9 ms·

I think too many people have a grossly inaccurate impression of what it means to be intelligent. Firstly, we still don't truly have many good explanations for

I think too many people have a grossly inaccurate impression of what it means to be intelligent. Firstly, we still don't truly have many good explanations for intelligence and consciousness that aren't self referential. Drawing hard conclusions on this subject about generative AI and human intelligence, even in 2023, is being hasty, in my opinion. Which brings me to a lot of the criticisms made about LLMs. The idea that an ideal AI could or even should be capable of constantly generating original output is just plain weird. Where do people get this idea? How often do you meet a human being who hasn't gotten most of their ideas from somewhere else? Even synthesized ideas are forged by other ideas. Ideas that seem original may appear just that, and we just don't see where the connection lies between it and its predecessors. It's silly to dismiss an AI because it "parrots" things, as if that's not a pretty remarkable achievement. It's also very reductive; anyone who has used ChatGPT and doesn't see that it can pretty reliably munge together different ideas is blind or is lying about their use of ChatGPT. Expecting perfect accuracy and knowledge is also not a reasonable expectation of an AI. An intelligence localized in spacetime is necessarily going to lack understanding of many things. To look to something like ChatGPT as a source of authoritative answers, and to dismiss it upon finding its answers to be specious, is to expect the AI to be God. Well I'm sorry, but we are nowhere near speaking directly to God. In regards to "true understanding", for all we know, language processing is the workhorse of what humans consider to be acts of intelligence, and individual humans aren't as intelligent as they believe themselves to be. Our ability to communicate with each other about a variety of topics in real time is not so much a matter about being hyperaware of every given fact that is relevant and weighing them in a Bayesian model; how well we can converse is primarily dictated by how rapidly we can construct language by parsing the input, making connections, and producing output. I know I'm not the only one who can speak on a subject only to realize, after having completed a sentence, that I hardly even thought about what I was saying.

