Retrieval-augmented generation (RAG) is becoming a critical focus in company discussions about generative artificial intelligence (GenAI), as organisations look to leverage real-time external data to contextualise and refine the accuracy of generated responses.
Companies are focused on utilizing this technology to enhance user experience that allows customers to integrate the real-time proprietary data sources with general-purpose AI models, reveals the Company Filings Analytics Database of GlobalData.
Misa Singh, business fundamentals analyst at GlobalData, comments: “While many companies are still concentrating on GenAI, others have already started seeing RAG as an opportunity. The discussions mainly revolved around driving better results, minimising AI hallucinations and providing technical support to products.
“Companies are also collaborating to integrate their offerings to build RAG pipelines and answer questions more promptly.”
Amazon.com has built Amazon Bedrock, which along with the broadest selection of large language models (LLMs) also has RAG to expand model’s knowledge base. This allows to safeguard what questions applications will answer and agents to complete multistep tasks.
Progress Software discussed about its new Progress Data Platform, which is the integration of MarkLogic and Progress Technologies. This platform features both semantic and vector capabilities to power RAG in AI application. The platform provides contextual links to corporate information and knowledge within the responses. Users can see where the answers came from, what they mean, and how they are relevant to the business. This drives dramatically better results of GenAI powered responses and minimizes AI hallucinations.
Innovation Technology Group mentioned focusing on introducing and polishing the RAG and Agent engineering frameworks to provide strong technical support for the iterative upgrades of ChatDoc, ChatRobot Pro and other products. Pure Storage Inc revealed in its earnings call that it sees RAG as an opportunity.
Confluent has partnered with Pinecone for its vector database architecture “Pinecore serverless”. This integration allows customers to build RAG pipelines that will allow customers to bring together the real-time state of their proprietary data sources with general purpose AI models.