The chip company Nvidia wants to expand its leading role in technology for artificial intelligence. Company boss Jensen Huang presented a new generation of Nvidia’s computer platform. Nvidia sees the system, called Blackwell, as “driving a new industrial revolution” through AI. The system is four times more powerful than the current generation Grace Hopper when it comes to training artificial intelligence.
Nvidia’s computer systems dominate AI training in data centers. The group also wants to expand its role in generating content using artificial intelligence. The “Blackwell” system is 30 times better than “Hopper,” emphasized Huang on Monday at the in-house GTC developer conference in San José. Nvidia also has new software for this that can also be used via interfaces via the cloud.
Huang was convinced that in the future most content will not be retrieved prefabricated from storage, but that AI software will generate it fresh based on the current situation. Nvidia developed the computer system for this future. For example, you will be able to talk to buildings via chatbot instead of looking at data somewhere.
AI for weather forecasts
With Grace Hopper, among other things, the chatbot ChatGPT could have been trained within three months with 8,000 Nvidia chips and a power consumption of 15 megawatts, said Huang. With Blackwell you can do this in the same time with 2,000 chips and 4 megawatts of electricity.
ChatGPT is the chatbot from the developer company OpenAI, which triggered the current hype around artificial intelligence over a year ago. Such software is trained with enormous amounts of information and, on this basis, can, for example, generate sentences at the linguistic level of a human and images from text specifications. Nvidia also wants to use the technology to generate weather forecasts, said Huang. For this purpose, the group is developing a world simulation called “Earth 2”.
Nvidia bundles its chips for Blackwell into computer systems with around 600,000 components and weighing more than 1.3 tons, emphasized Huang. The Nvidia technologies originally developed for graphics cards have long been proven in computing work for applications with artificial intelligence. Competitors such as Intel and AMD have so far been unable to catch up. This allows Nvidia’s business – and stock market value – to grow rapidly. Major AI companies such as Microsoft, Google and Amazon are already planning to use Blackwell.
From simulation to reality
Nvidia’s new “AI superchip” is named after American mathematician David Blackwell. Nvidia also wants to promote the use of so-called “digital twins”, in which companies can simulate their entire business on the computer. In the future, before you build something in the real world, you will first simulate it digitally, emphasized Huang.
With this concept, Nvidia is also making Apple’s Vision Pro computer glasses more useful for use in companies. Nvidia’s Omniverse 3D environment, which allows companies to create “digital twins” of their products and factories, is coming to the Apple device. The technology is often used in the design of objects. Huang demonstrated it on Monday at the in-house developer conference GTC, using the example of the South Korean car manufacturer Hyundai, among others. This meant that a designer could look at different color variants of a car model in different environments – and also virtually sit behind the steering wheel. With the $3,499 (around 3,200 euros) Vision Pro, Apple wants to establish a so-called “spatial computer” that can be used to display digital content within the real environment.
Also read:
What does the EU’s new AI law mean?
Nvidia also relies on robots. “Everything that moves will be robotic,” said the company boss. The goal is for robots to be able to learn simply by observing people. “The ChatGPT moment for robotics may be just around the corner,” Huang said. Simulations in which robots learn for the real world should play a central role.