Nvidia unveils AI-powered humanoid robotics platform, devoid of cloud dependence

NVIDIA has announced the launch of Jetson Thor, a groundbreaking AI platform designed for autonomous on-device computing. The mastermind behind this innovative technology is none other than NVIDIA, with its founder and CEO Jensen Huang playing a prominent role in its development and announcement.

Jetson Thor is designed for on-site operation, allowing neural networks, language models, and visual models to be processed locally. This means that service robots in retail or hospitality can spontaneously react to guests, while in logistics, the routes of transport robots can be adjusted in real-time. In industry, robots with Jetson Thor can react to changing production conditions without relying on external servers.

One of the key advantages of Jetson Thor is its energy efficiency. Nvidia has developed the Blackwell architecture, which aims to excel in energy efficiency, providing higher performance per watt. This reduction in energy consumption is particularly beneficial for robots operating continuously.

Jetson Thor offers up to 7.5 times the AI computing power of the current flagship Jetson Orin. It comes with a Jetson AGX Thor Developer Kit, which includes a Jetson T5000 module with 128 GB of memory (LPDDR5X). This powerful hardware allows companies to port existing AI models to the new hardware or develop new applications without having to rebuild the infrastructure on Jetson Thor.

The Jetson Thor platform is also beneficial for healthcare applications. Robots equipped with Jetson Thor can assist with care or diagnostics, recognizing visual anomalies, using voice commands, or precisely controlling their movements, even when the company network is congested.

Jetson Thor's platform is also secure, as sensitive data like camera images do not necessarily have to enter external networks, reducing the risk of data leaks. Hybrid architectures are provided for Jetson Thor, allowing local preprocessing of sensor data and optional cloud transfer for long-term analysis or model updates.

Moreover, Jetson Thor enables multimodal reasoning to be performed directly on the device. It can perform up to 2070 TFlops in FP4 sparse computations, making it a versatile tool for various applications, including humanoid robots, mobile service robots in logistics centers, autonomous machines in factories, and precise assistance systems in medicine.

In conclusion, Jetson Thor is a significant leap forward in AI technology, offering unprecedented power, energy efficiency, and versatility. Its wide range of interfaces, integrated frameworks, and secure on-device computing make it an ideal solution for a multitude of industries, from retail and healthcare to logistics and manufacturing.

Nvidia unveils AI-powered humanoid robotics platform, devoid of cloud dependence