Google publishes RT-2, a visual language action model, which allows robots to learn to throw trash

CTOnews.com Google today unveiled Robotics Transformer2 (RT-2), a new artificial intelligence (AI) model that can help train robots to understand tasks such as garbage disposal.

Google said RT-2 is a visual language action model that converts information and images collected on the Internet into actions that robots can understand, helping to improve robot learning.

While tasks like picking up trash may sound simple to humans, robots need to learn a range of knowledge. For example, robots must first be able to identify which items constitute trash and then know how to pick them up and throw them away.

RT-2 allows the robot to use knowledge from the network to help it understand how to complete tasks, rather than programming the robot to complete those specific tasks. Even if the robot has not been explicitly trained in the exact steps, it can learn to complete new tasks with RT-2.

Google has no immediate plans to release or sell robots using the new technology, but it could eventually be used in warehouse logistics or home assistant robots.

CTOnews.com Enclosed RT-2 Model Details: Click here to go