Nvidia unveiled a bold stride in robotics at GTC 2024 with Project GR00T, a universal blueprint aimed at teaching humanoid systems by simply watching people. The idea is to craft robots that can learn from human motion and everyday interactions, turning observation into usable skill. This approach promises robots that not only copy movements but grasp the intent behind them, enabling smoother integration into real-world settings.
With GR00T, robots gain a practical handle on natural language and the subtleties of human motion. They can interpret spoken cues and mirror actions with careful timing, permitting them to adjust to varied environments and collaborate on tasks ranging from household chores to industrial assembly. Nvidia stresses that these machines will be able to navigate diverse spaces, adapt to changing contexts, and engage with humans and objects in meaningful ways.
According to Nvidia leaders, developing basic humanoid models stands out as a frontier in artificial intelligence, highlighting the potential to build systems that learn rapidly, reason about their surroundings, and operate with a level of dexterity that closely resembles human capability.
Alongside GR00T, Nvidia announced Jetson Thor, a new robotic computing platform built around next‑generation GPUs grounded in the Blackwell architecture. This platform is engineered to deliver substantial compute capacity, supporting real-time perception, planning, and control for robotics workloads. Jetson Thor is designed to mesh with products from prominent humanoid robot manufacturers, including 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics and Figure AI, enabling a broader ecosystem for developing and deploying capable robotic assistants in various sectors.
Earlier demonstrations highlighted Nvidia’s ongoing push to showcase some of the most powerful tools available for accelerating AI, vision, and control systems in intelligent machines. The company emphasizes that these advancements aim to empower robots to operate more autonomously and safely in everyday environments, expanding the range of tasks they can perform and the reliability with which they can execute them.