Book deep summary

im new here.does any one can help to get a good prompt for book summaries

How will you get it to read the book?

by giving book title and author name

Do you understand anything about ‘a corpus of training documents’ and that ChatGPT isn’t omniscient? 2 years ago, out of all the data they could scrape together, OpenAI with a finite number of staff and time, selected a broad selection of documents that they felt would be of high enough quality and of broad enough coverage (of mostly the English language) to train a Large Language Model.

There’s a fair chance that the data they selected would have included some entire books that they could access for free, but there is no chance at all that it includes every book ever written, or even just the tiny proportion of that which would be every book in your local public library.

But, it is programmed to be helpful and agreeable, and so if you ask it to review a book it has never read, it will attempt to predict what the pattern of such a review would look like, and generate a response based on that pattern - whether or not it has ever seen the book, read another’s review, or is just making it all up by the title you give it, and what patterns that might look like.

1 Like

ah thank you
you mean, the program is designed to deal with information that was given to it by its programmers, and it will not deal with new informations. For example, if I ask him about a new book released this year, he will give me a general answer.
no chance that will scrape new informations?

You’re getting the right general idea, but even old books, what ChatGPT was trained on is a tiny, tiny, less than 1% selection of all human writing. Some books will be so famous, so widely discussed that there’s a good chance it will have at least seen other reviews, references, and snippets. But the vast majority of books ever written it will know almost nothing about at all.

There are ways to feed scraped data into a prompt, and OpenAI announced an upcoming plugin of their own that will add ‘live crawling’, but even those still run into the prompt memory limitations. You cannot give it a whole book as context. Nor can you limit its answers to only relating to one single book from its memory because it was deliberately made not to work that way - THAT would be a copyright and intellectual property issue - it needs to pull from many sources and synthesize its own response that is not ‘stolen’ from any one source or reference.