Oracle keeps AI focus with database updates, new data lake

Heating up

HeatWave is a MySQL database that that allows customers to query and analyze data within the database environment so that they don't have to extract, transform and load data before using it to inform decisions.

Competing platforms include Amazon Redshift, Databricks, Google BigQuery, Snowflake and Teradata.

HeatWave GenAI is a feature within HeatWave and is designed to enable users to build AI models and applications using the data stored in the database. Capabilities included when the feature was initially launched were in-database LLMs, automated in-database vector storage, scalable vector search and HeatWave Chat, an AI-powered assistant that enables users to have natural language interactions with data.

LLM inference batch processing aims to help developers improve application throughput by executing multiple requests simultaneously, rather than just one at a time. Automatic vector store updating, meanwhile, provides AI application developers with the most current data available by automatically updating object storage.

More new HeatWave GenAI features include multilingual support so that similarity searches can be performed on documents in any of 27 languages when developing applications, support for optimal character recognition so developers can include scanned content saved as images when training applications, and JavaScript support to more easily let users build AI chatbots.

Like Mueller, Shawn Rogers, an analyst at BARC US, noted that the new HeatWave GenAI features add significant value because they help simplify developing AI models and applications.

"Heatwave GenAI enables customers to de-risk AI-driven projects through a highly integrated service that removes much of the complexity surrounding creating AI applications," he said. "Built-in LLMs and easy vector store creation help customers avoid do-it-yourself pitfalls without [requiring] extensive AI skill sets."

In particular, automated vector store updating is a significant addition, Rogers continued, calling it "an excellent feature in Heatwave."

Beyond HeatWave GenAI, Oracle updated numerous other HeatWave database features. Highlights include the updates to HeatWave Lakehouse and AutoML, according to Mueller.

New HeatWave Lakehouse capabilities include the ability to write results to object storage so that users can more easily and cost efficiently share and store query results. Also included is automatic change propagation to ensure that users always have access to the most up-to-date data.

New HeatWave AutoML features include increasing capacity so users can train larger machine learning models than was previously possible, data drift detection so developers can know when models need to be retrained, and topic modeling that enables users to more easily discover insights in their text data.

"HeatWave Lakehouse is critical," Rogers said. "[It enables users] to combine HeatWave and lakehouse data, which is key because enterprises need to rely on lakehouses for insights, and even more with AI. And the HeatWave AutoML [update] is very important to keep down the cost of a more powerful -- but therefore also more complex -- database."

In addition to new HeatWave capabilities, Oracle revealed that a free version of the database is now available in the Oracle Cloud Infrastructure (OCI) Always Free Service, enabling organizations to get started with the database by developing and running small applications at no cost.

Other new capabilities

Oracle's HeatWave updates, many designed to better enable developers to build AI models and applications, are just one aspect of the tech giant's push to improve the AI development experience for its customers.

Another development is its plan to develop and deliver Oracle Intelligent Data Lake as a foundational part of the Oracle Data Intelligence Platform.

Oracle expects Intelligent Data Lake to be available on a limited basis at some point in 2025. Once available, its aim will be to combine data orchestration, warehousing, analytics and AI in a unified environment powered by the OCI to more easily enable customers to use data from diverse sources.

Data is growing at an exponential rate. So is the complexity of data and the number of sources from which data is collected. Tools that address that volume and complexity with more advanced capabilities than those built to handle the lower data volumes and more simplistic data of the past are the appropriate next step for vendors such as Oracle, according to Rogers,

"The upcoming addition of Oracle Intelligent Data Lake is a logical step forward for the company," he said. "Nearly all enterprise customers have a highly diverse data ecosystem, and the integration of Oracle's data intelligence platform and OCI clearly provides additional flexibility and function. Customers optimizing their architecture to take advantage of AI will also benefit."

Specific features of Oracle Intelligent Data Lake include generative AI-powered experiences to enable conversational data analysis and code generation, integration capabilities that enable users to combine structured and unstructured data, a data catalog, Apache Spark and Apache Flink for data processing and native integrations with other Oracle platforms, as well as with open source tools.

Like the pending development of Intelligent Data Lake, new AI-powered applications in Fusion Data Intelligence now in preview are aimed at helping Oracle customers derive greater value from their data.

Like many data platform vendors, including Databricks and Snowflake, Oracle has made it a priority to provide users with prebuilt applications specific to individual industries to streamline exploration and analysis.

Now, the tech giant plans to infuse Oracle Cloud Human Capital Management (HCM) and Oracle Cloud Supply Chain Management (SCM) with AI capabilities to further improve the time it takes to reach insights in what Rogers called a "meaningful way."

A new tool in HCM called People Leader Workbench is designed to help organizations achieve business goals by adapting their talent strategy to changing business needs. Meanwhile, a new tool in SCM called the Supply Chain Command Center aims to provide recommendations that better enable organizations to quickly respond to changing supply, demand and market conditions.

"Many companies have long found the time gap between insight and action challenging," Rogers said. "Fusion Data Intelligence … helps Oracle clients close that gap in a meaningful way. Intelligent AI-powered applications are critical for companies looking to deploy AI in business systems for faster, accurate and actionable insights."

Finally, GenDev is intended to provide customers with a cohesive environment for generative AI application development by combining previously disparate tools in Oracle Database 23ai and adding new features.

Among the new features are support for more LLMs including integrations with Google Gemini and Anthropic Claude, improved retrieval-augmented generation (RAG) capabilities, access to Nvidia GPUs and synthetic data creation.

Next steps

With Oracle focusing intently on providing customers with the capabilities to develop and deploy generative AI models and applications, Mueller said it's important that Oracle do so for not only customers deploying on Oracle's own cloud, but also users of other clouds.

Many large enterprises use different clouds for different operations. In addition, they still keep some data on premises and in private clouds. Therefore, as Oracle scales out its generative AI development capabilities, it needs to do so for users of any cloud infrastructure.

"[Oracle needs to] make sure [deployment] is the same across Azure, Google and more clouds," Mueller said. "[They need to] provide multi-cloud management tools, dig deeper in functionality. … Whatever the most popular use cases are, Oracle needs adoption."

Rogers, meanwhile, suggested that Oracle needs to focus more on cost control and clear pricing.

Cloud computing costs were higher than many enterprises expected even before the surging interest in generative AI over the past two years. Now, vital functions such as vector search and storage, developing and running RAG pipelines and deploying LLMs are adding new workloads and their corresponding costs.

"Cost control and transparency must be at the forefront of Oracle's strategy as it continues to add to and integrate its technologies with AI," Rogers said. "Enabling a wider community of users to leverage AI will require simple cost controls to deliver value."

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.