[sinuhe69]

I wonder if one day we will have an AI that reads, summarizes and catalogues all the published books? A super librarian :) Imagine being able to ask questions like: "What have they written about AI in the 21st century?". Even better: "What did people not think of when they pursued AGI in the 21st century, which later led to their extinction?" ;)

[pillefitz]

Since most foundational models have been trained on illegally acquired books, this info should be already baked in.

[sinuhe69]

They had access only to just one tiny bit of the entire written words of the worlds. Not all books are available in electronic formats. You can see from the visualization in the article that we don’t even have the titles for a lot of published books with ISBN. And even so, books with ISBN comprise only a fraction of the entire ever written books. Not too mention books in various (minor) languages of the worlds.