| Google's NotebookLM is designed for this use case. Blog post: https://blog.google/technology/ai/notebooklm-google-ai/ (or try it at https://notebooklm.google/) I'm not sure if there's a length limit, but basically you upload a PDF and it'll try to "ground" the LLM in that. > A key difference between NotebookLM and traditional AI chatbots is that NotebookLM lets you “ground” the language model in your notes and sources. Source-grounding effectively creates a personalized AI that’s versed in the information relevant to you. Starting today, you can ground NotebookLM in specific Google Docs that you choose, and we’ll be adding additional formats soon. (It now supports PDFs and plaintext, from GDrive or elsewhere) ------------------ In my personal experiments, though, it didn't really seem to be very different than just uploading a file into ChatGPT (in a custom transformer). Sometimes one or the other would be higher-quality; both would still possibly hallucinate. Importantly, neither one is strictly limited to the source text; they both still have prior outside knowledge. What was cool, though, was NotebookLM's ability to cite passages in the source. Here is an example of your question: https://share.cleanshot.com/pSnz4CXC Their UI is a bit buggy, and sometimes questions just... disappear. That's why in the screenshot I asked it the same thing twice. But still, might be a "good enough" way to get the gist of a new source that's not already part of the training set. |