NVIDIA, the world’s most valuable chipmaker, announced a raft of updates to its software offerings that aim to make it easier for a wider variety of businesses to use generative artificial intelligence (AI).
The chip company officially released what it calls NIMs, or Nvidia inference micro services, which are software packages that take care of many of the logistical issues needed to use AI for a specific purpose. Generative AI, the technology behind chatbots, voice recognition and other automated human interactions with computers, often requires the orchestration of numerous pieces of hardware, software and information retrieval. Many companies do not have the expertise so Nvidia is trying to do that for them, for a fee.
Nvidia founder and chief executive officer Jensen Huang showcased his company’s latest offerings on Monday (Jul 29) at the Siggraph conference in Denver. He’s at the graphics technology conference – the latest in a stream of appearances at public events – seeking to persuade as many industries as possible to start using his technology and make AI pervasive in the economy. Later, he is scheduled to appear on stage at the conference with Meta Platforms CEO Mark Zuckerberg.
Nvidia’s chips have become central to the surge in building new systems being deployed to support AI computing. The company’s revenue doubled last year and is on course to double again in the current fiscal year.
The software and services are included in its Nvidia AI Enterprise product, which costs US$4,500 per graphics processor used per year. The software and services are designed to work on Nvidia hardware.
“The Nvidia NIM is a comprehensive solution for deploying generative AI simplified for developers, but built for applications at scale,” said Kari Briski, Nvidia’s vice-president of product management for AI and HPC software development kits.
Nvidia had made about 100 of the inference micro services available in preview and is now releasing completed versions. For example, Getty Images Holdings’ services is getting improved high-resolution image generation by making the software better at understanding text prompts. Shutterstock’s Edify three-dimensional image generator is going live with the ability to respond to text or images.
Most AI is used by knowledge workers to help with digital tasks, according to Nvidia. To get generative AI into broader circulation, Nvidia is providing software and services that will allow, for example, users of Apple’s Vision Pro headset to create virtual worlds. So-called virtual twins will be used for tasks such as training robotic computers to act more such as humans, cutting out the need for developers to do it manually, Nvidia said. BLOOMBERG
