Comment by walterbell

1 day ago

Did you find something incorrect in the two paragraphs that you read? Here they are:

  During my recent journey to Berlin—a city where every street corner whispers secrets of a divided past—I uncovered the chilling reality of "Zersetzung."

  This was not just another research assignment; it was an immersive exploration into the dark art of psychological warfare.

This is the personal website of a human being. The first two paragraphs described a human traveling to Berlin. Does ChatGPT travel often?

7 comments

walterbell

Reply
moritzwarhier  15 hours ago

I should have read your post in full before my commenting (it was late in the evening here).

Can't describe now what exactly I meant about the writing and normally I refrain from accusing LLM usage because of such gut feelings. So sorry.

Of course people have published LLM texts on their personal sites, but that doesn't mean you do.

So far I only skimmed that post, will re-read it later.

  • walterbell  14 hours ago

    Not my website, but the author does not appear to be fictional. If they are, it would be good to know how to figure that out.

    • moritzwarhier  14 hours ago

      I was exclusively talking about the writing style. Assuming a text would have been written by an LLM would of course not preclude it from being a rewording /summarization of real events or other people's writing.

      I see you posted another submission about this topic, and I can read that more easily, it sounds more natural to me.

      Make of that what you will, can't engange deeper rn, have to go back to work.

      1 reply →

actionfromafar  1 day ago

If you ask it nicely, yes.

  • walterbell  1 day ago

    Is the claim that the author is lying on their own web site about their travel, using ChatGPT to write fictional travel anecdotes?

    Would you care to provide any evidence to support the assertion that this personal website was written by an LLM?

    Data science project: index HN articles which trigger comments about AI/LLM/ChatGPT, with a combination of automated analysis and manual labeling to identify false positives, extract signal-to-noise ratio of HN comment detection of LLMs over time.