During its GTC 2024 event, NVIDIA has unveiled a series of cutting-edge innovations and projects in the field ofartificial intelligence (IA) and computing advanced.
Among the most significant announcements, the project named stands out Project GR00Ta platform model for humanoid robots for general use. Robots powered by GR00T, acronym for Generalist Robot 00 Technologywill be designed for understand natural language and emulate movements, observing human actions, quickly learning coordination, dexterity and other skills to navigate, adapt and interact with the real world.
In his presentation during the GTC, the CEO Jensen Huang showed several robots successfully completing various tasks. Also refer to videos published on X.
Project GR00T: the model for humanoid robots and Isaac robotic platform
The company led by Huang removes the veil from a platform – Project GR00Tin fact – fundamental for the creation of humanoid robots capable of interacting safely and naturally with the surrounding environment and with human beings.
NVIDIA then presented Jetson Thora System-on-a-Chip (SoC) aimed at humanoid robots, along with significant improvements to the Isaac robotic platform.
NVIDIA CEO Huang noted that creating a baseline model for general-purpose humanoid robots is one of the most exciting challenges to solve in AI. Several enabling technologies can be combined to make this significant leap forward.
NVIDIA Isaac
The new efforts intersect with NVIDIA Isaaca powerful end-to-end platform for developing, simulating, and deploying AI-enabled robots.
The NVIDIA solution allows you to build robots from scratch for specific uses using AI software. Isaac offers a wide range of tools, including Isaac AMR forcomplete autonomy of robots, Isaac Sim for scalable robotic simulation and synthetic data generation and Isaac ROS GEMS providing pre-trained models and accelerated in hardware. It also includes tools like PEOPLE Toolkit for the fine-tuning of pre-trained AI models e TensorRT for optimizing high-performance inference steps.
It SoC Jetson Thor
With Jetson Thor, NVIDIA begins to close the circle by making robots increasingly versatile and not “confined” to specific uses. The SoC, in fact, was created as a new computing platform capable of performing complex tasks and interacting safely and naturally with people and machines. It has a modular architecture optimized for performance, power and size.
The chip includes a GPU next generation based on architecture NVIDIA Blackwell with a transformer-based engine capable of expressing 800 TeraFLOPS for running multimodal generative AI models such as GR00T in 8-bit floating point.
With an integrated functional safety processor, high-performance CPU cluster, and generous bandwidth, Jetson Thor significantly simplifies design and integration efforts.
Enhancement of the ABCI-Q supercomputer for quantum research
NVIDIA representatives then announced the enhancement of the supercomputer ABCI-Q for quantum research. Developed by Fujitsu, it will rely on over 2,000 H100 GPUs and over 500 nodes interconnected with Quantum-2 InfiniBand. This platform will also integrate CUDA-Qproviding powerful simulation tools and the ability to program hybrid quantum and classical computing systems.
On the other hand, as Tim Costa (NVIDIA) points out, researchers demand high-performance simulations to tackle the toughest challenges in quantum computing. The technology developed by the Santa Clara-based company will represent an important milestone and accelerate the development of integrated quantum supercomputers.
The application fields: healthcare, biology, chemistry and transport
As part of the program NVIDIA Healthcarethe company launched more than 20 new generative AI microservices to advance drug discovery, medical technology and digital health.
These are tools that will enable healthcare companies around the world to benefit from the latest advances in generative AI, from anywhere and on any cloud platform. Suite of new services includes enhanced NVIDIA NIM AI models and workflows with industry-standard APIs to help build and deploy cloud-native applications, including advanced image processing, natural language and speech recognition, and industry studies of biology, predictions and simulations.
Con NVIDIA DRIVEInstead, the idea is to support the next generation of the most advanced vehicles. Starting from means of transport traditional, through robotaxis and autonomous delivery vehicles, the NVIDIA platform works to improve movements and safety with artificial intelligence.
The computing platform in-vehicle DRIVE Thor provides cockpit functions, safety support, advanced automation and autonomous driving. Several important players in the automotive industry have already adopted the new advanced tool.