Skip to main content

What is the good way to describe what RAG does to non-technical SMB customers?

Hi Vitaliy I’m a relatively non-technical person myself, so I’ll answer in my own words. Think about an LLM that you may have used before… maybe ChatGPT, Gemini, or something. When you ask it a question, its been trained on a massive amount of data; basically like the entire internet. So it will return an answer to you that it thinks its best according to how its been trained, referencing the entire internet. That is an over-simplification of course, but you get the idea. With RAG, I can say, “Hey LLM Model, I only want you to look at this specific document or website when you give me an answer, not the entire internet.”

Here is an example. Lets say that you’re a company that makes a fancy coffee machine. You want to create a chatgpt style chatbot for customers or internal employees quickly get troubleshooting information on your coffee maker because it breaks all the time (like mine does). If a customer went to chatGPT they might get fed information about other coffee makers… that doesn’t help. But if you use RAG, you can have your chatbot only reference your own technical documentation and the end user will only get served relevant information and hopefully better coffee.

Here is a good video that IBM put out a couple years ago that is really easy to follow. Hope this helps! 

 


Reply