Re:Our next-generation model: Gemini 1.5
by resouer on Thursday February 15, 2024 @12:09PM
(#39385221)
> The 10M context ability wipes out most RAG stack complexity immediately.
This may not be true. My experience of the complexity of RAG lays in how to properly connect to various unstructured data sources and perform data transformation pipeline for large scale data set (which means GB, TB or even PB). It's also worth to note this is a critical path rather a "nice to have" helper, because the quality of data and this pipeline is the major factor of your final generated the result. i.e., in RAG, the importance of R >>> G.
This may not be true. My experience of the complexity of RAG lays in how to properly connect to various unstructured data sources and perform data transformation pipeline for large scale data set (which means GB, TB or even PB). It's also worth to note this is a critical path rather a "nice to have" helper, because the quality of data and this pipeline is the major factor of your final generated the result. i.e., in RAG, the importance of R >>> G.