Nvidia’s ‘Cosmos’ AI helps humanoid robots navigate the world
Nvidia today announced it will release a line of platform AI models called Cosmos that can be used for training human figureindustrial robots and self-driving car. While language models learn to generate text by training on large numbers of books, articles, and social media posts, Cosmos is designed to create images and 3D models of the physical world. .
In a keynote presentation at the annual CES conference in Las Vegas, Nvidia CEO Jensen Hoang gives examples of Cosmos used to simulate operations inside a warehouse. Cosmos was trained on 20 million hours of real footage of “people walking, moving their hands and manipulating things,” Jensen said. “It’s not about creating creative content but about teaching AI to understand the physical world.”
Researchers and startups hope that these types of platform models are possible Providing robots for use in factories and homes more sophisticated capabilities. For example, Cosmos can create boxes of realistic video footage falling from shelves inside a warehouse, which can be used to train robots to recognize accidents. Users can also fine-tune models using their own data.
Nvidia said several companies are already using Cosmos, including humanoid robotics startups Agility and Figure AI as well as self-driving car companies like Uber, Waabi and Wayve.
Nvidia also announced software designed to help different types of robots learn to perform new tasks more efficiently. This new feature is part of Nvidia’s existing Isaac robot simulation platform, which allows robot builders to take a small number of examples of the desired task, such as grasping a specific object and generates large amounts of synthetic training data.
Nvidia hopes that Cosmos and Isaac will attract companies that want to build and use humanoid robots. Jensen appeared on stage at CES using life-size images of 14 different humanoid robots developed by companies including Tesla, Boston Dynamics, Agility and Fig.
Along with Cosmos, Nvidia also announced Project Digits, a “personal AI supercomputer” worth $3,000 can run large language models of up to 200 billion parameters without the need for cloud services like AWS or Microsoft. They also announced the highly anticipated next-gen RTX Blackwell GPUs and upcoming software tools to help build AI agent.
