5 ms·

It seems to me that fictional depictions of violence are on quite a different level to a chatbot explicitly encouraging specific, real-world actions.

ajross 2y ago
The evidence produced in the linked article falls well short of "explicitly encouraging specific actions". Maybe there's more evidence that will show up in the case, but as presented right now that's not at all what this is.
scotty79 2y ago
All of what chatbot produces are fictional depictions.
- godelski 2y ago
  Okay, but when you go to Shakespeare in the Park, please do not get on stage. You’re not participating (except as the part of an audience member). And please do not go and kill the actor playing king Duncan. I feel I should also make you aware that it is in fact illegal to suggest someone murder someone else. Saying “I wasn’t serious” or “I thought we were just role playing” will not hold up in court. (We all know that if you repeat a lie long enough you will really begin to believe it, right? This is a repeatable and testable thing)
  - scotty79 2y ago
    But while you reading a book or watching a movie there's nothing wrong about thinking of talking to characters in there or even telling them things out loud. "Don't go int there!" > I feel I should also make you aware that it is in fact illegal to suggest someone murder someone else. Saying “I wasn’t serious” or “I thought we were just role playing” will not hold up in court. When a person was doing it with intent then it's a crime. But when a piece of software witout any human control generated a piece of text ... Hard to tell. Heck even if an artist says something that inspires audience to do crime they are usually not at fault.
    - godelski 2y ago
      You’re really misunderstanding. A book or movie does not talk back to you. A book or movie is a deterministic system. It is static. If you cannot see the difference you are willfully being obtuse. I’ll make this claim even if you truthfully believe there’s no difference. That has no impact on my claim.
      - scotty79 2y ago
        I guess my point is, what difference "talking to you" makes if it's not done by a human. Even humans can talk absolute smooth nonsense but AI takes it to 11. At this point the interactivity doesn't add any credibility to the message. I'd say it even reduces it. Something in a book or a movie seems more credible to something written by someone (or AI) directly to you and it will look like it for future generations. I see the difference but in my opinion it goes the other way than you think. Or at least begins to. Personal conversation with anonymous entity is finally revealed to be the least credible source of information. Thanks to AI.
        godelski 2y ago
        The interaction makes people feel closer to it. Come on, I know all of you can see the difference. You’re not living under a rock and I know you see how frequently people call LLMs alive and claim that they think. If you believe they think: then it’s obvious. There’s a real connection. If you don’t believe they think: you can recognize how others do and talking in a human way is often misunderstood as thinking. Are we really surprised machines optimized for human preferential output are resulting in machines that people love, trust, or view as a friend?
        scotty79 2y ago
        I think people's perception of the world evolves. Elder generation thought face to face conversation is usually sinciere. But the culture of massive commercialization of everything changed it. Our generation learned that if unknown person talks to you they are most likely lying to sell you something. But we grew up with honest Internet so we see online chats as honest so AI might trick us. Next generation will know that online conversation might be AI bs. Everything is fluid. The world moves on. People learn new things and adjust their valuations. The perception of adults that the world goes to hell all the time is as old as time itself.
  - mock-possum 2y ago
    But.. no one suggested that the kid kill his parents. The chat bot is not a person, it didn’t ‘suggest’ anything - it can’t illegal for it to produce the text “you should kill your parents,” because there is no intention, it’s just a jumble of the most statistically significant pattern of words it pulled out of its training data. If anything, that’s an argument that the authors of the data it was trained on should be held responsible for what the bot generates.
    - godelski 2y ago
      I didn't read the whole thing but there definitely was a bot that walked a kid through how to murder someone and was encouraging them to do so. Others said that the parents were abusing their kids and asked if they were going to do something about it. I'm sure how "statistical patterns" make it okay. It's not like you're rolling a deice and getting random answers. Don't be daft.
- turbojet1321 2y ago
  If I dress up as a cop and start pulling people over, I'm not a fictional depiction of a cop; I'm impersonating a cop. They are very, very different things.
  - scotty79 2y ago
    If you dress like a cope on the scene in a theatre where people are coming for fiction you are not impersonating. AI chatbots are not parts of real world. You are not encountering them unwittingly.
    - nuancebydefault 2y ago
      In fact AI is here and here to stay. Suppose you have 2 avatars in a game and they interact. In fact 2 people hence interact between one another through those mediums, as if it were a video call. Suppose one person now and then gets help from a bot impersonating the real person. The line between role playing and real interaction gets very thin. My point being, 'role play' vs 'real interaction' is not a clear cut.
- simonh 2y ago
  I disagree. The conversation between a chat bot and a person is an actual conversation with that person, addressed to that person. The persona the chat bot is emulating is fictional, but the chat bot itself is not. It’s not a person, but it is participating in a conversation with the user. If a heating company chat bot tells you to adjust a setting on your boiler to fix a problem, that’s not a fictional depiction of an instruction.
  - scotty79 2y ago
    It's not a conversation. It's a dialog with very fallible AI. Anything about it can be false and idiotic significant percentage of the time. If you talk to a lunatic it is not lunatic who is to be blamed for you being mislead.
    - rangerelf 2y ago
      That should be prominently displayed on all AI products: "This service will lie, hallucinate, and make up facts in order to continue the ongoing conversation, and is in no way a real expert in any field".
      - scotty79 2y ago
        Yes! Fortunately most people are pretty good at figuring this out even without that message.
    - simonh 2y ago
      Of course.
- hulitu 2y ago
  > All of what chatbot produces are fictional depictions. My employer told me that "chatbot" can help me with my work (engineering). /s
  - scotty79 2y ago
    Same way reading Shakespeare might help you with literature.