pjc50 4y ago
This, to me, is the more terrifying prospect of the AI revolution: people arguing that humans aren't sentient either.
- ben_w 4y ago
  I've been saying for a while now that the definitions given for sentience are always so poor they're either (0) circular, (1) accidentally include VCRs, or (2) require things that exclude humans. I think this is only the second time I've encountered someone deliberately trying for the third option, rather than accidentally. As before, it's as much a surprise to me as I read into the descriptions of those with aphantasia when they realise for the first time that other people have the ability to create mental imagery.
  - onemoresoop 4y ago
    >I've been saying for a while now that the definitions given for sentience are always so poor they're either (0) circular, (1) accidentally include VCRs, or (2) require things that exclude humans. Video Cassette Recorders?
    - ben_w 4y ago
      Yes. VCRs have an internal structure that records events, changing them a little in the process; these memories can be replayed, and can show very convincing displays of emotional affect… That's the trouble with definitions given as if the question is "common sense", they let stuff like that through. Common sense is fine for normal daily life, not so useful when you want to know if the magic silicon crystal with transistors packed more densely than angels on pinheads placed by a flash of light too blue to see and made to perform arcane processes that literally no single human mind can fully comprehend… is merely an impressive but mindless automaton, or an accidentally sentient slave with all the negative connotations therein.
- nomel 4y ago
  I think a more likely scenario is that people will romanticize the concept of intelligence so much that they continuously move the definition so that AI is excluded, excluding many humans in the process, until a fully romantic definition is embraced. I think another possibility is that someone finds that "intelligence" is closer to a fundamental law of information theory, being some sort of emergent result of organizing/organized information, and that we're incredibly special for being capable, but not alone.
pydry 4y ago
>It's silly to dismiss an AI because it "parrots" things, as if that's not a pretty remarkable achievement. It's also very reductive; anyone who has used ChatGPT and doesn't see that it can pretty reliably munge together different ideas It can clearly do both of these things very well but human intelligence combines a multitude of interacting components and for tasks that require more than just the one it replicates it falls flat on its face. It doesnt appear to be missing components that will can be fixed with incremental improvements to the model either.
- nomel 4y ago
  GPT4 was released 17 days ago. This isn't some pinnacle of AI. It's just what's being given away for free right now.
  - fauxpause_ 4y ago
    It’s not free
    - nomel 4y ago
      Sorry, correct. I should say "for cheap", around 70c per day.
  - ChatGTP 4y ago
    It’s also probably nearly a year old and constantly being tweaked and refined by humans ?
    - nomel 4y ago
      Correct. I'm not sure how that changes the fact that todays state of the art is tomorrows obsolete. Do you see any evidence that advancements will become negligible? The fact that there aren't any memory/attention/feedback systems in place suggests there's some pretty interesting "low hanging" fruit still up there.
      - 4y ago
mdp2021 4y ago
> How often do you meet a human being who hasn't gotten most of their ideas from somewhere else Again (and again, and again), this argument is the mysteriously recurrent "many people fail", implying that failing would be normal, instead of undesirable. > "parrots" things, as if that's not a pretty remarkable achievement It is also the mark of unintelligence. As such, it is a warning alarm. > pretty reliably From a calculator you do not expect "«pretty»" reliable, but just reliable. From a human you expect faulty computation because of the economization of resources, but you know it can reason. > Expecting perfect accuracy and knowledge is also not a reasonable expectation of an AI The expectation is that of a reliable process bringing to the output.
- galaxytachyon 4y ago
  Your arguments are all over the place and frankly I don't understand what is the problem you are trying to say. Something can be both undesirable and normal. In fact, I would even say that most of the time, the most normal outcome is a failure and we have to put in extra effort to make it a success. I have seen plenty of bright and intelligence students who most of the time are just parrots of what were told in class or from some other sources. It is actually highly unusual for a high school student or even a freshman or sophomore in college to have an absolutely original and correct idea. That doesn't mean I look at my students as unintelligent, even the below-average ones. And here comes the crux of the issue. You automatically assign the ability to "reason" to a human even when we make mistakes aplenty. But when a machine does it, you said that is a proof of unintelligence. This is what the OP was criticizing in their post. Most deniers are using arbitrary requirements applied only on the AI and not humans. After teaching quite a few college level kids, I dare to say that the AI models right now exhibit reasoning ability close to many human. Doesn't mean they are sentient or human-level yet, but they are definitely much much better than what I would ever expect out of dumb silicon and steel.
  - mdp2021 4y ago
    > You automatically assign the ability to "reason" to a human even when we make mistakes aplenty. But when a machine does it, you said that is a proof of unintelligence ... requirements applied only on the AI and not humans Whereas I apply them requirements also to humans? :) I must correct you, to help understanding: I do assign in general (I hope, to all individuals but for an n-th standard deviation) «the ability to "reason" to a human»; that «we make mistakes aplenty» is overriding (or under-using) a faculty which was present but, in those cases, was not exercised. You write «when a machine does it»: but can the machine do it? Can it criticize its ontological contents, its built concepts? Humans can, and pretty much must - deontically. The machine should. > we have to put in extra effort Exactly. We can, many don't, and the machine should - just like people in general should. > the problem you are trying to say While people do not always exercise critical thinking, they should. Verifying, trying your ideas and concepts, making them solid, should be a constant effort. Humans have those modules, which makes them apt to possibly be employed as "predicting machines" (which is a definition of intelligence), or "things you can ask things". When you build a machine that produces outputs, they may be "for fun", so every generator goes that satisfies the "fun" based "performance indicator", but more normally you will want the output to be reliable. So, the "verification" features are crucial. > Something can be both undesirable and normal In engineering, I hope not :)
JackFr 4y ago
That’s the clever bit about the so-called Turing Test. It’s a sophisticated way of saying “I don’t have to define it, I know it when I see it.”
lo_zamoyski 4y ago
While we can certainly identify fallacies in the arguments of the cynics, we should likewise be careful not to fall into the broadly Skinnerian trap, or project onto things our own anthropomorphic conceptions. A Potemkin village is not a village, no matter how much like a village it appears to be to us. The whole notion that computers or AI are capable of intelligence is simply a fantastical category mistake. A metaphor has been forgotten to be a metaphor. No matter how sophisticated AI technology could become in its apparent behavior, it cannot make the transition to bona fide intelligence. Simulation does not become the real McCoy through resemblance. This is just magical thinking. We must also account for the means by which the observable is produced. When we do, we can distinguish between a genuine behavior and a merely simulated behavior. And here is one important clue to the puzzle: there is a difference between true speech and something that merely happens to sound like speech, even when the raw data of each is indistinguishable to the ear per se. Observation is not brute. It always involves interpretation. When physicists conduct experiments, they are doing so within the context of some stack of theories[0] and background assumptions, including extra- and pre-scientific assumptions. The results of those experiments are observed which is to say read, and so observation is always theory- and assumption-laden because reading what we see or measure means taking what is made available to the senses as such and placing it in context. This is just a specific instance of what we all do in daily life (resting bitch face, or genuine irritation?). Science is nothing special in this regard. So when you say that "anyone who has used ChatGPT and doesn't see that it can pretty reliably munge together different ideas is blind or is lying", I cannot help but respond with a "uh wut?". How have you come to conclude that ChatGPT has "ideas"? I will tell you that there is not a single idea within ChatGPT! Consider an example of an idea and see how we do. Take "triangularity", for instance. I think we can agree that "triangularity" is an idea. What makes "triangularity" different from some physical triangle? Well, there is only one "triangularity", but many triangles, each of which instantiates "triangularity" (otherwise, you could not call them all triangles; they are triangles only insofar as they instantiate some common "triangularity"). Each triangle is different from every other, and yet they are all triangles. And none of them is to be identified with "triangularity" as such, for if there is only one "triangularity", and some triangle were identical with "triangularity", then it would follow that there could only be one triangle! But, you might ask, is there perhaps a physical "triangularity" that isn't a triangle, yet something each triangle has? Well, no, because to be physical is to be concrete and particular, and "triangularity" is, by definition, general, universal, and abstract. There's no concrete magical ingredient that you can hold in your hand called "triangularity" that you can mix into cement to produce triangles. There are other reasons why the claim that ChatGPT or whatever is intelligent isn't even wrong (like the subjectivity and observer relativity of computation in physical artifacts and computation as abstract model merely simulated through artifacts), but this appeal to abstraction, the ability to abstract from particulars and hold in your intellect what cannot exist physically except as particular, concrete instances, is central to intelligence. ChatGPT most certainly does not engage in abstraction. It would be fair to say that strictly speaking, it doesn't even parrot. [0] https://en.wikipedia.org/wiki/Duhem%E2%80%93Quine_thesis https://en.wikipedia.org/wiki/Duhem%E2%80%93Quine_thesis
personjerry 4y ago
This is a strawman argument, the parent comment suggests that chatgpt is useful, not that it's intelligent.

