Google's DeepMind Introduces Advances for Safer, More Efficient AI-Powered Robots

New systems aim to improve robots' understanding of their environment, speed up decision-making processes, and ensure safety with a 'Robot Constitution'.

• Google's DeepMind team has introduced a set of advances to make AI-powered robots safer and more efficient in domestic environments.
• The first advance, AutoRT, is a system that uses large foundational models to give robots a better understanding of their environment and suggest tasks they can perform.
• DeepMind has also developed a 'Robot Constitution', inspired by Isaac Asimov's 'Three Laws of Robotics', to ensure the robots avoid tasks that could potentially harm humans or animals.
• Two other systems, SARA-RT and RT-Trajectory, aim to speed up decision-making processes and provide clearer instructions for physical actions, respectively.
• DeepMind's advancements have been tested over seven months, safely orchestrating up to 20 robots simultaneously and up to 52 unique robots.

3 Articles

Business Insider The Verge TechCrunch