How to use Retrieval Augmented Generation (RAG)

preview_player
Показать описание

What is Retrieval Augmented Generation (RAG) and how does it enhance generative AI capabilities in apps? Watch along as Googlers Aja Hammerly and Jason Davenport discuss this architectural approach to building AI applications including its components, benefits, and steps involved in a basic RAG workflow.

Chapters:
0:00 - Intro
0:22 - What is RAG?
1:06 - RAG dataflow
4:32 - Improving quality of RAG systems
6:05 - Conclusion

#GoogleCloud #GenerativeAI

Speakers: Aja Hammerly, Jason Davenport
Products Mentioned: Cloud - AI and Machine Learning - Gemini
Рекомендации по теме
Комментарии
Автор

Love this video style and information. We should have more content like this!

RobertoSilva-mv
Автор

I always thought RAG was more complex and changed the behaviour of the model somehow, thanks for this high level view

lostpianist
Автор

Great video! Do you have more resources to delve into RAG and fine-tuning LLM models? Thank you!

ivanrodriguezc
Автор

It seems to me that the value of a RAG is the ability to work on private data that needs to remain private. I am concerned about the Laissez-faire attitude with the data in these videos. It is great, but make a video where you might label it as "How to use Retrieval Augmented Generation with private data and how to keep it private".

grokism
Автор

When the vector DB in a RAG model fails to retrieve matching context, how does the LLM handle the response? Does it rely entirely on its pretrained knowledge from the prompt? Should the LLM Response Without Context or Fallback Policies or Hybrid Retrieval

SathyaPrakashMC
Автор

Didn’t know Dexter works at google now.

vee_grilla
Автор

The world have moved on from vanilla RAG a long time ago

kevinhe
visit shbcf.ru