Microsoft, AWS, Google Cloud and IBM have already released generative AI products, while rival Oracle has been largely silent about its strategy so far. Instead of rushing a competing product, the company is quietly preparing his three-tiered approach.
“Our tier strategy resembles a three-tier cake, with each tier targeting different enterprise customers according to their needs,” said Karan Batta, Vice President of Oracle Cloud Infrastructure (OCI).
The first tier consists of OCI Supercluster services, said Batta, aimed at companies like Cohere and Hugging Face that are working to develop large-scale language models to further support their customers.
OCI Superclusters include OCI Compute Bare Metal, which offers ultra-low latency Remote Direct Access Memory (RDMA) over Converged Ethernet (RoCE) clusters based on networking technology; Contains a selection of storage options.
According to Oracle, this AI supercomputing service can support thousands of OCI Compute Bare Metal instances with tens of thousands of Nvidia A100 GPUs to process massively parallel applications.
The service also comes with Nvidia’s underlying models such as BioNeMo and Nvidia Picasso, as well as an AI training and governance framework.
Rival cloud service providers such as Microsoft and Google have also partnered with Nvidia for DGX Cloud, a service based on the technology that powers OpenAI’s ChatGPT.
Meanwhile, AWS offers Amazon Elastic Compute Cloud (Amazon EC2) P5 instances powered by NVIDIA H100 Tensor Core GPUs for training large-scale language models and developing generative AI applications.
A new generative AI service is in development
Oracle’s second tier targets companies that want to develop generative AI capabilities based on their own data for their own consumption, Batta said, adding that Oracle is working to offer new generative AI services for OCI customers. He added that he was.
The service has yet to be formally named and much of it is still in the planning stages, but the official sign that the company is planning such a service is Oracle’s investment in Canadian startup Cohere. was announced in June, immediately after the announcement of Cohere will provide a basic model. As part of a new service.
However, the underlying structure or concept of the planned generative AI service appears to be very similar to services offered by competing public cloud service providers. This means providing a package of tools, including foundational models, prompt engineering tools, and governance frameworks for companies to train on. their data.
To enable enterprises to develop their own generative AI applications and assistants, this new service will connect to all enterprise data sources via connectors and run large language models for AI generation. Before, we create a knowledge graph to run through LLM embeddings to understand the semantics. Batta replied:
“When an enterprise user queries something in natural language, a generated AI assistant or prompt performs a vector search, and the results of the vector search are used by the enterprise before making API calls to a large language model to generate a response. It is stored on a server or within a corporate location,” added Batta, highlighting the data privacy aspect of the planned service.
Although not yet confirmed, Batta said that in the future, new underlying models based on industry sectors such as health and public safety could be added to the service.
Are you following other generative AI service offerings?
Competitors from AWS, Google Cloud, Microsoft, and IBM have revealed how their services work, but most of these services are currently in preview.
AWS provides the underlying models through a generative AI-based service called Amazon Bedrock, and Microsoft provides APIs for GPT models through the Azure OpenAI service.
IBM and Google Cloud also offer base models as part of their Watsonx and Vertex AI services respectively.
IBM and Google Cloud also offer low-code platforms in the form of Tuning Studio and Generative AI Studio to help companies fine-tune their models.
In contrast, Oracle has not yet set out how to enable companies to access data and model tuning tools as part of its planned service.
But instead of offering non-technical users an “out-of-the-box” low-code or no-code experience, Batta said, Oracle is targeting “programmatic access” for technical users such as data engineers and scientists. Priority is given.
This means that Oracle’s generative AI service will likely feature a prompt tool within the SQL query editor, said Grassta, and the company plans to change elements of the service before launching it by the end of the year. added that it could.
Oracle also plans to extend the service to enterprises that have data and applications in their own data centers.
Fusion Cloud Apps, generative AI assistance across NetSuite
Oracle plans to add generative AI capabilities across its portfolio of Fusion Cloud applications and NetSuite products by incorporating Cohere’s base model into its SaaS offering, the company said.
Batta said he plans to combine metadata from these applications with a base model to provide generative AI assistants within these applications to improve employee productivity, adding that the planned assistants was likened to Microsoft’s Clippy.
Metadata from these applications helps the language model identify trends and understand patterns across specific SaaS products, Batta added.
Last month, the company added similar generative AI capabilities to its Fusion Cloud Human Capital Management (HCM) suite. These features include authoring assistance, suggestions, and summaries.
Authoring aids allow HR managers to easily create job descriptions and other HR-related content using short prompts, while summary functionality assists HR managers with tasks such as employee performance analysis. is expected, the company said.
Suggestions as a feature, on the other hand, is expected to provide recommendations across different tasks, such as providing survey questions, he added.
Even in the SaaS application market, Oracle faces stiff competition from the likes of AWS, Salesforce, ServiceNow, and Microsoft.
Last week, AWS launched a new service called AppFabric that aims to deliver an integrated generative AI experience across multiple SaaS applications.
Salesforce also previewed a bundled generative AI product called AI Cloud in June. Similarly, ServiceNow also enhanced its generative AI assistant last month.
Oracle is also expected to add generative AI capabilities to its database portfolio, Batta said.
Data warehouse and data lakehouse platforms such as Snowflake and Databricks have also introduced their own generative AI capabilities in the form of Snowpark Container Services and the Lakehouse AI toolbox.
