OpenAI Is Giving Exactly the Same Copy-Pasted Response Every Time Time ChatGPT Is Linked to a Mental Health Crisis

Comments

• andallthat@lemmy.world ⁨8⁩ ⁨months⁩ ago

  ah, dear old copy/paste… It’s funny that even OpenAI doesn’t trust ChatGPT enough to give LLM-generated more personalized answers.

  source
• some_guy@lemmy.sdf.org ⁨8⁩ ⁨months⁩ ago

  This feels like a great time to recommend a song by a parody-hate band, S.O.D. Please understand that this band was formed by Scott Ian, of Anthrax, in the 80s. This was a time when you could mock hateful racists and people understood that it was a joke. I wouldn’t support a band saying that now, because I’d consider the excuse that it was a joke to be a front for their actual beliefs, as we’ve seen with people who are “just asking questions.”

  Kill Yourself.

  Bonus trivia: Scott Ian now plays with Mr Bungle. Just as S.O.D’s titular song was called Speak English or Die, Mr Bungle now plays a song called Habla Español O Muere (Speak Spanish or Die). If you can’t judge that the former was a parody by the evolution of the theme, I don’t know what to tell you.

  source

  • 4shtonButcher@discuss.tchncs.de ⁨8⁩ ⁨months⁩ ago

    Amazing read, thank you for this trivia comment

    source

    • some_guy@lemmy.sdf.org ⁨8⁩ ⁨months⁩ ago

      Thanks, I’m glad someone enjoyed it.

      source
• pewgar_seemsimandroid@lemmy.blahaj.zone ⁨8⁩ ⁨months⁩ ago

  i thought character ai, chai, dopple and the million other ones like them was more to blame for this

  source
• Electricd@lemmybefree.net ⁨8⁩ ⁨months⁩ ago

  LLMs can’t be fully controlled. They shouldn’t be held liable

  source

  • BananaIsABerry@lemmy.zip ⁨8⁩ ⁨months⁩ ago

    Perhaps we should also hold the rope, knife, and various chemical manufacturers responsible.

    The bridge architect? He designed a bridge that people jumped off of, so he’s at fault for sure.

    source
  • JakenVeina@midwest.social ⁨8⁩ ⁨months⁩ ago

    Well, yeah. The people who host them for profit should be held liable.

    source
  • Epzillon@lemmy.world ⁨8⁩ ⁨months⁩ ago

    “Ugrh guys, we dont know how this machine works so we should definetly install it in every corporation, home and device. If it kills someone we shouldnt be held liable for our product.”

    Not seeing the irony in this is beyond me. Is this a troll account?

    If you cant guarantee the safety of a product, limit or restrict its use cases or provide safety guidelines or regulations you should not sell the product. It is completely fair to blame the product and the ones who sell/manifacture it.

    source

    • Electricd@lemmybefree.net ⁨8⁩ ⁨months⁩ ago

      Safety guidelines are regularly given

      If people purchase a knife and behave badly with it, it’s on them

      Something writing things isn’t comparable to a machine that could kill you. In the end, it’s always up to the person doing the things

      I still wonder how ClosedOpenAI forcefully installed ChatGPT in this person’s home. Or how it is installed because they don’t have software… Quit your bullshit

      source

      • -> View More Comments
  • scratchee@feddit.uk ⁨8⁩ ⁨months⁩ ago

    Neither can humans, ergo nobody should ever be held liable for anything.

    Civilisation is a sham, QED.

    source

    • Electricd@lemmybefree.net ⁨8⁩ ⁨months⁩ ago

      Glad to hear you are an LLM

      source

      • -> View More Comments
  • surewhynotlem@lemmy.world ⁨8⁩ ⁨months⁩ ago

    I made this car with a random number generator that occasionally blows it up. Its cheap so lots of people buy it. Totally not my fault that it blows up though. I mean yes, I designed it, and I know it occasionally explodes. But I can’t be sure when it will blow up so it’s not my fault.

    source

    • Electricd@lemmybefree.net ⁨8⁩ ⁨months⁩ ago

      Comparing an automated system saying something bad with a car exploding is really fucking dumb

      source

      • -> View More Comments
• Not_mikey@lemmy.dbzer0.com ⁨8⁩ ⁨months⁩ ago

  Ah, I was hoping this meant chatgpt would give canned responses, ie. “Seek help”, whenever it detected it was being used for mental health issues, which it should. But no it’s just open air flipping off anyone who asks why there chatbot pushed a person to suicide.

  source

  • interdimensionalmeme@lemmy.ml ⁨8⁩ ⁨months⁩ ago

    It should also refuse to answer, any medical question, any engineering question, any finance question, anything that requires the responsibility of an accredited member of the professional–managerial class, to read the question, read the answer, and then decide if the question is allowed all and if he will re-write, change or simply block replies to the question entirely.

    Also such question should starting be 160$USD a pop

    source
  • WanderingThoughts@europe.pub ⁨8⁩ ⁨months⁩ ago

    Reminds me of the time that Twitter, when it was still Twitter and freshly taken over by Musk, replied to all slightly critical questions with a poop emoji.

    source

    • AngryRobot@lemmy.world ⁨8⁩ ⁨months⁩ ago

      As far as I know, emailing their public affairs address still does this.

      source