Natural language and nature
It’s funny: modern large language models (LLMs) are supposed to be superb at natural language. They are literally the most comprehensive model of it.
Yet they fail miserably when you ask them to write text. I mean they do write it, they can even make it factually correct, with references and all that.
But that ethereal human touch just isn’t there. Pretty much never. I guess you could say the texts produced by them are not natural.
I still quite cannot pinpoint exactly what’s wrong sometimes, but it just is. Even when it’s “perfect”.
Maybe it’s something related to the nature of how our mind works. Or maybe how the nature itself works.
Fascinating.