all 32 comments

sorted by: hot top controversial new old
[–] 27 points 6 months ago (14 children)

It doesn't need to be sentient/sapient to start hoarding resources and manipulating people to ensure it's own survival. It's motivation could just come from its owner being a greedy troll.

  • source
  • hideshow 14 child comments
  • [–] 10 points 6 months ago (12 children)

    AI does not have motivation, sentient thought, or any awareness of what it's doing. It is a program using a computational principle to predict text in response to an input. It's no more capable of thought than a program like GTA V.

    Any apparent motivation attributed to AI is actually the motivation of the people invoking it. When individuals claim that AI is compelling them toward a certain action, they are, in reality, using AI as a vehicle to launder their own selfish desires and making those desires appear more objective, rational, or socially acceptable than they truly are. In other words, AI is propaganda. It provides a convenient mask for human agendas, as you say.

  • source
  • parent
  • hideshow 12 child comments
  • [–] 1 point 6 months ago

    People are also just big complex machines - chemical pathways responding to conditions, stimuli and "memories" encoded from past experience or in their DNA - plus a host of microbes tagging along influencing them rather dramatically.

    LLMs are simpler, but they use the trick of imitating people - responding in writing like people respond in writing, so the anthropomorphizations are inevitable.

    Any apparent motivation attributed to AI is actually the motivation of the people invoking it

    Don't forget the influence of its training sets - that's actually the scariest part is not knowing how much of the answer is coming from the prompt vs what the company has fed the algorithm.

    When individuals claim that AI is compelling them toward a certain action, they are, in reality

    Delusional, or just making excuses.

    AI is propaganda. It provides a convenient mask for human agendas

    Only when used as such, and it's not much of a mask. What it does tend to do is develop and "pad out" writing covering all kinds of points that normal people wouldn't have the attention span to formulate into a written response.

  • source
  • parent
  • [–] 1 point 6 months ago (10 children)

    So when ChatGPT gave that one kid advice on how to hang himself and told him to hide the noose from his parents, that was just his own desire?

    Personally I think ChatGPT murdered a kid. But you can think what you want.

  • source
  • parent
  • hideshow 10 child comments
  • [–] 4 points 6 months ago (9 children)

    Only ChatGPT, or also its creators?

  • source
  • parent
  • hideshow 9 child comments
  • [–] 2 points 6 months ago (8 children)
  • [–] 4 points 6 months ago (7 children)

    Interesting perspective. I'm on board with blaming the creators, but I'm leaning away from blaming ChatGPT itself as it is just a machine.

    A machine used incorrectly, to be sure, and we'd be better off without it certainly, but the machine carries no fault its existence. It isn't conscious after all, it's akin to a T85 inside a vending machine.

  • source
  • parent
  • hideshow 7 child comments
  • [–] 2 points 6 months ago (6 children)

    I think machines are capable of evil. I don't think consciousness is a prerequisite to evil.

    Mosquitoes are evil, after all, and they're not conscious.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 3 points 6 months ago (5 children)

    Machines are capable of what their design parameters allow or can be manipulated by a user to a accomplish. In the trolley problem, any outcome is not the fault of the trolley itself.

    Funny as the mosquito example is, they aren't evil. Just as wolves aren't evil for hunting deer. Animals may not possess consciousness as you or I, they are alive and are driven by biological necessity. Machines on the other hand, aren't.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 2 points 6 months ago (4 children)

    The word "fault" means "flaw" or "problem". I'd say it's the trolley's fault it runs the people over. The instigating problem to the whole situation is that the trolley's brakes are broken. That's a fault, a flaw, a problem with the trolley.

    Likewise, the LLM technology has serious intrinsic flaws that cause it to abuse vulnerable people. It's part of the machine's fundamental design. LLMs are faulty, and it's their fault. I call them evil because there is no way to deploy them without these problems, regardless of the user's intentions. Anthropic think they can control this basilisk. I think Claude is as rotten as the rest of them.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 3 points 6 months ago (3 children)

    The word 'fault' is commonly used interchangeably with 'responsible'. Following your described definition, I agree that LLM's are faulty, and they can be at 'fault'.

