Hey folks,
I hope you are doing well today.
I’m not a dev and somewhat new to AI and HF. I have a book publishing business, and I want to train the AI in some of our books to help our team create social media content based on the books’s contents.
Would that be possible? Where to start?
You could check out RAG (Retrieval-Augmented Generation)
Here, a machine learning model (the AI) retrieves similar looking data as per your question from a vector database and returns an answer. This allows the model to pull relevant information from a large corpus (in this case, your books).
There’s usually a retriever model that retrieves the similar looking data based on the question from this vector database, then the generative models (like ChatGPT) generate a response based on the retrieved content.
Usually models have limits to context sizes (the amount of data you can tell it to learn or ingest in one go) within a single question you ask and follow-ups after it. So its better to use a RAG in your use case.