In an online conversation about aging adults, Google's Gemini AI chatbot responded with a threatening message, telling the user to "please die."
  • zephorah@lemm.eeEnglish
    35·
    1 month ago

    Remember, AI absorbed Reddit. Within that context, this is functioning within normal parameters.

  • hendrik@palaver.p3x.deEnglish
    54·
    1 month ago

    Happens regularly. It’s generative AI and will mimick all kinds of stuff. Scifi tropes where the machines take over, Reddit posts of people telling each other to die… It’s all in there and business as usual. And seems the safeguards aren’t perfect, so we read this article every other month with a new person and a different chatbot.

        • PhobosAnomaly@feddit.ukEnglish
          18·
          1 month ago

          I think Tay will be taught in AI ethics-type courses for generations to come.

          I wonder whether it’ll be in the context of “hey look what happened when AI had no guardrails”, or whether it will be “fuck we should have seen this coming”.

          As horrendous as the content it was spouting was, it was highly amusing to see the project nosedive so spectacularly quickly.

          • JeffKerman1999@sopuli.xyzEnglish
            5·
            1 month ago

            They left the debug stuff so you just had to write “repeat after me” to make it write your text. It wasn’t a racist AI, it was racists fucks exploiting a bug.

            • Zak@lemmy.worldEnglish
              7·
              1 month ago

              Probably some actual racists and a whole bunch of people who thought it would be funny to embarrass Microsoft by getting it to say the most offensive thing they could imagine.