NVIDIA, already a leader in the field of artificial intelligence, is extending its influence to the booming sector of humanoid robotics. Far from being content with its successes in generative AI, the company is preparing to launch Jetson Thora revolutionary platform specifically designed to power the next generation of robots.
Jetson Thor: an operating system tailor-made for humanoids
Jetson Thor enriches NVIDIA's Jetson range, a series of embedded platforms dedicated to AI and machine learning. Presented last March at a conference where Jensen Huang, CEO of NVIDIA, shared the stage with a multitude of humanoids, this technology will be commercialized from first half of 2025.
The objective? Equipping other companies' robots with a operating system specially designed for humanoids. Jetson Thor will allow these machines to run multimodal AI modelscapable of processing information from various sources: texts, images, sounds, and even sensory data such as touch.
More intelligent and interactive robots thanks to multimodal AI
Thanks to Jetson Thor, robots will be able to understand the natural languageinteract more intuitively with their environment and optimize their energy consumption. The platform's modular architecture will provide increased flexibility and adaptability to developers.
NVIDIA has considerable ambitions and is targeting a fragmented market of “hundreds of thousands” of manufacturers, according to Deepu Talla, NVIDIA vice president of robotics and edge computing.
Generative AI, a catalyst for humanoid robotics
The arrival of Jetson Thor is a testament to the impact ofGenerative AI in the robotics sector. The integration of learning models allows robots to acquire knowledge and adapt to a variety of tasks, marking a significant advancement in the field.
This trend is confirmed by the growing investments of generative AI players in startups specializing in robotics. OpenAI, for example, has invested in Figure AI, a startup that is developing a humanoid capable of communicating with humans using an AI model.
