I finally figured out why LLM replies give me such a high "uncanny valley" vibe.

It's the usage of expressions as "conversational flavour enhancers" that simply don't make sense coming from a machine.

From elements that are supposed to increase your trust in the reply like "that's a great question", to expressing feelings that the LLM is incapable of having: "it's fascinating how", or "it's crucial to", ending in both-sideism "a single XX can be Y and Z at the same time".

Basically LLM companies are trying to manipulate their customers with NLP tactics so they won't stop using their products...