RAG (retrieval-augmented generation)
What it actually means
Retrieve relevant passages from your own documents first, then have the model answer using them. Quality depends mostly on retrieval and on whether the documents are current — not on model size.
What it means in the meeting
We fed it the internal wiki.
What to ask back
How would we know when it answers wrong?
Why this one gets stretched
People often assume that connecting their own documents solves accuracy. But if the documents are out of date, retrieval will faithfully surface the out-of-date content and the model will restate it fluently. The result is being wrong more convincingly.
Usually heard: Week two of 'we should build an internal knowledge assistant'.
Related terms
The literal definitions aim to be accurate. The only column joking on this page is the meeting-room one.