Your experiment is contaminated: If a piece of training document said that AI texts are overly verbose, and then announced that the following is a piece of AI-written text, it’d be a natural guess that the document would continue with overly verbose text, and so that’s what an autocomplete engine will generate.
Due to RLHF, AI is no longer cleanly modelled as an autocomplete engine, but the point stands. For science, you could try having AI assist in the writing of an article making the opposite claim :).
I have not tried asking GPT-4o to write concisely, but then, the “writers” of the articles I have in mind clearly haven’t. There are a few within the last few days. I’m sure people can guess which they are.
In my experience, they talk like that regardless of the claim being made unless I specifically prompt for a different writing style (which has mixed success).
Your experiment is contaminated: If a piece of training document said that AI texts are overly verbose, and then announced that the following is a piece of AI-written text, it’d be a natural guess that the document would continue with overly verbose text, and so that’s what an autocomplete engine will generate.
Due to RLHF, AI is no longer cleanly modelled as an autocomplete engine, but the point stands. For science, you could try having AI assist in the writing of an article making the opposite claim :).
I did that once.
I have not tried asking GPT-4o to write concisely, but then, the “writers” of the articles I have in mind clearly haven’t. There are a few within the last few days. I’m sure people can guess which they are.
In my experience, they talk like that regardless of the claim being made unless I specifically prompt for a different writing style (which has mixed success).