In the realm of science fiction, the existence of humanlike artificial intelligence (AI) often coincides with a physical manifestation, typically taking the shape of androids or robots. This concept may not remain trapped in the depths of imagination for much longer, as several leading tech companies, Nvidia among them, aim to materialize this idea.
Nvidia, a global powerhouse in AI chip manufacturing, recently announced its aspiration to fast-track ’embodiment,’ a process of giving AI models physical forms. “Building foundation models for general humanoid robots is one of the most exciting problems to solve in AI today,” said Nvidia CEO Jensen Huang.
In a spectacular reveal at Nvidia’s annual GTC conference, Huang introduced Project GR00T. Named after the iconic Marvel character, GR00T (Generalist Robot 00 Technology) is envisioned as a universal basis for humanoid robots. The intention behind this AI model is to serve as the ‘brain’ of robots, enabling them to learn skills and solve a broad spectrum of tasks swiftly.
The term ‘Artificial General Intelligence (AGI)’ often crops up when discussing the potential of Project GR00T. It refers to a form of AI that could theoretically learn any task a human being can, without any prior specialized training. While some experts believe AGI is a distant reality, Nvidia’s audacious goal seems to indicate a different belief.
“Human robotics is likely easier,” stated Huang, explaining that a wealth of imitation training data is available due to humans and robots sharing similar construction. This equivalence means that training data captured from human movement can assist AI models in controlling robot movement, enhancing their balance and dexterity.
Another key advantage of humanoid robots is their adaptability to human-centric environments. Human figures can operate in environments we are accustomed to designing and can interact with objects and interfaces (like tools, appliances, and furniture) intended for human use.
Nvidia’s Project GR00T is part of a larger robotics initiative that also includes Jetson Thor. This new computer platform, based on Nvidia’s Thor system-on-a-chip (SoC) and part of the Blackwell GPU architecture, is designed to power the next generation of humanoid robots.
With these ambitious ventures, Nvidia continues to challenge the boundaries of AI and robotics technology, moving one step closer to bringing the concept of embodied AI from the realm of science fiction to our everyday reality.