LLMs break the internet (Changelog Interviews #534)
Changelog Master Feed
00:00
How to Train a Language Model on Your Own Documentation
Open source projects providing little LLMs alongside their docs or with their docs that are trained on their docs. Astro is one of these where they'll have a little train language model deal where you can chat with it about and just knows everything about the Astro docs. embeddings search which is a way of doing semantic searches so you can take the user's question and plotted in 1500 dimensional space against the previously plotted documents then find the stuff that's semantically closest to it but even if none of the keywords match it should still catch the documentation talking about the sort of higher level concept than talking about.
Transcript
Play full episode