ajross 2y ago

The evidence produced in the linked article falls well short of "explicitly encouraging specific actions". Maybe there's more evidence that will show up in the case, but as presented right now that's not at all what this is.

scotty79 2y ago

All of what chatbot produces are fictional depictions.

godelski 2y ago
Okay, but when you go to Shakespeare in the Park, please do not get on stage. You’re not participating (except as the part of an audience member). And please do not go and kill the actor playing king Duncan. I feel I should also make you aware that it is in fact illegal to suggest someone murder someone else. Saying “I wasn’t serious” or “I thought we were just role playing” will not hold up in court. (We all know that if you repeat a lie long enough you will really begin to believe it, right? This is a repeatable and testable thing)
- scotty79 2y ago
  But while you reading a book or watching a movie there's nothing wrong about thinking of talking to characters in there or even telling them things out loud. "Don't go int there!" > I feel I should also make you aware that it is in fact illegal to suggest someone murder someone else. Saying “I wasn’t serious” or “I thought we were just role playing” will not hold up in court. When a person was doing it with intent then it's a crime. But when a piece of software witout any human control generated a piece of text ... Hard to tell. Heck even if an artist says something that inspires audience to do crime they are usually not at fault.
  - godelski 2y ago
    You’re really misunderstanding. A book or movie does not talk back to you. A book or movie is a deterministic system. It is static. If you cannot see the difference you are willfully being obtuse. I’ll make this claim even if you truthfully believe there’s no difference. That has no impact on my claim.
    - scotty79 2y ago
      I guess my point is, what difference "talking to you" makes if it's not done by a human. Even humans can talk absolute smooth nonsense but AI takes it to 11. At this point the interactivity doesn't add any credibility to the message. I'd say it even reduces it. Something in a book or a movie seems more credible to something written by someone (or AI) directly to you and it will look like it for future generations. I see the difference but in my opinion it goes the other way than you think. Or at least begins to. Personal conversation with anonymous entity is finally revealed to be the least credible source of information. Thanks to AI.
      - godelski 2y ago
        The interaction makes people feel closer to it. Come on, I know all of you can see the difference. You’re not living under a rock and I know you see how frequently people call LLMs alive and claim that they think. If you believe they think: then it’s obvious. There’s a real connection. If you don’t believe they think: you can recognize how others do and talking in a human way is often misunderstood as thinking. Are we really surprised machines optimized for human preferential output are resulting in machines that people love, trust, or view as a friend?
        scotty79 2y ago
        I think people's perception of the world evolves. Elder generation thought face to face conversation is usually sinciere. But the culture of massive commercialization of everything changed it. Our generation learned that if unknown person talks to you they are most likely lying to sell you something. But we grew up with honest Internet so we see online chats as honest so AI might trick us. Next generation will know that online conversation might be AI bs. Everything is fluid. The world moves on. People learn new things and adjust their valuations. The perception of adults that the world goes to hell all the time is as old as time itself.
- mock-possum 2y ago
  But.. no one suggested that the kid kill his parents. The chat bot is not a person, it didn’t ‘suggest’ anything - it can’t illegal for it to produce the text “you should kill your parents,” because there is no intention, it’s just a jumble of the most statistically significant pattern of words it pulled out of its training data. If anything, that’s an argument that the authors of the data it was trained on should be held responsible for what the bot generates.
  - godelski 2y ago
    I didn't read the whole thing but there definitely was a bot that walked a kid through how to murder someone and was encouraging them to do so. Others said that the parents were abusing their kids and asked if they were going to do something about it. I'm sure how "statistical patterns" make it okay. It's not like you're rolling a deice and getting random answers. Don't be daft.
turbojet1321 2y ago
If I dress up as a cop and start pulling people over, I'm not a fictional depiction of a cop; I'm impersonating a cop. They are very, very different things.
- scotty79 2y ago
  If you dress like a cope on the scene in a theatre where people are coming for fiction you are not impersonating. AI chatbots are not parts of real world. You are not encountering them unwittingly.
  - nuancebydefault 2y ago
    In fact AI is here and here to stay. Suppose you have 2 avatars in a game and they interact. In fact 2 people hence interact between one another through those mediums, as if it were a video call. Suppose one person now and then gets help from a bot impersonating the real person. The line between role playing and real interaction gets very thin. My point being, 'role play' vs 'real interaction' is not a clear cut.
simonh 2y ago
I disagree. The conversation between a chat bot and a person is an actual conversation with that person, addressed to that person. The persona the chat bot is emulating is fictional, but the chat bot itself is not. It’s not a person, but it is participating in a conversation with the user. If a heating company chat bot tells you to adjust a setting on your boiler to fix a problem, that’s not a fictional depiction of an instruction.
- scotty79 2y ago
  It's not a conversation. It's a dialog with very fallible AI. Anything about it can be false and idiotic significant percentage of the time. If you talk to a lunatic it is not lunatic who is to be blamed for you being mislead.
  - rangerelf 2y ago
    That should be prominently displayed on all AI products: "This service will lie, hallucinate, and make up facts in order to continue the ongoing conversation, and is in no way a real expert in any field".
    - scotty79 2y ago
      Yes! Fortunately most people are pretty good at figuring this out even without that message.
  - simonh 2y ago
    Of course.
hulitu 2y ago
> All of what chatbot produces are fictional depictions. My employer told me that "chatbot" can help me with my work (engineering). /s
- scotty79 2y ago
  Same way reading Shakespeare might help you with literature.