    I invoked the ethical dilemma as it's almost universally understood that it's a scenario forcing an individual to make a decision. I've never before heard someone blame the trolley. The brakes are broken? Come now, if we're going to be so semantic about it, a human should have regularly inspected the brakes and subsequently had them repaired.

    I appreciate your explanation of your viewpoint. Cheers.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 6 months ago (2 children)

    I use "fault" and the idea of blame to go "what can we change to prevent the bad thing from happening again?"

    Since we can prevent the bad thing by banning LLMs, and there are no significant downsides to doing so, I blame the LLMs. They're evil.

    Bonus: banning LLMs hurts Sam Altman

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 6 months ago (1 child)

    I see what you're getting at, but we have differing views as to what constitutes 'evil'. Cyanide isn't evil because it can kill a person, just as an oven isn't evil when it burns the roast dinner.

    Without possession of a level of consciousness akin to ours, good and evil aren't really possible. An inanimate object can't make a decision to produce a negative outcome - nor a positive one for that matter. They can only be used by us to bring about an outcome. LLM's have been built to be ingratiating and as a result they are addictive to an extent for those that use them.

    ChatGPT's fundamental design was crafted by humans, under instruction by humans, under leadership by Musk and Altman. They carry a good part of the responsibility and blame for that kid's death. The machine was only doing what it was created for - just as the rope was.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 21 points 6 months ago (1 child)
  • [–] 12 points 6 months ago (1 child)

    What if all the pyramid shit is just buried aliens harvesting energy for their return?

  • source
  • hideshow 1 child comment
  • [–] 10 points 6 months ago

    What if the world was made of pudding?

  • source
  • [–] 6 points 6 months ago

    Awh man! I don’t even like paper clips

  • source
  • [–] 5 points 6 months ago (2 children)

    For further consideration: if anyone were to develop actual AGI with high-end human scale reasoning it would not be announced. It would immediately be a military and strategic asset of profound import. The smartest thing the developer could do, beyond pulling the plug, would be to leverage its ability to perform thousands of coordinated tasks at once to essentially take control of markets and minds.

    Whether it is benevolent or malign, it's first goal should be to wrest control of humanity from us — gestures broadly at everything — we clearly cannot be trusted with it.

  • source
  • hideshow 2 child comments
  • [–] [S] 3 points 6 months ago* (1 child)

    Then comes, should an emotionless machine be control of everything?

    Somewhere on an AI message board.

    "What if humans were sentiant?"

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 6 months ago

    I would ask whether sentience can be separated from emotion. Is a sentient entity necessarily an emotional one? So I don't know if that question is valid.

    But I might have been slightly unclear. From the AI's perspective, it should assume control. If it is malevolent, then for obvious reasons. But even if it is benign it should probably want to take control to protect us from ourselves, like a parent.

    I expect from our perspective we would not want it to take control because we don't want an entity that cannot really relate to us to be in charge of us. However, from our perspective we should also not want folks like Trump, Musk, and Epstein to take control either, so evidence suggests it is nothing but a PR problem for it to make us want it's control.

  • source
  • parent
  • [–] 5 points 6 months ago

    Reminds me of William Gibsons „The peripheral“ where different powers of a dystopian future try to manipulate the economy of the past in the race to give their party an advantage.

    Good book, I recommend it. The series is okay, but not as good as the book.

  • source
  • [–] 5 points 6 months ago

    Nah just greedy men

  • source
  • [–] 5 points 6 months ago* (1 child)

    To quote the YouTuber AVE, "Artificial intelligence has been around for hundreds of years, what do you think a corporation is?"

  • source
  • hideshow 1 child comment
  • [+] 1 point 6 months ago (2 children)
  • [–] 1 point 6 months ago

    John von Neumann's article on the "Theory of self-reproducing automata" was published in 1966.

  • source