Nvidia Empowers Humanoid Robots with Learning via Apple Vision Pro Guidance

Last March, Nvidia made a big move by introducing the GR00T project. It was like shining a bright light in the world of humanoid robots, making it the center of attention and a really important moment. Many famous robot companies joined right from the start, including new ones like 1X Technologies, Agility Robotics, and Apptronik, as well as well-established ones like Boston Dynamics, Figure AI, Fourier Intelligence, Sanctuary AI, and Unitree Robotics. They all worked together to explore the technology of humanoid robots.

Fast forward to the 2025 CES, where Nvidia's CEO, Jensen Huang, unveiled more about the GR00T project. He showed a key part called Blueprint. This new feature is very important because it is based on imitation learning, which is a great way to teach robots new skills, especially for humanoid robots. Why? Because these robots are designed to look like humans doing tasks, like our "technological shadows."

Imitation learning is like a real-life "copycat" game. Humans show a movement, and the robot copies it like a good student. This method is perfect for systems that want to automate tasks, like repetitive jobs on a factory line or busy tasks in a warehouse. These are the places where humanoid robots first show their skills. Remote operation is also very important. It's like an invisible connection between people and robots. Imagine being far away, but you can still control a robot to learn new skills by turning your movements into digital ones. Isn't that awesome? With the powerful GR00T Blueprint tool, users can just wear Apple's Vision Pro to easily create movements. These movements turn into digital twins, and the robot can practice them in a simulated world until it gets really good, just like an athlete training hard for a great performance.