Contact us
Earhartstrasse 17, 8152 Glattpark
Mobile: +41 76 577 32 59
Work requests
Mobile: +41 76 577 32 59

Rabbit r1: Your personal AI assistant for seamless app integration

Rabbit's pocket-sized AI "r1" makes everyday tasks child's play.

Rabbit strives to go beyond the productive possibilities of large language models and establish artificial intelligence as an integral part of everyday life. To this end, Rabbit will launch the r1, a personal AI device that performs tasks by using various applications on behalf of the user. The device was unveiled yesterday at the CES 2024 and has generated a great deal of interest online, with millions of views of the promotional posts.

The device, designed by Swedish technology company Teenage Engineering, is reminiscent of the Playdate handheld console, which was also designed by Teenage Engineering. With a size comparable to a Post-it, the device fits easily into any pocket. It uses a comprehensive language model from OpenAI to interpret the user's voice commands. Rabbit's goal, however, as mentioned earlier, is to go beyond the productive capacities of large language models (LLMs) and move to a more proactive and engaging AI.

r1 uses user credentials for seamless application access

Rabbit's unique selling point is its software, a specially developed AI model that learns to operate applications on behalf of the user. This AI model, known as the Large Action Model (LAM), enables the user to perform tasks such as "Order an Uber driver to the office, play motivational music on the way and inform the team about my lateness." LAM then uses the appropriate apps to complete these tasks.
LAM's ability to coordinate apps enables an experience that goes beyond that of a smartphone. "We've reached a point where our smartphones are crammed with hundreds of apps whose complex user interfaces don't communicate with each other," explains Jesse Lyu, founder and CEO of Rabbit. "As a result, users often feel frustrated and lost."

The model was originally trained by analyzing thousands of recorded user sessions in various apps. Users can grant r1 access to their apps by entering their credentials via a setup page on their laptop or desktop.

Lyu explains that the features available via APIs are often limited and extending API functionality is rarely a high priority for app developers. Therefore, Rabbit decided to access the apps directly via the user credentials. In theory, the LAM model can learn everything a user can do within an app.

How do app developers react to Rabbit's strategy?

One inevitably wonders how app developers will react to Rabbit's approach. Rabbit allows users to use an AI agent as a proxy to use their apps. According to the company, it is unlikely that this proxy will show ads or encourage in-app purchases. Lyu does not believe that app developers will have a problem with their services continuing to be used, but in a new way.

Later this year, Rabbit will introduce a "teach mode" that will allow users to directly train the LAM model to use their main applications. Via a web portal, called "rabbit hole", users can show the model how they use their apps and train it in this way. Rabbit explains that by observing user interaction with certain apps, the model can understand and mimic human actions on computer interfaces. This enables the model to learn new tasks reliably and quickly. For example, a video editing program could be taught to remove watermarks from photos.

Rabbit claims that the design and features of the r1 allow it to perform user tasks faster than a modern smartphone. The handheld device has a 2.88-inch touchscreen display, weighs 115 grams and has a scroll wheel to scroll through "cards" with text information on AI tasks. A rotating camera enables computer vision tasks such as recognizing food ingredients. The device also supports video calls and offers Wi-Fi and cellular connectivity, a MediaTek processor, 4 GB RAM, 128 GB storage, a USB-C port and a SIM card slot. The battery should last a whole day.

Lyu announces that the r1 will be available for $199 and will ship by Easter (March 31).


Write comment

Your email address will not be published. Required fields are marked with *

This website stores cookies on your computer. Cookie Policy