Nabugu - stock.adobe.com
Oracle has taken on the Microsoft-OpenAI partnership by launching a Cohere-powered generative AI service with unique features.
This week, Oracle rolled out its Oracle Cloud Infrastructure (OCI) service in beta at the company's CloudWorld conference in Las Vegas. Oracle reported it would develop a generative AI service with Cohere in June, a week after joining the startup's $270 million funding round.
Generative AI is poised to transform businesses by providing intelligence on internal and external data through natural language user interactions. The Microsoft and Oracle services, underpinned by large language models (LLMs), let companies integrate the services into business applications to significantly improve customer support, office productivity and business operations.
Oracle plans to embed generative AI services across its business applications, including Oracle Fusion Cloud, Oracle NetSuite and industry-specific applications, such as its healthcare product Oracle Cerner.
The fundamental difference between the Oracle and Microsoft generative AI services is Cohere's second representation model, Gartner analyst Sid Nag said. Both services have LLMs trained on a specific body of data and then fine-tuned to provide more tailored output to queries.
Cohere's representation model adds a deeper text analysis. "[Representation] can be used for understanding language, which in turn can have a variety of advantages, such as sentiment analysis," Nag said.
Sentiment analysis is a technique for determining the feelings and opinions behind the text.
Another difference between the two generative AI services is Oracle's use of retrieval augmented generation. The technology can update LLM outputs and make them more specific to a particular industry or organization, Nag said.
"That means the generative AI system can provide more contextually appropriate answers to prompts, as well as base those answers on extremely current data," he said.
The Oracle-Cohere beta service will let enterprises try ready-to-use pre-trained models or create a custom model that organizations can fine-tune using their data on a GPU cluster with up to 16,000 processors.
Oracle has launched an interface for organizations that want to try the various models. The company hasn't said when customers can use the service for production environments.
"The key that we're focusing on with this service is ensuring that customers retain complete control of their data and understand exactly how it's being used," said Clay Magouyrk, executive vice president of OCI development, in his CloudWorld keynote Wednesday.
The beta interface lets users refine model prompts and parameters until they are satisfied with the results. They then will get code to integrate into applications.
Enterprises can also create a copy of a pre-trained model to train it on their data sets.
Cohere provides developers with open source libraries and an API for accessing its models. By partnering with the company, Oracle shows it's ready to work closely with developers, something Microsoft has done for decades, said Paul Nashawaty, an analyst at TechTarget's Enterprise Strategy Group.
"[Cohere developer tools] are enablers that if Oracle didn't have them, they would be at a competitive disadvantage, especially against somebody like Microsoft and OpenAI," Nashawaty said.
The OCI-Cohere service offers many use cases, Oracle said. Enterprise can create text for new products, marketing campaigns or emails to a sales client. They also can build a chatbot-style user interface for custom support.
Other use cases include data extraction from contracts, tables and job applicants. The models can also summarize lengthy documents.
OCI hosts the services in its Midwest region centered in Chicago.
Oracle and Microsoft are competitors and partners in the cloud market. Last week, the companies reported that Oracle would offer its Exadata database hardware on Microsoft Azure. The offering reflected customer demand for accessing services on multiple cloud providers, the companies said.
Antone Gonsalves is editor at large for TechTarget Editorial. He has deep and wide experience in tech journalism. Since the mid-1990s, he has worked for UBM's InformationWeek, TechWeb and Computer Reseller News. He has also written for Ziff Davis' PC Week, IDG's CSOonline and IBTMedia's CruxialCIO, and rounded all of that out by covering startups for Bloomberg News. He started his journalism career at United Press International, working as a reporter and editor in California, Texas, Kansas and Florida. Have a news tip? Please drop him an email.