195
An Experiment Put LLMs in Charge of Radio Stations. You’ll Never Guess How It Went (gizmodo.com)
submitted 4 months ago by to c/technology@lemmy.world
101 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 29 points 4 months ago (24 children)

LLMs/Chatbots confabulate statistically probable texts, there's no compassion possible.

Don't fall into the AI-marketing trap of "we don't know what's happening in the black box, so we have to assume there's consciousness in there". The systems produce convincing deceptive language, but all signs of intelligence or compassion anyone sees in them is just an anthropomorphic projection.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 24 child comments
    •
  • [–] 2 points 4 months ago* (last edited 4 months ago) (21 children)

    This is a semantic argument, they obviously mean it emulates compassion better than a real human, and given its issue with sycophancy this is undoubtedly true, even to a fault. There's no need to do this every time someone says an ai thinks or does some humany thing, everyone gets it, the language for saying these things is just clunky.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 21 child comments
    •
  • [–] 11 points 4 months ago (2 children)

    everyone gets it

    Disagree, plenty of people still need to hear it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 5 points 4 months ago (13 children)

    How is it a semantic argument? They're talking about how LLMs work on a functional level, not arguing the meaning of compassion itself. It's not hard to say that they emulate compassion and intelligence relatively well, applying human adjectives without any nuance just opens it up to being misinterpreted by people who don't know any better.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 13 child comments
    •
  • [–] -2 points 4 months ago (11 children)

    It's semantic because it's really about language. Who cares that it's not doing that like a human would, everyone who knows anything knows that and they were clearly using language in a less cumbersome way.

    yes, everyone already knows what you're saying, but it doesn't matter and serves no purpose other than making it difficult to talk about their behaviours. The only workaround for this would be inventing new terms for when an ai does a behavior that resembles a human one. It'd be very cumbersome and add no value to any conversation.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 11 child comments
    •
  • [–] 4 points 4 months ago (10 children)

    This is assuming that the average person has a solid grasp of the inner workings of an LLM, which unfortunately isn't the case. Regardless, it would only be a semantic argument if they were shifting the meanings of the relevant words to support their argument, which they evidently weren't doing here.

    LLMs don't think, they predict patterns in language mathematically, making them functionally incapable of human capacities like compassion and intelligence, both of which require a conscious mind to be displayed. To use words that go against that without being precise is to imply the opposite. It's simply a matter of describing it accurately.

    If anything, considering it 'AI' is a semantic argument because it implies there's some form of higher thinking occurring under the surface, which there clearly isn't. It would be like if I said my PC was intelligent because it has a CPU. Obviously we've passed the point of using a better term, but it's still unfortunate we've decided on that because it's inherently misleading.

    It’d be very cumbersome and add no value to any conversation.

    I think you're using cumbersome in an unnecessarily negative way since it's very much an inevitable feature of the concept at hand. Yes, it's cumbersome, like all controversial fields of study. Things like that work themselves out over time. Until then we'll just have to deal with it without misleading anyone.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 10 child comments
    •
  • [–] -1 points 4 months ago (9 children)

    What exactly is the harm in people being mislead in this way, as long as they still know about the risks of hallucination, in your eyes?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 9 child comments
    •
  • [–] 3 points 4 months ago (5 children)

    When is being mislead not a bad thing? In a perfect world, there would be none of that. Of course we don't live in a utopia, but I'd prefer if we avoid spreading skewed understandings of anything at all as much as possible. It's a matter of principle.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] -1 points 4 months ago (4 children)

    Nobody does this when people say their computer is "thinking" when it's running slow, I just don't see the necessity of pointing this out every time the topic is brought up.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 1 point 4 months ago (3 children)

    Ideally people who say that aren't misled into believing their computer is thinking in the same way that a human is

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 0 points 4 months ago (2 children)

    Nobody thinks they are the same, some wonder if it can be made to do equivalent things. Nobody needs to hear that what they are doing isn't thought for the same reason nobody needs to hear that my computer isn't actually thinking when it's running slow.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 4 months ago* (1 child)

    I don't think this is true. People are out here having straight up AI psychosis. Do you really think there aren't fools who believe their computer is actually thinking like a human?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 0 points 4 months ago

    I don't think a single one of them believe that pattern prediction vs a human brain is meaningfully distinct, so this argument is worthless to them, it'd be like saying "nuh uh", you're not engaging with their beliefs, so, how would this help?

  • permalink
  • fedilink
  • source
  • parent
  • [–] 3 points 4 months ago (2 children)

    Having an inaccurate view about something so fundamental to the topic leads you to predict reality incorrectly and make bad decisions

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] -1 points 4 months ago (1 child)

    Nobody does this when people say their computer is "thinking" when it's running slow, I just don't see the necessity of pointing this out every time the topic is brought up.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 1 point 4 months ago

    I agree it's unnecessary to point out. Using anthropomorphising shorthand to talk about technology is extremely common, and AI is no different; saying an AI is "thinking" or whatever is fine. But there is a difference between using that language as shorthand, and actually holding misconceptions about what is really happening. So saying that it's fine for someone to be misled and use that language is different than just saying the language makes sense to use.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 3 points 4 months ago* (3 children)

    It's not semantic – it's completely different things happening if there's real consciousness and compassion present based on lifeforms on one hand or a mere simulation of that in form of a text output on the other hand that only superficially looks like there's something intelligent.

    People regularly fall for the illusion and project their own feelings into the machine while reading the text output of an LLM. Many are not capable of differentiating and the chatbots are designed in a way to make it more and more difficult to recognize synthetic output.

    Humans are good in projecting their own feelings into things they see, just look at all the cat or dog owners who believe they can read the thoughts of their "babies" from their facial expressions.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 0 points 4 months ago* (2 children)

    It's semantic because it's really about language. Who cares that it's not doing that like a human would, everyone who knows anything knows that and they were clearly using language in a less cumbersome way.

    yes, everyone already knows what you're saying, but it doesn't matter and serves no purpose other than making it difficult to talk about their behaviours. The only workaround for this would be inventing new terms for when an ai does a behavior that resembles a human one. It'd be very cumbersome and add no value to any conversation.

    to those not capable of understanding this or who disagree with you, what you're saying wouldn't convince them anyway, you're just adding noise to these conversations.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 4 months ago (1 child)

    I'm sorry that you seem to have given up on people. I still think discussing things is worthwhile but from your text I get the impression that you don't believe in social discourse.

    I think it's a fact that people learn through and with language and that the symbols we humans use are what constructs our reality. So it's very important what kind of language we use, how we use it and to be precise in our terms and vocabulary.

    People are not readymade, everybody has to learn and learning never stops and people are not all on the same level. So no: not everyone knows the same, as you seem to assume.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 0 points 4 months ago

    I absolutely do, I'm just tired of this being posted every time someone says an AI is doing a thing, it's overplayed, social discourse is obviously valuable, this just isn't a valuable point to make.

    People also wouldn't do this if I said my slow computer was "thinking".

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 4 months ago (1 child)

    Anthropic have actually looked at how their LLMs reason. Don't use them for anything important.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 1 point 2 months ago

    The "reasoning" which is displayed, such as step-by-step explanations or chain-of-thought responses, is only a simulation of reasoning, generated as text based on patterns. It's in the same form of any text production an LLM is capable of. When an LLM mimics "reasoning," it is generating text that looks like reasoning because it has seen similar patterns in its training data.

  • permalink
  • fedilink
  • source
  • parent