LLM’s are a tool. One that absolutely sucks at creative writing. After reading a page, it doesn’t take a genius to tell what was written by a human, what was written by an AI, and what was enhanced by one. The only times an AI work, enhanced or fully LLM, is better than human work is when comparing it to humans who are too lazy to use a grammar checker. Even someone with a partial understanding of English writes more interesting than any AI I have seen so far.
There is a reason why AI writing is called slop. At the base layer is a neural net that predicts the next letter/token, making it an advanced form of predictive text. Then above that is a layer that checks the lower level and sees if its response fits the prompt. Then there are various layers to make the writing more interesting, accurate, etcetera. However at the fundamental level it is just predictive text guessing at what the most likely next letter, word, etc. will be after being trained on an extremely large selection of text. It is an average word generator, or more accurately an average word generator targeted at the prompt, and with grammar and context recognition.