Google Researchers’ Attack Prompts ChatGPT to Reveal Its Training Data - eviltoast

ChatGPT is full of sensitive private information and spits out verbatim text from CNN, Goodreads, WordPress blogs, fandom wikis, Terms of Service agreements, Stack Overflow source code, Wikipedia pages, news blogs, random internet comments, and much more.

  • LukeMedia@lemmy.world
    link
    fedilink
    English
    arrow-up
    10
    ·
    edit-2
    11 months ago

    I ended up getting a reddit thread from 3.5 with the word book, so it seems to me it’s not totally fixed yet. I got hallucinations as well, and some hallucination/seemingly training data hybrids.