Contrary to your original comment there is lots of room for uncertainty, especially seen globally, in this area. While the copyright / fact extraction framing the author chose might not be the correct one in the years to come, this is certainly an area of volatility - at least here in the EU. Your comment w.r.t. "convenience of access" is pretty restricted here, no ML algorithm just boots up, grows legs, and magically walks into a public library to read books for fact finding. The process of extracting facts from books includes lots of steps that are subject to IP and other legislation over here (from crawling the content in the first place if your source is the internet, duplicating/storing the content if it is subject to copyright, ...).

While I'm not trying to suggest that your ultimate conclusion might not turn out to be correct, especially in the context of US legislation, that's far from settled and I expect to see quite a few lawsuits whenever this hypothetical actually affects a major player with some market power.