Oracle bets on vector search capabilities to drive enterprise AI value

Oracle has revealed ‘Oracle AI Vector Search’ as part of the latest iteration of its converged database offering, giving enterprise customers a greater level of control over unstructured data stores. 

The entire release, dubbed ‘Oracle Database 23ai,’ contains over 300 additional features, though the cloud giant specifically highlighted vector search capabilities as a key highlight.

AI Vector Search will enable Oracle customers to search through unstructured datasets with more clarity and without needing to move or duplicate private business data.

Enterprise users will be able to search documents, images, and relational data “based on their conceptual content” rather than specific pieces of information such as words, pixels, or data values.

“AI Vector Search makes it possible for LLMs to query private business data using a natural language interface and helps LLMs provide more accurate and relevant results,” the firm said.

Developers can then easily add search capabilities to new and existing applications, including mission-critical applications that operate sensitive or security-intensive workloads.

Oracle wants to deliver greater value from enterprise data

Oracle said the launch is focused largely on simplifying the process of utilizing unstructured data.

By bringing “AI algorithms to where the data lives, instead of having to move the data to where the AI algorithm lives,” the firm’s new offering allows enterprises to streamline their use of unstructured data, which is becoming increasingly important for AI development.

“You have all of your enterprise business data sitting in Oracle Database - we're bringing the AI to where the business data resides,” Steve Zivanic, VP of database and autonomous services, said at a pre-briefing for the release.

Moving unstructured datasets around to operate vector search capabilities can be complex, and Oracle claims its new offering will lessen the load of this process.

“When you try to use a separate vector database, you have to try to maintain data consistency between the vector database and the business database,” Zivanic said. “That's not an easy challenge."