Google Cloud · Generative AI Leader
Grounding, RAG, and sampling
Generative AI Leader · Lesson 11 of 14 · 1:29:27
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the Google Cloud Generative AI Leader exam. In this lesson: grounding and its data (first-party enterprise data, third-party data, world data), the steps of RAG and how they shape the output, Google Cloud's grounding offerings (prebuilt RAG with Agent Search, RAG Engine, grounding with Google Search), and the sampling settings: tokens, temperature, top-p, safety settings, and output length. Includes three original practice questions.
Preview the topics
- Cymbal Retail’s assistant grounds its answers about store policies in Cymbal’s own documents, and its answers about national holidays in public web information. Which kinds of data are these, in that order?
- After Cymbal Retail launches RAG, its assistant quotes an old return policy that still sits in the indexed documents. What explains this?
- Cymbal Retail’s marketing team wants more varied, creative slogan ideas each time it runs the same prompt. Which setting should it raise?