Rabbit has recently demonstrated a new AI agent designed to interact with Android applications. This showcase comes after the company’s previous device, the Rabbit R1, failed to deliver promised capabilities upon its launch.
The demonstration featured a „generalist Android agent“ that operates on a tablet, executing tasks such as finding YouTube videos and gathering cocktail recipes. Users input commands on a laptop, which the AI translates into actions on the tablet.
During the demo, the AI was tasked with downloading the game 2048 and learning how to play it. While it managed to complete the tasks, its performance exhibited some inconsistencies, such as sending messages one line at a time instead of in a single block.
Despite these quirks, the AI agent represents progress, building on the LAM Playground, a web agent introduced last year. Rabbit has been releasing updates aimed at enhancing the functionality of its AI agents, including the ability to train them for specific tasks.
The company indicated that the current demonstration only showcases the core functionalities of the Android agent. Further details regarding its upcoming multi-agent system are expected to be revealed in the coming weeks.
For more information, visit the original article on The Verge.