Anyone know if these LLaMA models can have a large pile of context fed in? Eg to have the "AI" act like ChatGPT with a specific knowledge base you feed in?
Ie imagine you feed in the last year of chatlogs of yours, and then ask the Assistant queries about the chatlogs. Compound that with your Wiki, itinerary, etc. Is this possible with LLaMA? Where might it fail in doing this?
(and yes, i know this is basically autocomplete on steroids. I'm still curious hah)