Summary
In this chapter, we greatly expanded the data that’s available to our AI models by using the RAG methodology. Besides allowing AI models to use large amounts of data when building prompts, the RAG methodology also improves the accuracy of the model: since the prompt contains a lot of the data that’s required to generate the answer, models tend to hallucinate less.
RAG also allows AI to provide references to the material it used to generate a response. Many real-world use cases require models to manipulate large quantities of data, require references to be provided, and are sensitive to hallucinations. RAG can help overcome these issues easily.
In the next chapter, we will change gears and learn how to integrate a Semantic Kernel application with ChatGPT, making it available to hundreds of millions of users. In our example, we will use the application we built in Chapter 5 for home automation, but you can use the same techniques to do that with your own applications...
