Oracle Database 23ai Brings the Power of AI to Business Data

To say that databases run the world is no exaggeration. Every time we use a credit card, make a phone call, ship a package, buy groceries, pay taxes, or fill a prescription, a database in the background makes that action possible and manages all the related data. As applications and types of data have expanded over the years, databases have evolved to meet those challenges.

The latest long-term Oracle Database release is Oracle Database 23ai. It includes a new feature called AI Vector Search, which enables unstructured data to be quickly searched using a new data representation called vectors. A vector represents the semantic content or meaning of items like documents, videos, and images.

At its core, a vector is a sequence of numbers known as dimensions. These dimensions capture the essential features of documents, videos, and images. For example, in the case of a picture of a vehicle, the dimensions within the vector could represent the make, model, color, body type (SUV, sedan, F1 racer, etc.), style, and number of seats and doors. In reality, vectors are more complex. Machine Learning algorithms determine the content of the dimensions, and vectors often contain many hundreds of dimensions. This rich representation makes it both possible and easy to search for similar images of vehicles and other real-world objects using mathematical calculations.  

Vectors can also represent the semantic content of documents such as contracts, account statements, emails, documentation, websites, medical notes, etc. Vectors make it possible to search business documents based on the concepts in the document rather than the exact words in the document. Using vectors, a user’s natural language question can now be easily matched to the closest business document related to that question. This breakthrough capability enables endless possibilities for building more powerful applications and simpler human interfaces. 

However, the real power of AI Vector Search in Oracle Database 23ai is the ability to easily combine it with a relational search on business data to solve business problems. Users can now combine business data and vectors to answer user queries without moving or synchronizing data or managing multiple products. 

Combining vector search with business data search is necessary for the vast majority of business needs. For example, users are only interested in searching for products that are relevant to them, within their budget, available in the time frame they need, related to products they already have, etc. This information is stored and maintained in databases. It takes decades to build sophisticated business applications that manage financial data, retail data, communications data, supply chain data, government data, etc. Now, with Oracle’s new AI Vector Search capabilities, these applications can be easily extended with the ability to search on combinations of business data and documents, images, and other human-centric unstructured data.

The other key advantage of having AI Vector Search built into the Oracle Database is that every mission-critical feature of Oracle Database works transparently with AI vectors. You automatically get all the industrial-strength scalability, consistency, availability, security, and performance you have come to rely on with Oracle Database. Because of this, you can safely use AI Vector Search in the same mission-critical customer-facing applications that you use Oracle Database for today.

But the power of AI Vector Search is not limited to content search. It enables you to talk to the data in your database using natural language. Generative AI Chat systems have been trained on the entire content of the internet, but they know nothing about your private data, such as your financial data, your retail purchases, your medical data, etc. This data is securely stored in business databases. Oracle AI Vector Search improves generative AI by allowing business data to be queried using Generative AI Chat. We can now augment AI chat prompts with private database content, enabling chat systems to answer your questions about your private data and produce more relevant and accurate answers. This process is commonly referred to as retrieval-augmented generation (RAG).

Most of the world’s largest enterprises store their critical data in Oracle databases. Now, they can easily use Oracle AI Vectors to search their unstructured data and combine similarity searches with searches on their business data. They will also be able to use generative AI chat to ask natural language questions about their business data without the risk of other users seeing their private information.

To learn more about Oracle Database 23ai, please join us at Oracle DatabaseWorld: AI Edition on May 14th.