Nvidia CEO Jensen Huang and Snowflake CEO Frank Slootman expect the rapid advancements in artificial intelligence (AI) will redefine business models by allowing natural language queries and tackling more complicated issues such as supply chain management. The two vendors aim to accelerate this process by connecting an “AI factory” next to customers’ data warehouses.

Huang and Slootman had a fireside chat at this week’s Snowflake Summit 2023 to discuss generative AI’s impact on data innovation in the enterprise.

Huang boasted that the partnership between Nvidia and Snowflake “brings the world's best compute engine to the world's most valuable data.”

The collaboration leverages Nvidia’s GPUs, particularly for machine learning (ML) workloads, and integrates its AI Enterprise and NeMo LLM framework with the Snowflake Data Cloud. “At the core, the big revolution is about the combination of data plus AI algorithms plus compute engine,” he explained.

Nvidia and Snowflake aim to help businesses write AI applications using their own proprietary data. Huang noted the chip company brings AI capabilities to Snowflake’s data management platform, which allows customers to build a large language model (LLM) that enables them to interact with their data as though it were another human.

“The combination of a large language model plus knowledge base equals an AI application,” he said. “A large language model turns any data knowledge base into an application.”

Huang emphasized that the key is data. “You're sitting on a goldmine of natural resources — your company's proprietary data — and we're now going to connect it to an AI engine. And on the other end of that is just intelligence spewing out every single day. ”

Slootman echoed that their combined technology will push the boundaries of what questions users can ask of their data using natural language. “We're all right here that the intelligence is in the data and the models are able to extract the reasoning and intelligence from that data. And that will lead to those incredibly insightful, predictive relationships with data.”

Slootman responded to the question by highlighting that the quickest benefits of AI may not necessarily equate to the most value.

He noted the quickest generative AI application comes in the form of augmented queries, “because that's relatively easy to add.”

This phase one is “low hanging fruit” that allows users to interact with the data in new ways and doesn't require getting value out of data. “It's like searching on dashboards,” Slootman said.

And phase two involves tackling more complex questions, particularly in handling structured and unstructured proprietary enterprise data.

Slootman uses supply chain management as an example, noting supply chain consists of many different entities and the management problem has never been solved in the industry. He said reconfiguring supply chains based on events and using AI to solve problems that haven't been previously addressed could lead to significant advancements in the industry.

As large telcos and banks look to refine their business economics, “collapsing these enormous call center investments that we have, pricing optimizations in the world of retail; it's a redefinition of business models that people are going to see. It's exciting, but that's the real potential there. I think CEOs of large institutions are after that.”

Huang echoed that Nvidia's own supply chain is super complicated, as well as its design database.

“It's impossible for Nvidia to build our GPUs without AI anymore because none of our engineers can go through the number of iterations and exploration that AIs could do for us. And so when we were coming up with AI, our first application was ourselves,” he added.

Nvidia uses LLMs to answer pressing questions, identify exposures and suggest fixes, and then loop in humans to confirm whether the recommended fix is the best one.