NVIDIA Research showcases how Physical AI combines graphics, AI, and robotics at SIGGRAPH conference in Vancouver. New software libraries for physical AI, including NVIDIA Omniverse NuRec 3D Gaussian splatting libraries, Metropolis updates, and reasoning models like Cosmos Reason, are unveiled to advance AI capabilities in robotics and vision systems.

Physical AI development relies on creating lifelike 3D environments for training advanced AI systems, such as humanoid robots, in simulation. NVIDIA’s research in real-time rendering, computer vision, motion simulation, and AI reasoning over nearly two decades enables the construction of virtual worlds where robots can learn safely through trial and error, enhancing physical AI capabilities.

NVIDIA’s legacy in ray tracing and real-time computer graphics, combined with AI models for neural rendering, is essential for creating realistic virtual environments to train advanced physical AI systems. AI helps in the rapid reconstruction of 3D worlds from images, showcasing the integration of cutting-edge technology in physical AI simulations for various applications like agriculture and manufacturing robots.

Read more at NVIDIA: NVIDIA Research Shapes Physical AI