The most popular and comprehensive Open Source ECM platform
Let us know more
Smarter and More Versatile Robots From AI
Google has announced a new artificial intelligence model that can help it train robots to understand tasks like throwing out trash, choosing a drink, or moving objects to a specific location. The model, called Robotics Transformer 2 (RT-2), is a vision-language-action model that can learn from information and images on the internet and translate them into actions for the robot.
RT-2 is an improvement over the previous version of the model, which was based on Google’s large language model Bard. RT-2 can use rudimentary reasoning to respond to user commands, even if the robot has not been explicitly trained on the exact steps. For example, the robot can infer what makes a good improvised hammer (a rock) or what drink to offer an exhausted person (a Red Bull) by using knowledge from the web.
The new model also enables the robot to understand directions in languages other than English, such as Spanish or Japanese. This makes the robot more adaptable and versatile in different settings and scenarios. Google tested RT-2 with a robotic arm in a kitchen office setting and reported that it nearly doubled the robot’s performance on previously unseen tasks.
New AI models bring us one step closer to having robots that can perform complex and diverse tasks in real-life environments, without requiring extensive programming or supervision. The company does not have imminent plans to widely release or sell robots with the new technology, but eventually, they could be used in warehouses, factories, or as home assistants.
The future of robotics is bright and exciting, thanks to AI models. As models continue to learn and improve from more data and feedback, we can expect to see robots that can do more than just follow simple instructions. They can also reason, communicate, and interact with humans and other robots in natural and intuitive ways.
