Oracle's Latest AI Solutions Integrate and Manage Your Data Effectively

Oracle is making a compelling case for keeping your data right where it is, especially if that's within their databases, when you're looking to dive into the world of generative artificial intelligence. This approach could be a game-changer, particularly for enterprise users who are already invested in Oracle's ecosystem.

During a recent partner event in Dubai, Oracle rolled out the general availability of their OCI Generative AI Services. This managed AI service, which initially launched in beta back in September, is now ready for prime time. Alongside this, Oracle introduced two new offerings still in beta: OCI Gen AI Agents and OCI Data Science AI Quick Actions.

Oracle's pitch is clear: building your enterprise generative AI applications on your existing data infrastructure not only leverages your unique data more effectively but also saves you from shelling out extra cash for additional infrastructure. OCI, which stands for Oracle Cloud Infrastructure, encompasses the global network and compute resources, including the Oracle Autonomous Database and Nvidia GPU "super-clusters," that Oracle has invested billions in.

Erik Bergenholtz, Oracle's vice president for strategy and operations, emphasized the company's approach: "We're essentially bringing AI to the data." He highlighted how Oracle's Fusion applications, like ERP and HCM, house exabytes of data, and the company is now integrating generative AI into these vast data pools.

The practical benefits of building on Oracle's database, middleware, and Fusion apps are significant, according to Bergenholtz. He pointed out that while companies could opt to purchase additional software like a vector database such as Pine Cone, this approach adds cost and complexity. "The drawback, of course, is you have yet another piece of infrastructure that raises the cost of cloud, and you now have to actually move and potentially synchronize data across your originating data store," he explained. Using OCI services, however, "just eliminates that barrier, that friction, for our customers."

Steve Zivanic, vice president of Oracle's database and autonomous services marketing, added, "We don't want customers to move data, because the last thing they want is to move 500 terabytes just to get the advantage of generative AI."

The newly available OCI Generative AI service includes pre-built large language models (LLMs), such as Meta's open-source Llama 2 70-billion-parameter model. Oracle has also partnered with Cohere, a venture-backed startup in which Oracle has invested, to include three models in their service: Command for mainstream text-language functions, Summarize for document summary, and Embed for multi-language functions.

Bergenholtz stressed the importance of data privacy, noting that customer data used in OCI for training or fine-tuning models remains confidential and inaccessible to other Oracle customers. Since its beta phase, the service has added capabilities like content moderation, which is performed before prompts are submitted to the language model and after the model's responses are generated.

The service also integrates with the LangChain development framework, enhancing its utility for developers working with LLMs.

The OCI Gen AI Agents product aims to connect LLMs with other resources, such as a customer's proprietary data. The first agent, for retrieval-augmented generation (RAG), allows the language model to tap into data sources like OCI's OpenSearch and, soon, Oracle's Database 23c AI Vector Search and the MySQL Heatwave Vector Store. The AI agents service is set to begin beta testing this month.

The OCI Data Science Quick Actions, stemming from Oracle's 2018 acquisition of DataScience, offers a no-code approach to deploying and fine-tuning language models. It supports several frameworks for distributed training and uses object and file storage to manage model weights. Quick Actions will enter beta testing next month.

Oracle has observed various use cases during the beta phase of their OCI Gen AI service. A common application is automating responses to HR policy questions, like calculating remaining vacation days based on company policy and individual usage. Similar applications include answering healthcare insurance benefits questions. Another prevalent use case is in customer support, where RAG can help summarize cases and provide scripts for next steps, enhancing the user experience.

For those hesitant about integrating generative AI with their data, Zivanic outlined Oracle's approach: by embedding capabilities like vector search directly into the Oracle database and Heatwave, they're bringing the technology to the customers. He acknowledged that some organizations might start with side projects to get comfortable with the technology but predicted that over time, the advantages of a converged database would become clear, outweighing the complexities of managing multiple databases.