From the course: Implementing a Data Strategy for Responsible AI
Unlock this course with a free trial
Join today to access over 25,200 courses taught by industry experts.
Retrieval augmented generation
From the course: Implementing a Data Strategy for Responsible AI
Retrieval augmented generation
- [Instructor] Retrieval-augmented generation combines what we've already done to create customized task specific pre-trained models, AKA LLMs, with information retrieval techniques. It overcomes one of the main shortcomings of traditional LLMs: not having enough current, relevant, high-quality data to update the LLM. RAG sets out to improve the LLM's response by accessing and integrating realtime, contextualized information from vetted data sources to the user request. Many times, these sources are internal company documents or knowledge bases that have been pre-processed, translated into text embeddings, and stored in vector databases. Using internal company knowledge bases has other perks. The new data contains domain-specific knowledge that helps mitigate the infiltration of malicious content. It also provides greater transparency and explainability, since you can more likely trace the origins of the data easier. The data input is more reliable and accurate. Therefore, your LLM's…