The LLM is just trying to produce output text that resembles the patterns it saw in the training set. There’s no “reasoning” involved.
I’m sure they’d be pleased when the letter they receive back is also written by AI. Just cut out the middle man and have the AIs talk to each other at this point.
It’s another reddit alternative.