Oracle debuts cloud generative AI service to help enterprises deploy and fine-tune language models


Database giant Oracle Corp. unveiled its long-awaited Oracle Cloud Infrastructure Generative AI service today, launching it with various innovations that will enable big companies to leverage the latest advancements in generative artificial intelligence technology.

The new OCI Generative AI service is a fully managed and cloud-hosted offering that enables enterprises to leverage powerful large language models such as Meta Platform’s open-source Llama 2 and Cohere Inc.’s proprietary algorithms, and integrate them with existing systems. In doing so, companies should be able to automate many of the manual business processes currently performed by humans.

According to Oracle, OCI Generative AI service comes with support for more than 100 languages, offers an improved cluster management experience for graphics processing units, and provides flexible fine-tuning options.

Oracle said its new service will enable those models to be consumed via application programming interface calls to address tasks such as text generation, summarization and semantic similarity. Customers will be able to embed generative AI securely within their existing technology stacks with tight data security and governance protocols in place.

The company announced OCI Generative AI service in June, describing it as an alternative to Microsoft’s Azure OpenAI service, which allows customers to deploy large language models that respond to human prompts in a conversational way.

Although Cohere’s AI models and Llama 2 are known to be very powerful, they’re not all that useful for most businesses unless they can be enhanced and refined with their internal data. To that end, customers can use the OCI Generative AI Agents service that allows them to use their own data to augment the LLM’s capabilities.

Available now in beta test, the OCI Generative AI Agents service uses retrieval-augmented generation techniques to fine-tune generative AI models so they can provide much more useful, contextualized responses. It provides access to a RAG agent that utilizes enterprise search capabilities built on OCI Open Search to help LLMs grab the most up-to-date information to inform their responses and predictions.

Although the initial release supports RAG through OCI OpenSearch, the company plans to integrate a wider range of data search and aggregation tools, including Oracle Database 23c with AI Vector Search, as well as MySQL HeatWave with Vector Store. The offering will also support prebuilt agent actions across the company’s suite of software-as-a-service applications, including Oracle Fusion Cloud Applications Suite, Oracle NetSuite and Oracle Cerner.

Oracle said OCI Generative AI service is available in the Oracle Cloud in multiple regions, and can also be accessed on-premises through its OCI Dedicated Region platform. The service is also being integrated with Oracle’s portfolio of cloud-based business applications, including the Oracle Enterprise Resource Planning, Human Capital Management, Supply Chain Management and Customer Experiences apps. In addition, Oracle said its generative AI service will come to its database portfolio via its Autonomous Database Select AI feature.

In a related update, Oracle said it’s expanding the capabilities of its OCI Data Science platform, adding a new AI Quick Actions feature in beta next month that paves the way for no-code access to various open-source LLMs. The feature will help customers to build, train, deploy and manage LLMs from open-source libraries Hugging Face Inc.’s Transformers or PyTorch.

Greg Pavlik, OCI’s senior vice president of AI and data management, said the company is focused on using generative AI to solve real-world business use cases and pave the way for more widespread enterprise adoption of the technology. “To do this, we are embedding AI across all layers of the technology stack by integrating generative AI into our applications and converged database, and offering new LLMs and managed services,” he said. “Instead of providing a tool kit that requires assembling, we are offering a powerful suite of prebuilt generative AI services and features that work together to help customers solve business problems smarter and faster.”

The new service, with its API-based access, could be a compelling option for some enterprises, as it is somewhat unusual in that it’s being made accessible both in the cloud and on-premises, said Andy Thurai, vice president and principal analyst at Constellation Research Inc. That said, the analyst said, the service also has some big limitations, notably the fact it only supports LLMs from Cohere and Meta’s Llama 2, and it’s targeted at only a narrow range of use cases, such as text generation and summarization.

“In terms of its overall generative AI offerings, Oracle remains far behind Microsoft, Google and Amazon, which provide greater functionality and more flexible deployment options,” Thurai said. “However, the option to integrate with their ERP, HCM, SCM and CX applications running on OCI could make this offering attractive to a number of enterprises, if it’s priced right. But if not, AWS might win this easily, as its generative AI services are currently far ahead of Oracle’s.”

Although it may have limitations, Oracle’s generative AI service has three very big advantages, said Thurai’s colleague Holger Mueller. He said the first is that Oracle has built a cloud infrastructure that’s designed especially for database, with fast servers that connect to an even faster network, which happens to the sweet spot for AI too. “There’s lots of evidence for this, as both Microsoft and Nvidia are running native workloads on Oracle,” Mueller said.

The second advantage is the willingness of Safra Catz and Larry Ellison to invest when they see an opportunity. “Oracle has been investing in the cloud at a rate of 50% of its free cash flow for several quarters now, and the result is an attentive AI platform and cloud for enterprises,” the analyst said. “In addition, Oracle’s applications business ensures it has a very different experience in terms of building a generative AI infrastructure, which may make its service more interesting compared to its platform-as-a-service and infrastructure-as-a-service-focused competitors.”

The launch of the new service should at least help Oracle to establish itself as an option for enterprises in generative AI development. Already, the company is pushing to become a key player on the infrastructure side, providing cloud-based access to Nvidia Corp.’s most powerful GPUs.

Photo: Flickr CC

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy



Leave a Reply

Your email address will not be published. Required fields are marked *