Google’s DeepMind recently unveiled a groundbreaking AI model, RT-2, which translates vision and language into robotic actions. This new model, a vision-language-action (VLA) model, is a Transformer-based model trained on text and images from the web. RT-2 can directly output robotic actions, effectively enabling it to “speak robot.”
The pursuit of helpful robots has long been a challenging endeavor due to the complex, abstract tasks they need to handle in highly variable environments. Traditionally, robots have required training on billions of data points across every object, environment, task, and situation in the physical world. However, RT-2 takes a new approach by leveraging the power of AI to transfer knowledge from web data to inform robot behaviour. This allows RT-2 to recognise objects in context, distinguish them from similar objects, and understand how to interact with them.
RT-2’s ability to transfer knowledge from a large corpus of web data to robot actions represents a significant advancement in robotics. This capability enables robots to more rapidly adapt to novel situations and perform tasks they haven’t been explicitly trained for. In testing, RT-2 models functioned as well as previous models on tasks in their training data and almost doubled their performance on novel, unseen scenarios. This demonstrates that RT-2 allows robots to learn more like humans do, transferring learned concepts to new situations.
The introduction of RT-2 also has implications for the future of robotics in human-centred environments. While there is still much work to be done to enable helpful robots in these settings, RT-2 shows an exciting future for robotics just within grasp.
In addition to the advancements in AI-powered robotics, the global robotics industry is also grappling with safety and certification considerations related to AI in robotics. The European Machinery Product Regulation and the AI Act are currently under review, with proposed requirements for mandatory third-party certification of AI-enabled robots. This could impact any company selling robots on the European market, particularly SMEs and start-ups.
To stay current with the latest developments in robotics, it’s essential to follow industry news, trends, and insights. Here are six top robotics blogs to consider:
- Robohub: A global community covering all things robotics, featuring many different perspectives, including robotics research, start-ups, business, and education.
- Robotics.org: The Robotics Industries Association’s online presence, with a news stream and expert analysis of the latest issues.
- RoboGlobal News: A resource for the latest headlines in robotics, as well as occasional posts offering deeper insights from an investment perspective.
- Robotics Business Review: A comprehensive online robotics news and information resource covering all aspects of the business of robotics.
- The Robot Report: A blog headed up by roboticist Frank Tobe, reporting regularly on developments from across the industry.
- Robotics Industry News, Applications, and Trends at Robotiq: A blog focusing on collaborative robots and the technical issues surrounding robotic grasping, force sensing, and robot vision.
In conclusion, Google’s RT-2 model represents a significant advancement in AI-powered robotics, enabling robots to more rapidly adapt to novel situations and perform tasks they haven’t been explicitly trained for. As the global robotics industry continues to evolve, staying current with the latest news, trends, and insights is crucial for success.
