With the rise of AI chatbots powered by Large Language Models (LLMs) such as ChatGPT and Bing Chat, Google seeks to create a new advancement in the same domain with their Robotics Transformer 2 (RT-2). While it is also an AI model, RT-2 is powered by a unique Visual-Language-Action (VLA) which uses a Transformer-based model to learn its surroundings and environment without any training.

What this means is Google’s RT-2 can understand the world and humans talking to it as it can process textual and visual information from the web and translate it into robotic actions in real time. Google claims that the VLA AI model of RT-2 is the first of its kind that allows such advanced reasoning abilities without any training needed.

To put it into simpler terms, a use-case or test Google conducted with the RT-2 showed how the VLA-powered robot can easily determine what we mean by “trash” and throw it in a bin when commanded to.

Other trials conducted by Google showed positive results for RT-2. It scored 62% in unseen scenarios in the 6,000+ trials it went through. This is an impressive feat compared to the RT-1, its predecessor, which scored only 32%.

More testing and trials are being conducted to see how mature the RT-2 is and can be for real-world uses and applications but developments such as this one make us excited about the future and growth of robotics. We wonder how this unveiling can influence the trends and future of robots.