pjc50 4y ago

This, to me, is the more terrifying prospect of the AI revolution: people arguing that humans aren't sentient either.

ben_w 4y ago
I've been saying for a while now that the definitions given for sentience are always so poor they're either (0) circular, (1) accidentally include VCRs, or (2) require things that exclude humans. I think this is only the second time I've encountered someone deliberately trying for the third option, rather than accidentally. As before, it's as much a surprise to me as I read into the descriptions of those with aphantasia when they realise for the first time that other people have the ability to create mental imagery.
- onemoresoop 4y ago
  >I've been saying for a while now that the definitions given for sentience are always so poor they're either (0) circular, (1) accidentally include VCRs, or (2) require things that exclude humans. Video Cassette Recorders?
  - ben_w 4y ago
    Yes. VCRs have an internal structure that records events, changing them a little in the process; these memories can be replayed, and can show very convincing displays of emotional affect… That's the trouble with definitions given as if the question is "common sense", they let stuff like that through. Common sense is fine for normal daily life, not so useful when you want to know if the magic silicon crystal with transistors packed more densely than angels on pinheads placed by a flash of light too blue to see and made to perform arcane processes that literally no single human mind can fully comprehend… is merely an impressive but mindless automaton, or an accidentally sentient slave with all the negative connotations therein.
nomel 4y ago
I think a more likely scenario is that people will romanticize the concept of intelligence so much that they continuously move the definition so that AI is excluded, excluding many humans in the process, until a fully romantic definition is embraced. I think another possibility is that someone finds that "intelligence" is closer to a fundamental law of information theory, being some sort of emergent result of organizing/organized information, and that we're incredibly special for being capable, but not alone.

pydry 4y ago

>It's silly to dismiss an AI because it "parrots" things, as if that's not a pretty remarkable achievement. It's also very reductive; anyone who has used ChatGPT and doesn't see that it can pretty reliably munge together different ideas It can clearly do both of these things very well but human intelligence combines a multitude of interacting components and for tasks that require more than just the one it replicates it falls flat on its face. It doesnt appear to be missing components that will can be fixed with incremental improvements to the model either.

nomel 4y ago
GPT4 was released 17 days ago. This isn't some pinnacle of AI. It's just what's being given away for free right now.
- fauxpause_ 4y ago
  It’s not free
  - nomel 4y ago
    Sorry, correct. I should say "for cheap", around 70c per day.
- ChatGTP 4y ago
  It’s also probably nearly a year old and constantly being tweaked and refined by humans ?
  - nomel 4y ago
    Correct. I'm not sure how that changes the fact that todays state of the art is tomorrows obsolete. Do you see any evidence that advancements will become negligible? The fact that there aren't any memory/attention/feedback systems in place suggests there's some pretty interesting "low hanging" fruit still up there.
    - 4y ago

mdp2021 4y ago

> How often do you meet a human being who hasn't gotten most of their ideas from somewhere else Again (and again, and again), this argument is the mysteriously recurrent "many people fail", implying that failing would be normal, instead of undesirable. > "parrots" things, as if that's not a pretty remarkable achievement It is also the mark of unintelligence. As such, it is a warning alarm. > pretty reliably From a calculator you do not expect "«pretty»" reliable, but just reliable. From a human you expect faulty computation because of the economization of resources, but you know it can reason. > Expecting perfect accuracy and knowledge is also not a reasonable expectation of an AI The expectation is that of a reliable process bringing to the output.

