I’m actually curious to know how you’re hosting whisper Vs say just using the OpenAI api. I’m also wondering how offloading the transcription would to the client with something like https://github.com/xenova/whisper-web
That’s the old standby argument, and it may be right. I can’t really read John Barth or George Saunders the way I can read Richard Russo or Lionel Shriver or Kurt Vonnegut or Michael Chabon or Barbara Kingsolver. For me the experimental writers are very unpleasant to actually read. David Foster Wallace is just inside that frontier for me, and I can enjoy IJ. Bernard Malamud was pretty dark but I could hang in. But Paul Auster … I love what nonfiction writing I’ve seen, but the New York trilogy is so dark and Spartan it makes Joy Division look like disco.
Nitpick: I finally gave up on Pynchon, but is he really postmodern??
A lot of teams can do a lot with search with just LLMs in the loop on query and index side doing enrichment that used to be months-long projects. Even with smaller, self hosted models and fairly naive prompts you can turn a search string into a more structured query - and cache the hell out of it. Or classify documents into a taxonomy. All backed by boring old lexical or vector search engine. In fact I’d say if you’re NOT doing this you’re making a mistake.
Hi Zach, how do you think this architecture would perform for one longer document i.e. a novel of >50k words <100k? the queries would be about that one long document as opposed to multiple documents. any tips on how to approach my use case? thanks
Anyone remember how awesome This American Life was? They used to be about just that — an American [in their life] with an amazing story to tell. So simple and elegant. Then they went massively down hill around the time Trump was elected. They seemed to be “on the campaign trail” more often than not, or about some kind of grievance or “injustice”. I stopped listening.
Are there any best practices for doing RAG over, say, a novel? (50k-100k words) things that would make this unique compared, say, RAG over smaller docs or research papers:
- ability to return specific sentences/passages of a character while also keeping their arch in mind from beginning to end of story