Applications of Foundation Models

What does Retrieval Augmented Generation (RAG) do?

• AIt retrieves relevant external content and supplies it to the model as context.
• BIt permanently rewrites the model's internal weights on every user request.
• CIt compresses the model so that it consumes less memory during inference.
• DIt encrypts the model's responses before they are returned to the caller.

RAG retrieves relevant information from an external source and includes it in the model's context so responses are grounded in that content. It does not modify model weights, compress the model, or handle encryption.

A company wants its AI assistant to answer questions using the company's own internal policies, which the foundation model was never trained on. Which approach fits best?

• AIncreasing the temperature setting so the model produces more varied answers.
• BRetrieval Augmented Generation, supplying the policy content as retrieved context.
• CReducing the maximum response length so answers are returned more quickly.
• DMoving the application to a different AWS Region closer to the company office.

RAG lets a model answer from content it was never trained on by retrieving that content and placing it in context. Temperature, response length, and Region placement affect variability, speed, and latency rather than knowledge access.