The copyright infringement comes about later, when that mass of numbers is used to produce a topically related work. The same rules apply for humans -- see the concept of "clean room implementation".

It really doesn't. Prose isn't source code. Learning something, then later writing something else isn't copying.

My limited understanding of case law is that transformative use is still judged very human-centricly.

E.g. the courts take a dim view of any attempt to create a machine (in the abstract sense) that takes in copywritten works and churns out similar-but-uncopywritten works