AI’s peak value is in training with private data: Oracle’s Larry Ellison

[LAS VEGAS] For artificial intelligence (AI) to reach its peak value, models will need to be trained on private data, said Larry Ellison, executive chairman and chief technology officer of Oracle, on Tuesday (Oct 14) at the technology company’s AI World event in Las Vegas.

He noted that the value of AI models is no longer in their creation, but in their training, adding: “We’re certainly involved in training more multi-modal AI models than any other company.”

Organisations and individuals are spending vast fortunes on training AI models with publicly available data on the Internet, making this the largest and fastest-growing business in history, he said.

“Bigger than railroads, bigger than the Industrial Revolution – I mean, it is a whole new world that is dawning.”

Oracle has been able to tap the business opportunities presented by training AI models – which many aim to use to solve humanity’s biggest problems – from building data centres to providing its customers with access to such models.

The company is now constructing one of its biggest AI clusters for OpenAI: a 1.2-gigawatt data centre in Abilene, Texas, that spans 405 hectares. Part of it is already operational, and it will eventually house up to 500,000 Nvidia graphics processing units.

Although businesses want AI models that can reason with and answer questions directly related to their data, they want to keep this information private.

And this is indeed possible. “You can have your cake and eat it, too,” said Ellison.

Oracle has benefited from allowing AI models to access and reason with its own private data. Its first such project involved customer data, asking AI models questions such as which Oracle customers were likely to buy another product in the next six months, and which products those might be.

SEE ALSO

Oracle names insiders Clay Magouyrk, Mike Sicilia as co-CEOs in surprise move

With the answers to these questions, AI agents facilitated automation. They were able to send e-mails to prospective buyers regarding the products they could be interested in.

“It’s extremely interesting that (AI) can solve a problem like this so quickly, and tell us what the sales force should be concentrating on at Oracle over the next six months,” said Ellison.

However, unlocking enormous productivity gains with AI tools would require automating the entire ecosystem. Oracle is looking to help the healthcare sector do this with AI agents that can be deployed across all departments, from human resources and finance to patient care.

But Ellison also noted: “If we want to automate hospitals and clinics, those hospitals and clinics are not going to be very efficient if the people who regulate (them) are not also automated.”

For example, AI agents could determine the best drug to give a patient with the highest reimbursement level. This can also aid hospitals in obtaining credit facilities from banks, as the technology can compile accounts receivable for a lender to assess.

With administrative tasks handled by AI, doctors’ and nurses’ time can be “spent much, much more efficiently on patients”, Ellison said.