Comment by volkercraig

More than that, the entire structure of the study is pointless. They set up as a question/response and then had humans rate the response. That's literally what LLM's are trained to do, which ultimately is convincing a human to click the "I like this one better" button on it's response.