NVIDIA to create "the second Earth"?!

Nov 10, 2021 AI

 

At CTG 2021 fall conference, Jensen Huang, CEO of the Nvidia Corporation, once again expressed his obsession with Metaverse to the public.

In a nearly two-hour presentation, Jensen Huang described the major initiatives that NVIDIA is working on and how they will impact the industry, creation, and even the world.

Huang said frankly that " how to stimulate with Omniverse" was a theme throughout the presentation, which also included information on the latest technologies in quantum computing, conversational AI and natural language processing, new applications for virtual worlds, as well as how NVIDIA and its partners are building digital twin for factories, cities and entire regions.

A few days ago, NVIDIA tops $700 billion valuations. It looks confident for NVIDIA, a semiconductor company that plays a major role in AI, smart driving, and the "Metaverse".

 

NVIDIA Quantum-2

At CTG 2021 conference, NVIDIA updated many prominent new features and platforms, NVIDIA Quantum-2 being one of them.

 

To address the dilemma of time-consuming manufacturing of quantum computers and the lack of ultra-high-speed quantum simulators for verification, NVIDIA has created the cuQuantum DGX device, equipped with an acceleration library for quantum computing workflows that can use state vector and tensor network approaches to accelerate quantum circuit simulations.

Simulations that once took months can now be done in just a few days. With cuQuantum on DGX, the quantum computer and algorithm researchers can use the fastest computers available today to invent the computers of tomorrow. With 400Gbps of high bandwidth, NVIDIA Quantum-2 InfiniBand doubles the speed of the network and triples the number of network ports. This triples the performance and reduces the need for data center structure switches by a factor of 6 while reducing data center power consumption and data center footprint by 7%.

However, NVIDIA Quantum-2 is the most advanced end-to-end networking platform to date. It consists of the Quantum-2 switch, ConnectX-7 NIC, BlueField-3 DPU, and a full suite of software for this new architecture, and is a 400gbps InfiniBand platform.

NVIDIA Quantum-2 is a networking platform that combines the performance of supercomputers with the multi-tenant sharing of cloud computing, giving supercomputers native cloud-generated capabilities that will be better optimized.

 

Update the Omniverse

"To say that the Internet has changed everything is an understatement," said Jensen Huang.

People have needed to connect since ancient times, and while the nature of the Internet is a digital representation of the world, covering mainly 2D information such as text and pictures. However, the future may look very different with the creation of modeling technologies that create a whole new 3D world.

When talking about the Metaverse, Huang had some predictions about the Omniverse: "In the future, there will be many designers and creators who design digital things in virtual reality and the Metaverse, and then finish designing in the real world, including cars, bags, shoes and other products." He believes it will be a bigger market, a bigger field, and that the world in the Metaverse maybe a hundred times bigger than the real world.

At the CTG conference, Huang still expressed the same view and gave specific examples.

Things that work in Adobe will be able to connect to Autodesk through Omniverse, which allows designers to collaborate in a shared space. Changes made by one designer in one world will be updated in real-time to all associated designers, essentially like a cloud of shared documents for 3D design. "Omniverse will revolutionize the way that 40 million 3D designers around the world collaborate."

In addition, companies can build virtual factories in which they operate with virtual robots, and the virtual factories and robots are digital twins of their physical replicas, while the physical products they produce may be replicas of the digital versions of the products.

At the GTC conference in April, BMW showed how to build the digital twin of the Regensburg factory. Now their virtual factory has been extended to three other factories with a total area of 10 million sq. m. Omniverse can roll out new simulation capabilities whenever and wherever BMW needs them to make important decisions and introduce automated production processes to increase agility and optimize productivity.

During the presentation, Huang also updated four key features of Omniverse: Showroom, an Omniverse application with demos and sample applications that showcase the core technology; Farm, a system layer for orchestrating batch job processing across multiple systems, workstations, servers, and virtualization; Omniverse AR, which streams graphics to cell phones or AR glasses; and Omniverse VR, NVIDIA's first full-frame interactive ray-traced VR.

 

Virtual avatar

Huang has created a virtual avatar for himself called Toy-Me.

 

"You'll see that the avatar is built based on language processing from the largest language model currently trained, including the voice, which is also synthesized using my voice, and you can also see beautiful images based on ray tracing in real-time," said Huang. Toy-Me can fluently answer questions on climate change, astronomy, and biological proteins.

In addition to an avatar based on Huang, this time he also unveiled the NVIDIA Omniverse Avatar platform, which can generate artificial intelligence avatars.

Avatar is an interactive character developed by NVIDIA with the ability to ray-trace 3D graphics. It can keep an eye on people as they interact with the outside world, communicate with people on a wide range of topics, and even determine the intent of conversations.

Omniverse Avatar brings together NVIDIA's technologies in voice AI, computer vision, natural language understanding, recommendation engines, and simulation technologies. The platform opens the door to creating artificial intelligence assistants that can help handle billions of daily customer service interactions, including restaurant orders, bank transactions, reservations, etc., leading to greater business opportunities and increased customer satisfaction.

 

Earth-two

When it comes to NVIDA's future focus, it mainly includes biosynthesis, climate prediction, and other aspects that are closely related to the future of human beings.

"The future of drug discovery is end-to-end computational modeling of disease pathways involving genes, drug-target interactions, and off-target interactions, and we are witnessing the dawn of a revolution in biology with the convergence of the MillionX million-fold acceleration with machine learning for protein and chemical structure prediction, and Physics-ML simulation methods." Soon after the presentation, Huang expressed his expectations for the future revolution in biology.

As for climate science, predicting climate change to develop environmental and adaptation strategies is arguably one of the most difficult problems to solve today. Modeling the physical properties of the Earth's atmosphere, oceans and waters, ice, land, and human activities and their interactions is not an easy task, and it is still not possible to accurately predict the climate decades from now.

However, long-term climate prediction is essential, and humans need to "mitigate and adapt before it's too late". In NVIDIA's plans, it may be possible in the future to use physical principles and data from principle-based models and observations in Physics-ML models, optimized for training on multiple GPUs and multiple nodes to predict climate change in hyper-real time, create digital twin models of the Earth, or transfer their data to the Metaverse.

At the end of his speech, Huang also announced one more thing.

NVIDIA's last supercomputer was named Cambridge-1, or C-1, and next NVIDIA will start developing a new supercomputer, "E-2," or "Earth-two," the second Earth.

"All the technologies we have invented so far are essential to the realization of the Metaverse, and I can't imagine a more ambitious and important use for it."

This is an article from WeChat official accounts GeekPark (ID: geekpark), written by Yu Sanzhui , translated by Lexi Ren.