Covering Disruptive Technology Powering Business in The Digital Age

Written by: Martin Dale Bolima, Tech Journalist, AOPG.

It is full steam ahead for NVIDIA.

The chipmaker turned leader in all things Artificial Intelligence (AI) announced at NVIDIA GTC 2022 advancements to its already world-leading AI capabilities, with Co-Founder and President Jensen Huang doing the honours.

Hopping in on the Hopper Express

Most notable among Huang’s announcements is the planned rollout in October of the NVIDIA H100 Tensor Core GPU, which is in full production already.

First unveiled in April early this year, the H100 is based on the pioneering NVIDIA Hopper™ architecture and built with 80 billion transistors and powered by even more groundbreaking innovations from NVIDIA. Among these advancements making the H100 even better are NVIDIA’s new Transformer Engine, a second-generation multi-Instance GPU and confidential computing. The new H100 also features NVIDIA NVLink®, which helps accelerate even the largest AI models, including large language modelling and advanced recommender systems.

The world’s foremost computer makers—Dell Technologies, Lenovo, Hewlett Packard Enterprise, GIGABYTE and Fujitsu, among others—are already building systems powered by the H100. In total, NVIDIA expects more than 50 H100-powered server models out by the end of the year and several more in Q1 of 2023.

This development is another big step toward democratising AI and expanding its already extensive list of use-cases—of which large language modelling is among the most useful. Large language models are generally viewed as the tech behind chatbots. But they are more than that according to Ian Buck, Vice President and General Manager, Tesla Data Center Business, at NVIDIA.

“Large language models are not just used for building things like chatbots,” explained Buck in an exclusive pre-GTC 2022 virtual media brief. “They are also used for image generation, for recommenders of how we want to buy and surf what we want online, for translation to allow different communities to talk to each other and for text generation, summarisation and understanding language at a holistic level.”

Buck expects more, and the world should, too.

“We’re seeing large language models expand to the life sciences—understanding biology and chemistry and predicting the outcome of materials,” Buck added. “Large language models are one of the most important in AI today, and it [the advancement of AI] is only beginning. With that, we built the NVIDIA H100 to supercharge large language models.”

NVIDIA Gives a Hand to Healthcare

One industry already benefiting from AI is healthcare, and the former’s impact on the latter is poised to grow only bigger with the launch of NVIDIA BioNeMo Service, a biology-centric, AI-powered, open-source framework that can assist developers in generating and predicting molecules, proteins and even Deoxyribonucleic Acid (DNA). Its aim is to spur drug discovery by helping pharmaceutical companies fast-track the development of more effective drugs.

“The mission of drug discovery R&D teams across the world is to find a chemical or protein that interacts with a target to change its course and change the course of disease but exploring this base of chemicals and proteins is nearly infinite,” explained Kimberly Powell, Vice President of Healthcare at NVIDIA. “So, numerical methods and experimental methods alone aren’t enough. Today, we are averaging 50 drugs approved every year, and they cost about USD $1 billion dollars each.”

In other words, developing drugs that actually work takes a lot of time—and a lot more money. The use of large language models, however, can expedite the process and slash expenses significantly.

“Large language models give us new tools to more effectively explore this world of proteins and chemists,” Powell pointed out. “NVIDIA BioNeMo Service gives researchers access to pre-trained chemistry and biology language models… that will infer thousands of representations per second, which can then be used to train a specific task, like predicting protein stability or solubility.”

Put simply, NVIDIA is lending a helping hand to healthcare by helping in the search for cures. But that is not all. NVIDIA has also announced NVIDIA IGX for Medical Devices, an all-in-one platform that accelerates the development and deployment of real-time intelligent machines such as imaging devices.

Advancing Automotive Tech

NVIDIA is also heavily involved in the automotive sector, and its latest innovation for the industry—NVIDIA IGX for Industrial Automation—figures to advance automation further. This platform is similar to NVIDIA IGX for Medical Devices but with a focus on developing and deploying intelligent machines for industrial applications. It can, in particular, help accelerate the development of self-driving cars, which is one of the bigger challenges within the industry according to Danny Shapiro, Vice President of Automotive at NVIDIA.

“Autonomous vehicles are one of the most complex computing challenges of our time, and safety is our number 1 priority,” said Shapiro. “We need diverse and redundant sensors and algorithms which require massive compute. The industry has realised that greater computing performance in the vehicle means greater safety.”

Poised to help solve such complex challenges is NVIDIA DRIVE Thor, which will bring to the transportation industry most of NVIDIA’s advancements across accelerated computing, AI and Deep Learning. Thor leverages Hopper GPU, next-gen GPU and Grace CPU, along with multi-domain computing and NVLINK-C2C capability, and integrates NVIDIA’s Transformer Engine for good measure.

With these features, car manufacturers will soon be able to consolidate a car’s different computer systems—advanced driver assistance systems, driver monitoring, digital instrument clusters and infotainment—into a single system but at reduced costs. This capability, according to Shapiro, will equip car makers with “compute headroom and flexibility to build software-defined autonomous vehicles that are continuously upgradable through secure over-the-air updates.”

In other words, NVIDIA IGX for Industrial Automation, Thor and NVIDIA Drive—a neural reconstruction engine also announced at GTC 2022—may soon usher in the era of fully autonomous but safe vehicles.

The Omniverse, NVIDIA’s Metaverse, Is Evolving

Also announced at GTC 2022 were enhancements to the NVIDIA Omniverse. First showcased at CES 2022, NVIDIA’s own metaverse, the Omniverse, is advancing at an unprecedented pace.

Richard Kerris, Vice President of Omniverse at NVIDIA, noted that the metaverse itself is the 3D evolution of the internet, which means it requires greater computing power and more capabilities. Enter the Second Generation NVIDIA OVX, powered by NVIDIA’s next-generation GPUs and enhanced by networking technology that, according to Kerris, is set to “deliver groundbreaking graphics, AI and digital twins simulation.”

Kerry also announced the NVIDIA Omniverse Cloud, a comprehensive suite of cloud services for artists, developers and enterprise teams to design, publish, operate and experience metaverse applications anywhere. This next-generation cloud has four key applications: Omniverse Farm (a scaling engine for 3D workloads), Omniverse Replicator (for synthetic 3D data generation), Isaac Sim (for training and simulating robots) and DRIVE Sim (for training and simulating autonomous vehicles.

Pivoting from Platform to Software—All for AI

Upgrades to NVIDIA’s AI software frameworks, like Riva, were also announced at GTC 2022. These updates underscore yet again the company’s vision of democratising AI and its pivot to becoming more software-centric relative to bringing AI closer to enterprise use.

“It is important for NVIDIA to really empower the [AI] ecosystem with software stacks that are much closer to the use-case. And as a result of the last few years, we’ve really reinvented our company to become much more of a software company while maintaining our platform approach,” said Manuvir Das, Vice President of Enterprise Computing at NVIDIA. “What we have done at NVIDIA is we have developed a number of software frameworks that provide essential technology that can be used for a variety of use-cases that are very related.”

A Lot to Look Forward To

All this means the future of AI is bright—with use-cases in every industry and for practically any company that wants to benefit from it. And between the NVIDIA H100 now being in full production, the various upgrades announced at GTC 2022 and NVIDIA’s shift to a software-focused strategy, that future appears to be just around the corner.

(0)(0)