Video / Short
Why Your AI Keeps Getting It Wrong
Retrieval-augmented generation connects AI to current company knowledge so its answers can be more accurate without retraining the model.
Video summary
Pete Cooper explains retrieval-augmented generation, or RAG, as a way to connect a language model to a selected body of company information. He describes the retrieval step as collecting relevant documents and the generative step as the model producing an answer from that material.
The intended benefit is more accurate, useful answers without retraining the underlying model. Cooper presents RAG as a common way for organisations to make a large volume of knowledge easier to access. Its reliability still depends on the quality, currency and scope of the documents provided to the system.
Video transcript
Read the spoken content without loading the YouTube player.
What is RAG? Well, it stands for retrieval-augmented generation. It's essentially a technique that allows you to use the power of an LLM, like ChatGPT, but have it access only the data that you put aside for it. The retrieval augmentation is the part where you put all the documents into one place, and the generative part is the large language model. It could be ChatGPT or whatever you want to use. It's a commonly used technique for companies to give easy access to a large volume of data.