Post by account_disabled on Jan 29, 2024 8:43:35 GMT
Google has unveiled RT-2, a new AI model that can translate your commands into robotic actions without extensive training. Google-Pixel-6-Pro-Optimus-Prime-Robot-1 Although AI chatbots used with large linguistic models (LLMs) are popular these days due to the meteoric rise in popularity of Chat GPT , Bingchat , Lama Meta and Google Cold , this is only a small part of the AI landscape. Another area that has been actively explored over the years is robotics that use sophisticated techniques to replace or assist humans.
Now Google has announced a breakthrough in this domain in the form of a new AI model. Google has unveiled Robotics Transformer 2 (RT-2), the latest AI model with a specific goal: to communicate your desired action to the robot. It uses new techniques to achieve this goal, including a unique visual-language implementation (VLA) that Google claims is the first of its kind. Although several previous models such Country Email List as the RT-1 and PaLM-E have made progress in increasing the thinking ability of robots and making them trust each other, the robot-dominated world depicted by sci-fi movies is likely to look like this. which is a very far-fetched thing. the future RT-2 aims to reduce this gap between fiction and reality by making sure that robots understand the world around them with minimal or no support. Essentially, it is very similar to LLMs in that it uses a transformer-based model to learn about the world from textual and visual information on the web and then translate it into robotic actions, even in test cases where it is not clearly taught.
Google explained several use cases to explain RT-2's capabilities. For example, if you ask the powerful robot RT-2 to throw garbage into a box, it can easily understand what the garbage is, how to distinguish it from other objects in the environment, how to mechanically move it. and must be picked. until and how to throw it in the trash can, all without any special training in any of these activities. Google also shared some very impressive results from the RT-2 test. In more than 6,000 tests, the RT-2 was as adept as its predecessor.
Now Google has announced a breakthrough in this domain in the form of a new AI model. Google has unveiled Robotics Transformer 2 (RT-2), the latest AI model with a specific goal: to communicate your desired action to the robot. It uses new techniques to achieve this goal, including a unique visual-language implementation (VLA) that Google claims is the first of its kind. Although several previous models such Country Email List as the RT-1 and PaLM-E have made progress in increasing the thinking ability of robots and making them trust each other, the robot-dominated world depicted by sci-fi movies is likely to look like this. which is a very far-fetched thing. the future RT-2 aims to reduce this gap between fiction and reality by making sure that robots understand the world around them with minimal or no support. Essentially, it is very similar to LLMs in that it uses a transformer-based model to learn about the world from textual and visual information on the web and then translate it into robotic actions, even in test cases where it is not clearly taught.
Google explained several use cases to explain RT-2's capabilities. For example, if you ask the powerful robot RT-2 to throw garbage into a box, it can easily understand what the garbage is, how to distinguish it from other objects in the environment, how to mechanically move it. and must be picked. until and how to throw it in the trash can, all without any special training in any of these activities. Google also shared some very impressive results from the RT-2 test. In more than 6,000 tests, the RT-2 was as adept as its predecessor.