devices hitting the market, but all too frequently, those devices are limited to connecting to chatbots. In comparison, Rabbit functions more like a super app, offering a single interface that allows you to accomplish nearly everything. Rabbit OS may be to the app store what ChatGPT is to online searches. That dream is interesting, of course, but it has a thousand cautions and complexities.
Though it’s unclear from listening to Lyu discuss Rabbit OS and the R1, what the firm truly sees for this device is unclear. Despite the fact that it has a SIM card port and can make video conversations, it is not quite powerful enough to replace your phone. Although its main function is voice assistance, the gadget also contains a screen and a camera. It performs a variety of voice assistant functions in addition to being a voice assistant. Rabbit claims to have taken security and privacy into consideration while designing Rabbit OS, yet it still requests that you sign into some of your most frequently used services via its interface. According to Lyu, the R1 is the future of everything in one package as well as a cool accessory. Lyu states that the R1 will begin arriving in March, and the item is currently available for preorder. He may even be hoping that he can outsell Humane’s AI Pin in the marketplace.