• Awoo [she/her]@hexbear.net
    link
    fedilink
    English
    arrow-up
    52
    ·
    8 months ago

    Altman also said that OpenAI’s chatbot should be capable of reflecting back the worldview that its users want.

    “I think our product should have a fairly center-of-the-road, middle stance, and then you should be able to push it pretty far,” he said. “If you’re like, ‘I want you to be super woke’ — it should be super woke.”

    This means the same thing in the opposite direction. It also means that the model will be explicitly designed not to give “truth” (based on the majority of its training data) but instead simply to tell its users what they want to hear. By telling users what they want to hear it will become pipeline tool, leading people deeper and deeper into radicalisation over time.

    • Comrade_Bones [he/him]@hexbear.net
      link
      fedilink
      English
      arrow-up
      20
      ·
      8 months ago

      “I think our product should have a fairly center-of-the-road, middle stance, and then you should be able to push it pretty far,” he said. “If you’re like, ‘I want you to be super woke’ — it should be super woke.”

      Weird that he didn’t offer up the counter-example, “If you’re like, ‘I want you to be super racist’ — it should be super racist.”

      • Awoo [she/her]@hexbear.net
        link
        fedilink
        English
        arrow-up
        10
        ·
        8 months ago

        I suspect “super woke” is actually code for “communist”. I’m certain that one of the things they’ve noticed on the backend is communists deeply criticising the shit out of the ideology within the AI and then flagging the conversations as negative. These conversations getting flagged as negative turn up on their internal reviews because one of the internal processes will be “Is this flag for incorrect/biased information accurate?” and upon a human reviewer reading the conversation and finding the communist to be accurately calling out the bot for its inaccuracies and historical revisionism the reports get escalated as “false information”.

        The fascists that go through the same process don’t get escalated because when a human reviewer goes over the report claiming false information/inaccuracies the human reviewer looks at the conversation and flags the report as baseless.

        This is usually then followed up by the communist sending something like “everyone involved in this project should be in prison” or some more radical shit.

        This is repeating internationally in multiple languages.

          • Awoo [she/her]@hexbear.net
            link
            fedilink
            English
            arrow-up
            5
            ·
            8 months ago

            i live in a country where 34% of the population will vote for socialism and a significant majority of those people will be able to accurately define socialism in marxist terms. Similar numbers play out in the rest of europe, with some being even better.

            Respectfully, I think Americans underestimate how many leftists are screaming at this thing because the democrats have formed such a strong buffer against leftist growth.

        • Damarcusart [he/him, comrade/them]@hexbear.net
          link
          fedilink
          English
          arrow-up
          5
          ·
          8 months ago

          I suspect “super woke” is actually code for “communist”.

          I think it is more likely that “super woke” just means “not a shithead.” It’ll take the “middle of the road” approach in a US way, so extremely right wing, but you have a choice of right wing with some bandaids covering the worst parts, or full on mask off nazi.