Nvidia’s ‘Cosmos’ AI Helps Humanoid Robots Navigate the World

Nvidia announced today that it is releasing a family of fundamental artificial intelligence models called Cosmos that can be used for training. humanoidsindustrial robots and driverless cars. While language models learn to generate text by learning from a large number of books, articles and social media posts, Cosmos is designed to create images and 3D models of the physical world.

During a keynote address at the annual CES conference in Las Vegas, Nvidia CEO Jensen Huang demonstrated examples of using Cosmos to simulate activities inside warehouses. Cosmos was trained on 20 million hours of real-life footage of “people walking, moving their arms, manipulating things,” Jensen said. “This is not about creating creative content, but about teaching AI to understand the physical world.”

Researchers and startups hope that fundamental models like these can donate robots used in factories and homes more complex features. For example, Cosmos can create realistic videos of boxes falling from shelves inside a warehouse, which can be used to train a robot to recognize accidents. Users can also fine-tune models using their own data.

A number of companies are already using Cosmos, Nvidia says, including humanoid robot startups Agility and Fig 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. The new feature is part of Nvidia’s existing Isaac robot simulation platform, which will allow robot developers to take a small number of examples of a desired task, such as grasping a specific object, and generate large volumes of synthetic training data.

Nvidia hopes Cosmos and Isaac will appeal to companies looking to build and use humanoid robots. Jensen was joined on stage at CES by life-size renderings of 14 different humanoid robots developed by companies including Tesla, Boston Dynamics, Agility and Fig.

Along with Cosmos, Nvidia also announced. Project Digits, a $3,000 “personal artificial intelligence supercomputer.” which can run a large language model containing up to 200 billion parameters without the need for cloud services like AWS or Microsoft. The company also announced its highly anticipated next-generation RTX Blackwell GPUs and new software tools to help you create AI agents.