Google Unveils Robotics Transformer 2: A Breakthrough in AI for Training Robots in Specific Tasks

Google announced on Friday a new artificial intelligence model that could help train robots to understand specific tasks, such as disposing of garbage.

According to Google’s blog post, Robotics Transformer 2 (RT-2) is a model for vision, language, and motion that has been trained on information and images from the internet, which can be translated into actions for the robots.

While tasks like picking up trash may seem straightforward to humans, they require the robot to understand a sequence of steps.

For example, the robot must first recognize the components of the garbage, then learn how to pick it up and dispose of it properly.

Instead of programming the robot to perform these specific tasks, the Robotics Transformer 2 AI model allows the robot to use knowledge from all over the web to help it understand how to complete the task, even if it hasn’t been explicitly trained on the exact steps.

Google stated that the new model almost doubled the performance of robots in scenarios that were not possible with the previous version of the model.

The company also mentioned that the new version can use basic logic to respond to user commands.

The New York Times reported that the company currently has no immediate plans to launch or sell robots widely using this new technology.

However, in the long run, the technology could be used in warehouses or as home assistants, according to The New York Times.