On Friday, Google made an announcement about a novel artificial intelligence model called Robotics Transformer 2 (RT-2), designed to assist in training robots to comprehend tasks like disposing of garbage. According to a blog post by Google, the vision-language-action model is trained using internet data and images, which can then be translated into actions for the robot. While humans may consider tasks like picking up trash to be simple, it necessitates the robot to understand a series of actions. This includes recognizing what constitutes trash, knowing how to pick it up, and throwing it away.

Rather than programming the robot to perform these specific tasks, RT-2 enables it to utilize knowledge obtained from the web, aiding its understanding of how to complete the task, even without explicit training on the exact steps. Google stated that the new model has significantly improved robots' performance in previously unseen scenarios compared to the earlier version of the model, nearly doubling it. Additionally, the new version can employ basic reasoning to respond to user commands, Google added. The New York Times reported that while Google currently has no immediate plans to widely release or sell robots with this technology, they could potentially be utilized as warehouse assistants or home helpers in the future.