BestOfLemmy

Oh, that’s a very good point. I’ve already seen one LLM service that aims to summarize selected scientific sources and write articles based on them. You could feed it a bunch of historical texts and ask it to write an article based on them.

I’ve seen it sort of work with scientific articles, but those tend to be really short compared to historical records. AFAIK, even the normal GPT has some token limit, which you’ll run into during longer conversations. If you want to dump a bunch of diaries and logbooks on an LLM to answer a specific question about the motivations behind the development of flintlock rifles, you’re definitely going to run into some limitations.