Retrieval augmented generation
RAG is an architecture for optimizing the performance of an artificial intelligence (AI) model by connecting it with external knowledge bases. RAG helps large language models (LLMs) deliver more relevant responses at a higher quality.
Generative AI models while trained on large datasets are ultimately limited by the very knowledge they posses.
- Datasets can become outdated, when new conflicting information becomes available
- Some datasets a user might find relevant may not be public or meant for internal use
- The model can often benefit having access to real-time statistics to ensure less bias
All this and more can be made possible by using RAG.
https://www.youtube.com/watch?v=T-D1OfcDW1M
Benefits of using RAG
- More efficient AI implementation
- Access to domain-specific data
- Greater data security and privacy
- Lower risk of AI hallucinations
Advanced
RAG Evaluation
Chunking Strategies