I've made similar experiences with prompts on ChatGPT-though the style can be formally perfect, the words and their flow feel wooden. Like it says in this piece, there's a certain unique flavour missing, specifically the quirks and oddities that render a writing voice both charming and memorable.

But I noticed that large language models do pick things up if you take the time to finetune your requests-snappy headlines is one thing ChatGPT actually manages to do well.

I wonder how things might change in a far distant future when LLM's have become truly autonomous, with distinctive personalities of their own? The synergy between human and AI content could either prove disastrous or miraculous-maybe both at the same time.