During the Oracle DatabaseWorld at CloudWorld 2023, Oracle announced its plans to add semantic search capabilities using AI vectors to Oracle Database 23c. The collection of features, called AI Vector Search, includes a new vector data type, vector indexes, and vector search SQL operators that enable the Oracle Database to store the semantic content of documents, images, and other unstructured data as vectors, and use these to run fast similarity queries.
In addition, the announcement includes the latest updates to Oracle Database services and products:
Commenting on the release, leading industry analysts stated the following:
“The enormous simplicity of searching a combination of product, customer, and AI vector data in a 5 line SQL statement makes it clear that AI Vector Search is a huge winner for Oracle, the AI market, and its worldwide customers.”—Marc Staimer, Senior Analyst, Wikibon
“While generative AI makes it easy for consumers to access data, enterprises need to combine it with real-time internal data for it to be effective for executives. The best way to achieve this is by using Retrieval Augmented Generation (RAG) and a single fully integrated database that includes structured, unstructured, and vector data. Oracle 23c is the only Tier-1 supercloud database that supports all types of structured and unstructured data and also AI vector search.”—David Floyer, Co-Founder & CTO-Emeritus, Wikibon
“Vector search is fundamentally an OLTP feature, as it needs scalability, performance and reliability. Vendors like Oracle that have a proven track record of delivering mission-critical OLTP will be the winners in this market. Oracle’s DNA to deliver mission-critical performance to enterprises will be a key capability for its new vector-based offering, building on the trust and confidence that CxOs have developed for the Oracle Database.”—Holger Mueller, Vice President & Principal Analyst, Constellation Research
The Futurum Group
“Although some early vendors say vector search requires buying a separate product, it has become clear that it’s just another data type that will be added by database industry leaders such as Oracle. Vector search will rapidly become table stakes in the database market by the end of 2024. The winners will be market leaders such as Oracle that have a long-term track record of delivering mission-critical OLTP solutions.”—Ron Westfall, Research Director, The Futurum Group
“For enterprise customers with business-critical data, the vector search capability launching as part of Oracle Database 23c is a very timely and welcome addition. Adding a separate, specialized vector database to generate vector embeddings creates unnecessary data movement and complexity, particularly for organizations that have consolidated their operational enterprise data in an Oracle Database. Just as you can use graph, spatial and time series with Oracle Database, the release of 23c provides customers with the ability to add vector embeddings directly where their enterprise data is stored. This lets them seamlessly deliver generative AI (GAI) functionality such as Retrieval Augmented Generation (RAG) to better align model outputs with their operational data and their business requirements. It’s the smarter approach than trying a do-it-yourself amalgamation of sorts. ” —Bradley Shimmin, Chief Analyst, AI & Analytics, Omdia
“Oracle adds AI vector search to its industry-leading portfolio of data access methods, making it trivial to combine sophisticated operational enterprise data with AI vectors for building class-leading enterprise AI applications.”—Stephen Catanzano, Senior Analyst, ESG
“Vector search is one of those things you probably never knew you needed, but today it is a crucial part of every business’s Generative AI strategy. But do you really need another specialized database engine just to connect a chatbot to your existing business applications? With AI Vector Search, Oracle is introducing another data model to its existing database ecosystem, focusing on open SQL standard support instead of proprietary APIs pushed by other providers. With this new core capability, Oracle Database 23c ensures that enterprise customers are prepared for the AI revolution without the extra effort needed for moving or copying data from one database to another. At all times, they remain protected from attack surface exposures and other data mobility issues, maintaining their security and compliance posture.”—Alexei Balaganski, Lead Analyst & CTO, KuppingerCole Analysts
“With the launch of AI Vector Search, Oracle is bringing enterprise-grade capabilities to vector processing workloads. Class-leading capabilities such as partitioning, parallel execution, RAC, sharding and Exadata smart scan are now available for mission-critical vector database deployments. In the end, customers benefit from a converged database approach, where multiple algorithms can be combined in the same query for optimal business results.”— Richard Winter, CEO, Wintercorp
“Retrieval Augmented Generation (RAG), which is critical to keeping the large language models (LLMs) that power Generative AI timely and relevant, requires a place to store vectors. While we have seen several specialized vector databases emerge, for most enterprise use cases, it will be more cost-efficient and useful to store vectors with the proven databases that they already have. For instance, answering a query comparing sales growth across regions and asking why they are trending up or down is best answered by deciphering social network sentiment from the language model, then backing it up with hard sales numbers. With Oracle Database 23c, customers get the best of both worlds: secure, reliable access to enterprise data and the ability to use RAG to keep their Generative AI foundation models fresh and relevant.”—Tony Baer, principal, dbinsight
“Oracle Database 23c with vector embeddings is substantially simpler than cloning business-critical enterprise data to a standalone vector database and generating vector embeddings there. Oracle’s approach is significantly superior to adding another isolated database, as it eliminates data fragmentation, data artifacts and clones and the pains of data divergence. Clearly, enterprises who are serious about using their business-critical data for generative AI purposes will clearly benefit the most from Oracle’s industry-hardened approach.”—Steve McDowell, Principal Analyst & Founding Partner, NAND Research
Find more about the products and services included in the announcement through these blogs:
Youko Watari is a Product Marketing Director for Oracle database products. Youko spent most of her 25-plus-year career in data and analytics, and she has an end-to-end perspective in data platform implementation as well as analytics and application product development.