“Science and Technology Innovation Board Daily”, June 3: “The era of robots has arrived. One day, all moving objects will run autonomously.” In his keynote speech before the opening of the COMPUTEX conference, Hwang In-hoon made this judgment.

In the AI boom, Hwang In-hoon, as the head of Nvidia, affects the direction of attention of the market and industry with every move, word, and action. Where will AI stand next? This time, Hwang In-hoon pointed at the robot.

He emphasized the importance of Miracle Man and AI in future development — “The new wave of AI is physical AI. AI can understand the laws of physics and fight side by side with humans... It's really exciting that robots and physical AI are becoming a reality, not just in science fiction.”

▌What is “physical AI”?

Physical AI is physical AI in English, also known as “physical AI.” An article previously published in “Nature MachineIntelligence” by Aslan Miriyev of the Swiss Federal Laboratory for Materials Science and Technology and Mirko Kovač of Imperial College London defined “physical AI” in this way:

Physical AI refers to physical systems that can perform tasks commonly associated with intelligent organisms, and can achieve collaborative evolution of the body, control, morphology, movement execution, and perception.

In this speech, Hwang In-hoon said that most AI today do not understand the laws of physics and are not based on the material world. Generating images, videos, 3D graphics, and many physical phenomena requires AI that is based on physics and understands the laws of physics.

“The AI that works for us must understand the world model to understand how to interpret and perceive the world.” Researchers and companies around the world are developing robots powered by physical AI. These AI models can understand instructions and perform complex tasks autonomously in the real world.

▌The era of robots has arrived

Robotics is a pretty broad concept. Hwang In-hoon said that when he mentions “robots,” people usually think of humanoid robots, “but that's not entirely true. Everything is going to be a robot.” Robots will spread across all factories, factories will coordinate robots, and these robots will manufacture new robot products.

In Hwang In-hoon's vision, the next high-yield robot product manufactured by robots in a robot factory may be a humanoid robot, and the robot that is easiest to adapt to the world is also a humanoid robot.

“The robot era has arrived, and the next wave of AI has arrived.” Standing in front of various robot projection screens, Huang Renxun opened his arms and announced the opening of a new era. “Robotics driven by physical AI will completely change the industry. It's not the future; it's happening.”

Among Nvidia's products, there are many items related to robots, including the NVIDIA Omniverse platform for simulation applications, the Project GR00T humanoid robot general basic model, the Jetson Thor robot computer, and the NVIDIA Isaac platform. Companies such as BYD Electronics, Siemens, and Terida are also integrating Nvidia's autonomous robot technology into their own factories.

Nvidia has also formed a research group, GEAR, whose full name is Generalist Embodied Agent Research, which means “Universal Embodied Intelligence Research.”

Jim Fan, one of the group's leaders and senior research scientist at Nvidia, also made a similar statement to Huang Renxun at the same time: every moving machine in the future will be autonomous, and robots and simulated agents will be as ubiquitous as iPhones. “We're building a foundational intelligence: an AI with universal capabilities that can learn how to act skillfully in multiple virtual and real worlds. 2024 will be the year of robots, the year of game AI, and the year of simulation.”

It is worth mentioning that as the fire of AI burns from chatbots to humanoid robots, Nvidia is naturally investing more.

Last week, it was revealed that OpenAI had restarted the robot team that had previously been abandoned. A person close to OpenAI said that the newly formed robot team has been initially formed and has been in operation for about two months.

Peter Welinder, vice president of OpenAI and a former robotics team member, said, “We always plan to return to the robotics field, and our collaboration with Figure AI has shown us a path to explore the outstanding achievements humanoid robots can achieve driven by high-performance multi-modal models.”