Generative AI took another step forward today as NVIDIA announced a set of cloud services for enterprises to build domain-specific AI models. Called NVIDIA AI Foundations, the new offering includes cloud services for language, images, video and 3D models.

“Generative AI will reinvent nearly every industry,” said NVIDIA CEO Jensen Huang in his keynote presentation for NVIDIA’s Spring GTC conference, which began this morning. “The industry needs a foundry—a TSMC—for custom large language models.”

NVIDIA sees AI Foundations as exactly that. The so-called foundry includes three services: NeMo for human language models (like ChatGPT); Picasso for image, video and 3D generative models (like DALL-E and others); and BioNeMo for biology models such as protein structures.

“In the last few months [we’ve seen] the rising popularity of services like ChatGPT that are based on very, very large GPT models, where one model is used by millions of people every day. But when we work with enterprise companies, many of them are interested in creating models for their own purposes with their own data,” explained Manuvir Das, NVIDIA’s VP of Enterprise Computing, in a media briefing ahead of GTC.

Das says AI Foundations will allow enterprises to create, customize, optimize and deploy models based on their proprietary data. NeMo, for instance, will include pre-trained models such as GPT-8, GPT-43, and GPT-530 that companies can customize for applications such as enterprise search, supply chain forecasting, financial modeling and more. Picasso will allow engineering firms to develop their own text-to-image, text-to-video and text-to-3D models for product design, simulation and digital twin development.

NVIDIA AI Foundations services will be powered by the newly announced NVIDIA DGX Cloud, which provides browser-based access to NVIDIA’s AI systems through public clouds including Oracle Cloud Infrastructure, Google Cloud and Microsoft Azure.

Please click HERE to view the original article.