Oracle on Tuesday revealed updates to its OCI Generative AI service and introduced new services for enterprises looking to quickly make use of generative AI.
First introduced in September, Oracle Cloud Infrastructure Generative AI is a managed service, available via API, to integrate large language models into different use cases. The service is now generally available, with access to Cohere LLMs and Llama 2 70 billion parameter models.
The Cohere LLMs include multilingual capabilities that support more than 100 languages. The Llama 2 models will be added in the coming weeks, according to Oracle.
OCI Generative AI Agents and Data Science
Oracle also introduced OCI Generative AI Agents in beta, with the retrieval-augmented generation agent the only agent currently available. It enables users to pull information from enterprise data sources using natural language.
OCI Generative AI Agents is built on OCI OpenSearch. OpenSearch is a managed open source service that lets users ingest, search and analyze their data. With Generative AI Agents, customers can either ingest data in OpenSearch or use the OpenSearch repository.
Also, a self-check capability enables an agent to provide references and citations about where it got the information it provides to the user.
Future releases of Generative AI Agents will support other data search and aggregation tools, and access to Oracle Database 23c with AI Vector Search and MySQL HeatWave with vector store, according to Oracle.
Oracle also plans to introduce prebuilt agent actions across its SaaS applications, including Oracle Fusion, Cloud Applications Suite, Oracle NetSuite and the Oracle Cerner electronic health record platform, the vendor said.
Sid NagAnalyst, Gartner
Finally, Oracle revealed that OCI Data Science AI Quick Actions will launch in beta next month. The service gives customers access to a selection of open source LLMs from Meta and Mistral AI.
Making GenAI applicable
With the introduction of these new services, Oracle is looking to make generative AI applicable to enterprises, said Vinod Mamtani, the vendor's vice president of generative AI services.
"GenAI models are highly capable, but for us, it's about focusing on enterprise use cases and solving for it," Mamtani said in an interview.
However, both the OCI Generative AI service and OCI Generative AI Agents are similar to services offered by other AI vendors.
"This is all about catching up with the other hyperscalers, including the capability of leveraging and embedding GenAI across its applications portfolio, making these applications more intelligent," Gartner analyst Sid Nag said.
The OCI Generative AI service appears comparable to AWS Bedrock because it is an API capability that offers LLMs from Cohere and Llama, he added.
However, what makes Oracle's announcement notable is the tech giant's investment into infusing generative AI technology into its SaaS applications, according to Ventana Research analyst David Menninger.
"Where GenAI is going to have the first round of impact is going to be in applications," Menninger said.
By choosing to incorporate generative AI into its application portfolio, including its Fusion application portfolio, Oracle is focusing on where customers will get the most value, he added.
"Organizations are going to get more immediate value from the extensions they've put into the Fusion applications than they're going to get from building some sort of custom GenAI capability," Menninger said.
While custom applications are important, big tech vendors such as Microsoft already provide them to customers. However, by incorporating generative AI into applications for areas such as contact centers, consumer experience, and item descriptions or job postings, Oracle can help make the people in those roles more productive, according to Menninger.
And Oracle's approach in incorporating these applications is still in its early stages, he continued. For example, the OCI Generative AI Agents suite only has one agent available.
"More agents will be helpful in the same way that these application capabilities are helpful," Menninger said.
Moreover, some enterprise customers might not want to work only with Meta or Cohere. But Oracle's slow yet gradual rollout approach shows that cloud providers can't always do everything at the same time despite the fast pace of the market.
"I'm sure if they had unlimited resources, they'd love to bring everything out at once," Menninger said. "But that's never a reality."
Esther Ajao is a TechTarget Editorial news writer covering artificial intelligence software and systems.