The Humane AI Pin has collapsed, however Rabbit continues to be kicking. The corporate printed a weblog publish and video at present displaying off a “generalist Android agent,” slowly controlling apps on a pill in a lot the identical approach that Rabbit claimed its R1 machine would over a 12 months in the past. (It couldn’t, and may’t.) The work builds on LAM Playground, a “generalist internet agent” Rabbit launched final 12 months.
The engineers don’t use the Rabbit R1 in any respect for the demonstration. As an alternative, they sort their requests right into a immediate field on a laptop computer, which interprets them to actions on an Android pill. They process it with issues like discovering a YouTube video or finding a whiskey cocktail recipe in a cocktail app, gathering the components, after which including them to a Google Hold grocery checklist. At one level, they ask it to obtain the puzzle sport 2048 and determine the best way to play it, which it does, albeit slowly.
The mannequin typically does the issues they ask, generally properly and generally with quirks like sending a poem over WhatsApp one message at a time as an alternative of in a single block. One of many engineers wonders if they need to have requested it to make use of line breaks of their immediate, however they don’t return to attempt once more.
Rabbit’s AI agent is clearly nonetheless a piece in progress, because it has been since the R1 launched with nearly not one of the capabilities that founder and CEO Jesse Lyu introduced in January 2024. Rabbit has steadily rolled out updates, just like the capacity to coach its AI brokers to finish particular duties or immediate it to remake its personal interface. The examples it introduced at present are “solely the core motion loop an Android agent completes,” in line with Rabbit’s weblog publish. The corporate guarantees to share extra about its “upcoming cross-platform multi-agent system” in coming weeks.