Jensen Huang, CEO of Nvidia, took center stage at the annual GPU Technology Conference (GTC) this week, outlining a range of exciting initiatives that position the company at the forefront of the booming AI revolution.
Huang acknowledged the ever-growing demand for data processing in AI development. To address this, he announced the development of an even more powerful GPU than their existing Hopper line, hinting at a future with significantly stronger processing capabilities. Codenamed “Blackwell” after mathematician David Blackwell, this new hardware transcends the traditional GPU concept, functioning as a comprehensive platform.
Nvidia is making significant strides in robotics, with a particular focus on humanoid robots. Project GR00T (Generalist Robot 00 Technology), inspired by the iconic Marvel Comics character Groot, signifies a major leap forward in this field. Huang showcased several AI-powered robots, including the Jetson Thor platform, designed for complex tasks and natural human interaction. Notably, Nvidia is building a comprehensive system for training these robots using visual data like videos and VR simulations.
Nvidia’s Isaac Lab plays a crucial role in training GR00T robots at scale. Nvidia Osmo, a dedicated orchestration service, coordinates training and inference workflows across various Nvidia systems, ensuring seamless operation. The Isaac robotics platform received a significant upgrade with the introduction of Jetson Thor. Nvidia’s Isaac reinforcement learning gym provides a virtual environment for real-world adaptation training.
The concept of a digital twin – a virtual replica of a physical system – is being revolutionized by Nvidia’s Omniverse platform. Huang presented a demonstration of a digital twin representing a 100,000-square-foot warehouse bustling with autonomous systems – both human and robotic – interacting seamlessly. This technology has the potential to streamline physical plant design, engineering, and human-robot collaboration within industrial settings. Siemens, a global leader in industrial engineering, is collaborating with Nvidia to connect their Xcelerator platform to Omniverse, paving the way for the “industrial metaverse.”
Nvidia announced that Apple’s innovative Vision Pro headset integrates with the Omniverse platform, allowing users to experience the digital world in a whole new way. This integration, combined with the connection of various design tools to Omniverse, opens doors to groundbreaking design workflows.
Huang emphasized the inevitable integration of AI into the automotive industry, highlighting its role in enhancing safety, convenience, and efficiency. Nvidia’s technology spans the entire automotive production chain, from self-driving car systems to manufacturing processes. The Jetson Thor platform, slated for shipment next year, will be incorporated into vehicles from Mercedes and Jaguar Land Rover.
Nvidia is actively collaborating with numerous companies like 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics, and more to develop robust AI platforms for various applications. Huang concluded by underlining the immense potential of building foundation models for humanoid robots, expressing his belief that these advancements will propel the field of artificial general intelligence forward.
