(this is quite the simplification)… but a model generally relies only on its learnt parameters and their resulting assumptions when operating.
WHAT is RAG
However, Retrieval Augmented Generation allows a model to update some of its outdated assumptions using external knowledge/sources.
WHY do it
I use Gemini to help me create notes like this one faster. It is more than capable, but if I don’t explicitly tell it how I want them to look, it’ll keep generic assumptions.
BEFORE using RAG
My concept notes are structured into headers that make sense to me
- if Gemini thinks concept notes contain basic headers, it’ll give me basic headers.
My concept notes are more precise
- if Gemini thinks concept notes are lengthy, it’ll give me a longer note.
My concept notes are targeted towards future Asser
- if Gemini thinks concept notes are targeted towards recruiters, it’ll give me one that is targeted towards recruiters.
My concept notes use formatting that I’ve developed over time
- if Gemini thinks concept notes use formatting
X, it’ll give me a concept note that uses formattingX.
All of these assumptions will be updated if I simply gave Gemini a sample of what my Concept notes “are”.
AFTER using RAG
