Login
You're viewing the mstdn.social public feed.

Replies

  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Jul 7, 2026, 9:55 PM

    @heluecht chatbot translation hallucinations are already going *wild*

    comment from youtube:

    > I am not surprised. A while ago I wrote a mini-essay on a technical forum. Because there were native Spanish speakers in the thread, and I still speak passable Spanish from schooling, I got curious about how my essay would read when translated. I noticed the translation taking a really long time, way longer than usual, and thought, "This is ai now, isn't it? Isn't it!!??"

    > So my essay came back, and the first two short paragraphs were ok. But then the llm started rolling some kind of obnoxious snowball of intent down the hill and extrapolated all this stupid crap I wasn't even saying. It devolved all my claims, thoughts and wishes into exaggerated hyperbole, opinions became "demands" and everything was "I firmly and vehemently declare X, and anything less will be unacceptable." Eventually just straight up adding stuff I didn't say, just a big strung together flowery flower garland of poisonous ai slop. Yeah, I thought it was pretty distasteful.

    💬 1🔄 5⭐ 0
  • Jul 7, 2026, 10:22 PM

    @davidgerard Seems to me it has become a Monty Python sketch that starts out mostly sane and then descends further and further into a surreal mess until Graham Chapman appears dressed as a British army officer and yells "Stop it! Too silly!" Only without the funny bits.

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0