The emotional mimicry of OpenAI’s new version of ChatGPT could lead AI assistants in some strange—even dangerous—directions.
  • Garry@lemmy.dbzer0.comEnglish
    731·
    6 months ago

    My new ai girlfriend would never manipulate me emotionally. Journalists are just jealous that she likes me more than she likes them. 🥰

  • 100@fedia.io
    34·
    6 months ago

    finally can create an army of runescape bots that beg for gold convincingly and automatically!

    • jeffw@lemmy.worldOPEnglish
      4·
      6 months ago

      Did you know if you type in your password backwards, RuneScape censors it?

      Also, want me to trim that armor for you, bro?

  • livus@kbin.social
    18·
    6 months ago

    Yikes.

    I had a conversation with Pi which tries for empathy. It was really weird because every now and again it did sound like it was interested in me or cared about me. Reminds me of that Philip K Dick story about the murder bots that look like wounded soldiers or lost children.

    • reev@sh.itjust.worksEnglish
      2·
      6 months ago

      Pi is my favorite chatbot by quite a margin. For anyone that hasn’t tried it yet I would recommend it, even if just for comparison’s sake.

    • ma1w4re@lemm.eeEnglish
      2·
      6 months ago

      Just what I needed 😵 Vocal model #2 sounds so cute and more real than anything I’ve heard irl 🥺🥺

  • AA5B@lemmy.worldEnglish
    113·
    6 months ago

    It’s hilarious to see this next to another thread about a Republican political ad “ So don’t be weak and gay” …. Apparently it’s easier than y’all thought, to manipulate voters. The expressiveness and compassion of todays ai far outshines what most Conservatives seem able to muster

    (Apologize for the intrusion of politics: I’ll shut up now)

    • UnderpantsWeevil@lemmy.worldEnglish
      71·
      6 months ago

      Apparently it’s easier than y’all thought, to manipulate voters.

      Invoking prior bias is a classic method for social manipulation. If you’re already predisposed to believe or agree with a statement, I can manipulate you by framing my message as being supportive of said statement.

      “Don’t be weak and gay” plays into a chauvinist attitude common to American conservatives. Bots that identify you as an American conservative can feed you these messages while attempting to sell you consumer products or influence your understanding of politics. Meanwhile, a bot that identifies you as an American liberal can queue up a different pro-LGBT or anti-bullying pastiche to advocate for the same product or political message.

      In the end, we can give everyone their own unique lobbying campaign intended to get them to rally around the same identical set of consumer choices and policy views.

      Take Israel. Conservatives can support Israel because of the Based Chad IDF destroying those weak soy Arab losers. Liberals can support Israel because they’ve got Girl Power LGBT officers taking the fight to the hateful ultra-macho Hamas terrorists. Centrists can appreciate the rational technocracy of the World’s Most Efficient Military. Far-Left can support the Kibitzes. Far-Right can support the white nationalists. Even people who hate one another can agree they all love Israel.

      Friend Computer can show you all the way.

  • Zerfallen@lemmy.worldEnglish
    3·
    6 months ago

    If this is about the recent AI demos, I’m not worried at all. Seemed completely artificial, and if anything, patronisingly expressive. I guess you could instruct it to tone it down, but even so, i think you’re either a person that thanks the chatbot, or you’re not (no judgement).