Generative artificial intelligence (AI) is an exciting technology on its own, but the way access to large-scale infrastructure has changed is the real story, according to Red Hat's Sherard Griffin, senior director of engineering for AI services.

AI technology has been around for a while, but the infrastructure needed to run complicated AI workloads has finally caught up, Griffin told SDxCentral. "Generative AI is not about AI in its own right, in terms of what it's doing with generative AI. But it's thinking about how it's now so readily accessible," he said. "To me, that's the exciting part."

Red Hat's initial foray into the genAI space stemmed from the desire to "understand where the value is for customers," Griffin said. "It all started [with] the data." The vendor's Open Data Hub project, for example, focuses on helping customers and open source communities store and process data at scale, especially as they adopt Kubernetes and containerized workloads.

"We wanted to show [customers] how they could process all of their data at scale. ... Now you hear AI, right? If we have all this data, how can we use it to feed our models for machine learning? That dovetails into Open Data Hub," he explained.

Red Hat's Open Data Hub community has eagerly lined up to use open source for generative AI, according to Griffin, who cited the project's work on open source runtime engines and foundation models. "We're still continuing to navigate that open source landscape, finding the best-of-breed tools, but funneling it through our open source projects and ... continuing to be that trusted advisor for customers when it comes to open source," he added.

There are a plethora of AI tools available for enterprises, but none of them are a magic wand. As companies push to adopt AI, they need a clear goal, Griffin said. "You can't just sprinkle [AI] over everything, and all of the sudden the outcome is better," he said.

No two AI tools are the same, and he hasn't seen any one vendor emerge as a clear market leader. In order to identify the best AI technologies for specific use cases, Red Hat's Open Data Hub community works with other partners and open source groups to provide a selection of open source AI tools that are curated to train and "serve up" genAI models.

This is particularly ideal for companies "looking for more of a platform where it's not just about the generative AI model, but" the ability to pull data and models into their own data center environments, he said.

Red Hat, as an infrastructure company, sees its role in the AI arena as helping enterprises understand their goals and associated infrastructure requirements. "That's our sweet spot," Griffin said. "We want to make sure that as customers go into this journey ,they're doing it with eyes wide open, understanding that it takes a substantial amount of infrastructure to run these AI workloads, especially generative AI."

It's one thing for a company to pose a question to some GPT service or other, but running that internally in a data center is a different kind of challenge. "We're demystifying" those complexities and "helping customers navigate what that means," he said.