On Tuesday, NVIDIA founder and CEO Huang Renxun introduced a number of new NVIDIA technologies, which are expected to bring changes to the trillion-dollar industry. A number of NVIDIA's new technologies can help real and virtual world simulations in various fields, including autonomous driving, avatars, robots , and Earth climate modeling.
Huang Renxun delivered a keynote speech at the GTC online conference this Tuesday, bringing a number of latest releases and demonstrations such as NVIDIA Omniverse Avatar and NVIDIA Omniverse Replicator, and showing a virtual world simulation and collaboration platform for 3D workflow? —— NVIDIA Omniverse, and how Omniverse combines NVIDIA's various technologies.
He shared multiple use cases of the Omniverse Avatar platform, including Project Tokkio for customer support and Project Maxine for video conferencing.
Picture 3
NVIDIA CEO Huang Renxun demonstrated how Project Maxine uses the Omniverse Avatar platform to connect computer vision, Riva voice AI, and avatar animation and graphics to a real-time conversation AI robot-"Toy Jensen Omniverse Avatar"
Huang Renxun said: "How to use Omniverse to simulate warehouses, factories, physical and biological systems, 5G edges, robots, self-driving cars, and even digital twins of avatars is an eternal theme."
At the end of the keynote speech, he announced that NVIDIA will build a digital twin to simulate and predict climate change. The digital twin is called Earth Two, or E-2 for short.
Full-stack data center-level open platform?
Huang Renxun said that accelerated computing has promoted the development of modern AI, and the wave it has set off is flooding the scientific community and all walks of life around the world. And all of this is based on three types of chips-GPU, CPU, DPU, and a wide range of systems deployed from the cloud to the edge-DGX, HGX, EGX, RTX and AGX systems.
From graphics, AI, data science to robotics, NVIDIA has created 150 acceleration libraries for 3 million developers.
On GTC, NVIDIA also released 65 new and updated SDKs.
Huang Renxun said: "NVIDIA accelerated computing is a full-stack data center-level open platform."
When introducing NVIDIA Quantum-2, Huang Renxun called it "the most advanced network platform so far", and with the release of BlueField-3 DPU, the world will usher in cloud-native supercomputing. "The ultra-high performance, wide accessibility, and high security provided by Quantum-2 are exactly what cloud computing vendors and supercomputing centers need.
Cybersecurity is the primary issue for enterprises and the country. Huang Renxun has released a zero-trust framework based on three pillars to meet this challenge.
BlueField DPU isolates the application from the infrastructure. The latest BlueField SDK-NVIDIA DOCA 1.2 implements a new generation of distributed firewall. Huang Renxun introduced that NVIDIA Morpheus assumes that the intruder has entered its interior and will use the "super power of accelerated computing and deep learning to detect the activities of the intruder."
Omniverse Avatar and Omniverse Replicator
Huang Renxun said that with Omniverse, “we have the technology to create a new 3D world or model the physical world.”
NVIDIA also released the Omniverse Avatar, dedicated to helping developers create interactive characters using Omniverse, and enable this character to have visual and language skills, to be able to talk on various topics, and to understand the deep meaning of natural language.
Huang Renxun showed how Project Maxine uses the Omniverse Avatar platform to connect computer vision, Riva voice AI, and avatar animation and graphics to a real-time conversation AI robot-"Toy Jensen Omniverse Avatar".
He also demonstrated Project Tokkio, a virtual waiter in a kiosk in a self-service restaurant, who can see two customers, talk to them and understand their conversations.
Huang Renxun also demonstrated the ability of Project Maxine to add advanced video and audio features to virtual collaboration and content creation applications.
In a demonstration, a woman made a video call in English in a noisy coffee shop. After removing the background noise, she could clearly hear what she said. Her words were transcribed and translated into French, German and Spanish in real time. With Omniverse, the above-mentioned foreign languages all came from the mouth of the avatar, and the voice and intonation of the avatar during the dialogue were exactly the same as her own.
In order to help developers create the massive amounts of data needed to train AI, NVIDIA released Omniverse Replicator, a synthetic data generation engine for training deep neural networks.
NVIDIA has developed two Replicator synthetic data generation engines: Omniverse Replicator for general-purpose robots, for Isaac Sim, and Omniverse Replicator for autonomous vehicles, for DRIVE Sim.
Since its release at the end of last year, Omniverse has been downloaded 70,000 times by designers from 500 companies.
AI model and system
Huang Renxun released Nemo Megatron for training large-scale language models. He stated that this model "will become the largest mainstream HPC application ever."
As a key data structure in modern data science, graphs can now be projected into the deep neural network framework through the new Python software package Deep Graph Library (DGL).
NVIDIA Modulus released today is used to build and train physics machine learning models that can learn and fully comply with the laws of physics.
Triton is an inference server suitable for all workloads. Its upgraded version can now infer forest models and perform multi-GPU multi-node inference for large language models.
Huang Renxun released three new libraries:
·ReOpt-For the 10 trillion dollar logistics industry
·CuQuantum – Committed to accelerating quantum computing research
·CuNumeric-Committed to accelerating NumPy for scientists, data scientists, and machine learning and AI researchers in the Python community.
Huang Renxun also released NVIDIA Launchpad to help bring services based on NVIDIA AI technology to the edge. NVIDIA is working with data center giant Equinix to pre-install and integrate NVIDIA AI into data centers around the world.
robot technology
The NVIDIA Isaac ecosystem now has more than 700 companies and partners, and this number has increased 5 times in the past 4 years.
Huang Renxun announced that the NVIDIA Isaac robot platform can now be easily integrated into the Robot Operating System (ROS). This is a set of widely used software libraries and tools suitable for robotics applications.
Huang Renxun explained that Isaac Sim, built on the Omniverse, is the most realistic robot simulator ever.
"The goal of the simulator is to make it impossible for the robot to distinguish whether it is in a simulated environment or in the real world."? To this end, Isaac Sim Replicator can generate synthetic data for training robots.
Huang Renxun said that Replicator can simulate sensors, generate automatically labeled data, and create a rich and diversified training data set through a domain randomization engine.
Self-driving car
Huang Renxun said that all moving objects will be fully or almost completely autonomous. "By 2024, most new electric vehicles will have true autonomous driving functions."
NVIDIA DRIVE is NVIDIA's full-stack open platform for self-driving cars, and Hyperion 8 is NVIDIA's latest complete hardware and software architecture. Its sensor suite includes 12 cameras, 9 radars, 12 ultrasonics and a front lidar, and all processing is carried out through two NVIDIA Orin system-on-chips.
Huang Renxun introduced in detail several new technologies built into Hyperion, including Omniverse Replicator for DRIVE Sim. Omniverse Replicator is a synthetic data generation engine for autonomous vehicles based on Omniverse.
NVIDIA is currently running Hyperion 8 sensors, 4D perception, multi-sensor fusion based on deep learning, feature tracking and a new planning engine.
The interior of the car will also be completely changed. NVIDIA Maxine technology will redefine the way we interact with cars.
"Maxine will turn your car into a housekeeper."
Earth Two
Huang Renxun finally announced that NVIDIA will build Earth Two (E-2) to simulate and predict climate change. "In order to create Earth Two, we will need to use all the technologies invented so far."
About Manly Battery
Located in Shenzhen,China. A leading battery manufacturer mainly offering rechargeable Lithium batteries over 12 years ,widly used for Robostic ,if there is any project need to evluate ,pls feel free to send email to info@manlybatteries.com
