• wizardbeard@lemmy.dbzer0.com
    link
    fedilink
    English
    arrow-up
    28
    arrow-down
    1
    ·
    6 days ago

    What? They aren’t sentient. They aren’t sociopathic. They can’t do “use of deception”.

    These are (absurdly complex) word/phrase/semantic token association engines. They aren’t thinking. They don’t understand true or false.

    They’re dangerous and not to be trusted for factual information,but not for anything related to asimov’s law, or whatever you seem to think is going on under the hood.

    • Die4Ever@retrolemmy.com
      link
      fedilink
      English
      arrow-up
      3
      ·
      edit-2
      6 days ago

      yea maybe the prompt was like “make some graphs that show how much better GPT5 is, here’s the data:” and attached a csv file

      then it went too far with trying to follow the instruction of showing GPT5 as better