Amazon Web Services (AWS) recently updated its Generative AI development service with new features to help businesses create applications more easily. These updates include contextual grounding, connectors for Recovery Augmented Generation (RAG), and additional Large Language Models (LLM).
According to Vasi Philomin, Amazon’s Vice President of Generative AI, contextual grounding is a new feature aimed at reducing unwarranted responses from LLM-backed applications. It uses RAG and summarization applications to verify the accuracy of the response based on business data and user queries. This is similar to systems used by Microsoft and Google to assess the reliability of RAG applications.
AWS has also introduced RAG observability and evaluation capabilities within Amazon Bedrock, offering metrics like fidelity, response relevance, and semantic similarity to compare query responses. Additionally, AWS is releasing Guardrails for Amazon Bedrock as a standalone API for use with LLM outside of Bedrock, following requests from enterprise customers.
IBM has a similar strategy with Watsonx.governance, which will be available through a toolkit in May. While not directly comparable to the Guardrails API, IBM’s toolkit allows for recording model details throughout its lifecycle for evaluation purposes.
In terms of new features, AWS has added RAG connectors to Amazon Bedrock for a wider variety of data sources, including Salesforce, Confluence, and SharePoint. The ability to adjust Haiku by Claude 3 within Bedrock, currently in preview, is unique to AWS and highly requested by enterprises.
Other updates from AWS include vector search for Amazon MemoryDB and new capabilities for Agents for Bedrock. Agents can now retain memory across interactions, interpreting code for data-driven use cases like data analysis and visualization.
AWS also introduced AWS App Studio, a managed service powered by generative AI that allows business users to build and deploy apps using natural language cues. This service eliminates the need to learn low-code tools, competing with offerings from Google and Microsoft.
Lastly, AWS announced the general availability of Amazon Q Apps, a feature within Amazon Q Business showcased at AWS re:Invent last year. Amazon Q Apps enable users to generate AI-powered apps instantly based on company data by describing the desired app in a message.
In summary, AWS’s updates to their Generative AI service aim to simplify the development of applications using AI technology, while introducing new features to improve accuracy, connectivity, and usability for businesses of all sizes.
Article Source
https://www.infoworld.com/article/2515588/amazon-bedrock-updated-with-contextual-grounding-rag-connectors.html/amp/