This talk by Ben Clavié is the highest value per second talk I have ever watched on RAG Chapter summaries and additional links in next tweet

More information (slides, YT link, etc) see this link:

If anyone is interested in the RAG course mentioned in this talk check out this form

@bclavie @bclavie is hands down one of the most talented people I’ve come across. His perspective altered my own path.

@bclavie Thats a super interesting talk! We have built n productionized our MVP RAG at a healthcare client a month ago and we came down to exact architecture you suggested after a lot of research :) I would also add Multi-query retriever as a pre-retrieval step as well into MVP

@bclavie I wish it was made clear in the course that much of it would be made freely available, kind of impacts the value for spend (not all of us are FAANG/SF sw engs)

@bclavie @memdotai mem it

@bclavie This is superb. Thanks much.

@bclavie Thx so much.

@bclavie RAG is now like doing image recognition with classical algos. All these tricks to make it more efficient in one or other benchmark without ever making it a complete technique. RAG needs to be iterative:

@bclavie Came here after watching the recording on zoom. I just gotta say, the talk was 🔥. Thank you @bclavie and hoping to see more! Hopefully slightly more advanced stuff/case studies building from the stopping point. (synethic data gen & fine tuning embs/colbert would be awesome)