PhysicalAI. What’s that?
Came across this in #NVIDIA CEO Jensen Huang’s live keynote presentation at CES in Las Vegas. Below is the link to the full video.
While PerceptionAI focuses on understanding images, words, and sounds, and GenerativeAI focuses on creating them, AgenticAI (or digital coworkers as they are called) reasons, plans, and acts iteratively using multiple models to provide a response, PhysicalAI is talked about as the next frontier which aims to understand and simulate the physical world, including dynamics such as gravity, friction, inertia, geometric and spatial relationships, cause and effect and so on.
This will shape the way physical robots interact with us and operate. More sophisticated, and much more advanced.
And the big reveal..
Nvidia Cosmos, described as the “world’s first world foundation model”, has been designed to understand the physical world (trained on 20 million hours of video focusing on physical dynamics). Jensen said Cosmos has been made Open Source!! is available on GitHub and hope it will do the same for the world of Robotics and Industrial AI what Llama 3 has done for Enterprise AI.
Jensen talks about tokens as the building blocks for AI, impact of transformers on computing, neural rendering, synthetic data and more. A must watch presentation on where we are headed with AI.