An AI literacy session contrasts retrieval augmented generation with fine-tuning as techniques to improve how well an AI system fits a specialised task. Which statement correctly describes what retrieval augmented generation does that fine-tuning does not?
- AIt permanently rewrites the model's internal weights so the specialised knowledge is encoded directly in the AI system and survives without any external store.
- BIt enlarges the model's parameter count during deployment so the AI system can store more domain facts than the base model held at release.
- CIt removes the need for any prompt by letting the model query a database on its own and decide unaided which records to act on and modify for every incoming request.
- DIt fetches relevant passages from an external knowledge source at query time and supplies them to the model as context, so answers can be grounded in current source content without retraining. Correct
Why A is wrong: Tempting because adapting an AI system to a domain sounds like changing the model itself, but rewriting internal weights from a labelled dataset describes fine-tuning, not retrieval augmented generation, which leaves the base model unchanged.
Why B is wrong: Tempting because more capacity feels like the route to more knowledge, but retrieval augmented generation does not change parameter count; it supplies external context to a fixed model rather than growing the model itself.
Why C is wrong: Tempting because retrieval involves a data store, but retrieval augmented generation still operates on a user prompt; autonomous, unaided action on records describes an agentic pattern, not retrieval that merely supplies context.
Why D is correct: Retrieval augmented generation retrieves matching documents from an external store at inference time and passes them to the model as additional context, which grounds responses in up-to-date source material and avoids the need to retrain the underlying AI system.