human vs LLM : text more precise or jumblish

use of AI for anything to iterate on human work to make it better always makes it jumblish, untill and unless the text is even written. If there is idea, LLM can make it worse and seems to increase the entropy of whatever it is.

Scholarly human intelligence on the other hand, with focus and understanding of the text, can make sentences more aligned (to the need, which is to other scholars), hence, can actually reduce the entropy. Perhaps due to immense amount of goal-oriented-efficient-thinking that its brain allows.

As with text, it does not only apply to text, but any piece of information. That is to say LLM excel at certain tasks and humans at certain tasks.