Español
Português
Việt Nam
日本語
русский 
عربي 
Deutsch
한국어
Türkçe
Français
Italiano
Yesterday morning, NVIDIA GTC (GPU Technology Conference) 2024 AI developer event, known as the "AI wind vane" in the industry, was held at the San Jose Convention Center in San Jose, California, United States, laying the groundwork for another leap forward in AI.
As a global-level top AI event, this conference can be said to be in the limelight. At the conference, Jen-Hsun Huang, founder and CEO of NVIDIA, delivered a keynote speech entitled "Witnessing the Transformative Moment of AI", releasing the latest breakthroughs in the fields of accelerated computing, generative AI, and robotics, all of which can be called the king of the bomb level.
These include the AI chip Blackwell GPU (the world's most powerful chip), the new network switch X800 series, AI supercomputers, generative AI microservices ......
New humanoid robot base model and brain
In addition, NVIDIA is very bullish on embodied intelligence, "Developing a universal humanoid robotics base model is one of the most exciting topics in AI today."
Jen-Hsun Huang announced the humanoid robotics project GR00T at the conference, a modeling platform for humanoid robots designed to further NVIDIA's R&D in robotics.
Stemming from NVIDIA's Isaac robotics platform tool, robots powered by GR00T understand natural language and mimic rapid learning of coordination, dexterity, and other skills by observing human behavior to adapt to and interact with the real world.
NVIDIA has also developed a new "robot brain" computing chip, Jetson Thor, capable of performing complex tasks. Jetson Thor features a modular architecture optimized for performance, power and size, built on a next-generation Blackwell GPU, with bandwidth up to 100GB/s and AI compute performance up to 800 TFL. AI compute performance of 800 TFLOPs to run multimodal generative AI models such as GR00T. NVIDIA will develop production robots based on Jetson Thor.
At the conference, a pair of small Disney robots showed up to interact with Jen-Hsun Huang, and the robots used Jetson, NVIDIA's first AI chip for robots.
NVIDIA says it is currently building AI platforms for humanoid robotics companies such as 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics, Figure AI, Fourier Intelligence, Sanctuary AI, Unitree Robotics, and Yushu Technology and Xiaopeng Pengxing.
In addition, Isaac Lab, built on NVIDIA Omniverse, has been upgraded in order to provide software support for Project GR00T, allowing developers to use the platform to simulate robots to learn skills, interact with physical world simulations, and support thousands of robots for simultaneous training and simulation. The new features will be rolled out in the next quarter.
At the same time, NVIDIA also released a series of robot pre-training models, libraries and reference hardware: Isaac Manipulator, an accelerated library platform to assist in improving the sensitivity and accuracy of robotic arms, and the Isaac Perceptor software library to improve the depth perception of robots.
The wave of embodied intelligence is here.
In October 2023, NVenture, NVIDIA's venture capital arm, led a $32 million investment in Machina Labs, a robotics factory.
On February 24, 2024, NVIDIA announced the establishment of GEAR, the General Embodied Intelligence Research Laboratory, marking NVIDIA's formal entry into the field of embodied intelligence research and accelerating the process of AI embodiment.
Robots are excellent carriers of embodied intelligence, and Dr. Jim Fan, co-founder of NVIDIA's GEAR lab, also said on the X platform that 2024 will be the year of robotics, gaming AI, and simulation, and NVIDIA's entry is expected to accelerate the development of robotics.
As early as May last year, Jen-Hsun Huang proposed that "the next wave of artificial intelligence is embodied intelligence". That is, intelligent systems that can understand, reason and interact with the physical world, and humanoid robots are the most ideal carriers of embodied intelligence, NVIDIA is bound to have a layout.