"RoboCat" engine from Google will power future general-purpose robots

RoboCat will help pave the way for a new generation of more helpful, general-purpose robotic agents.

While robots are not new to the workplace, their capabilities are often limited. Most can only accomplish simple, precise activity over and over again; they lack the versatility of a human. A general-purpose robot that can pick up new duties with relative ease and consistency might mean more flexible automation and lower overall expenses. However, developing such a bot has proven extremely challenging; they have only lately began to take shape as true entrants into the workforce.

To address the issue, Google's DeepMind has created RoboCat, a self-improving AI agent that can learn to operate different robotic arms and execute varied jobs with as few as 100 samples, and it can create its own training data to assist sharpen performance. RoboCat gets its name from the AI model that powers it, Gato. To get RoboCat started, the team fed Gato a training set full of robotic arms performing hundreds of various tasks.

The more tasks RoboCat learns as a self-improving AI, the better it should get at learning new tasks. This is due to the AI having more experience to draw from, similar to how humans learn over time. RoboCat's ability to learn skills independently and swiftly self-improve, particularly when applied to multiple robotic devices, will help pave the way for a new generation of more helpful, general-purpose robotic agents.