Maybe the point they're getting at is that LLMs are kind of too smart to be human any longer. A bit like how software drummers added a sliding "Humanize" parame

Maybe the point they're getting at is that LLMs are kind of too smart to be human any longer. A bit like how software drummers added a sliding "Humanize" parameter so that the drumming was "off" a bit. ChatGPT needs to confuse "loose" and "lose" in its output, mistake the U.S. state "Georgia" with the country.