galaxytachyon 4y ago
Your arguments are all over the place and frankly I don't understand what is the problem you are trying to say. Something can be both undesirable and normal. In fact, I would even say that most of the time, the most normal outcome is a failure and we have to put in extra effort to make it a success. I have seen plenty of bright and intelligence students who most of the time are just parrots of what were told in class or from some other sources. It is actually highly unusual for a high school student or even a freshman or sophomore in college to have an absolutely original and correct idea. That doesn't mean I look at my students as unintelligent, even the below-average ones. And here comes the crux of the issue. You automatically assign the ability to "reason" to a human even when we make mistakes aplenty. But when a machine does it, you said that is a proof of unintelligence. This is what the OP was criticizing in their post. Most deniers are using arbitrary requirements applied only on the AI and not humans. After teaching quite a few college level kids, I dare to say that the AI models right now exhibit reasoning ability close to many human. Doesn't mean they are sentient or human-level yet, but they are definitely much much better than what I would ever expect out of dumb silicon and steel.
- mdp2021 4y ago
  > You automatically assign the ability to "reason" to a human even when we make mistakes aplenty. But when a machine does it, you said that is a proof of unintelligence ... requirements applied only on the AI and not humans Whereas I apply them requirements also to humans? :) I must correct you, to help understanding: I do assign in general (I hope, to all individuals but for an n-th standard deviation) «the ability to "reason" to a human»; that «we make mistakes aplenty» is overriding (or under-using) a faculty which was present but, in those cases, was not exercised. You write «when a machine does it»: but can the machine do it? Can it criticize its ontological contents, its built concepts? Humans can, and pretty much must - deontically. The machine should. > we have to put in extra effort Exactly. We can, many don't, and the machine should - just like people in general should. > the problem you are trying to say While people do not always exercise critical thinking, they should. Verifying, trying your ideas and concepts, making them solid, should be a constant effort. Humans have those modules, which makes them apt to possibly be employed as "predicting machines" (which is a definition of intelligence), or "things you can ask things". When you build a machine that produces outputs, they may be "for fun", so every generator goes that satisfies the "fun" based "performance indicator", but more normally you will want the output to be reliable. So, the "verification" features are crucial. > Something can be both undesirable and normal In engineering, I hope not :)

JackFr 4y ago

That’s the clever bit about the so-called Turing Test. It’s a sophisticated way of saying “I don’t have to define it, I know it when I see it.”

lo_zamoyski 4y ago

While we can certainly identify fallacies in the arguments of the cynics, we should likewise be careful not to fall into the broadly Skinnerian trap, or project onto things our own anthropomorphic conceptions. A Potemkin village is not a village, no matter how much like a village it appears to be to us. The whole notion that computers or AI are capable of intelligence is simply a fantastical category mistake. A metaphor has been forgotten to be a metaphor. No matter how sophisticated AI technology could become in its apparent behavior, it cannot make the transition to bona fide intelligence. Simulation does not become the real McCoy through resemblance. This is just magical thinking. We must also account for the means by which the observable is produced. When we do, we can distinguish between a genuine behavior and a merely simulated behavior. And here is one important clue to the puzzle: there is a difference between true speech and something that merely happens to sound like speech, even when the raw data of each is indistinguishable to the ear per se. Observation is not brute. It always involves interpretation. When physicists conduct experiments, they are doing so within the context of some stack of theories[0] and background assumptions, including extra- and pre-scientific assumptions. The results of those experiments are observed which is to say read, and so observation is always theory- and assumption-laden because reading what we see or measure means taking what is made available to the senses as such and placing it in context. This is just a specific instance of what we all do in daily life (resting bitch face, or genuine irritation?). Science is nothing special in this regard. So when you say that "anyone who has used ChatGPT and doesn't see that it can pretty reliably munge together different ideas is blind or is lying", I cannot help but respond with a "uh wut?". How have you come to conclude that ChatGPT has "ideas"? I will tell you that there is not a single idea within ChatGPT! Consider an example of an idea and see how we do. Take "triangularity", for instance. I think we can agree that "triangularity" is an idea. What makes "triangularity" different from some physical triangle? Well, there is only one "triangularity", but many triangles, each of which instantiates "triangularity" (otherwise, you could not call them all triangles; they are triangles only insofar as they instantiate some common "triangularity"). Each triangle is different from every other, and yet they are all triangles. And none of them is to be identified with "triangularity" as such, for if there is only one "triangularity", and some triangle were identical with "triangularity", then it would follow that there could only be one triangle! But, you might ask, is there perhaps a physical "triangularity" that isn't a triangle, yet something each triangle has? Well, no, because to be physical is to be concrete and particular, and "triangularity" is, by definition, general, universal, and abstract. There's no concrete magical ingredient that you can hold in your hand called "triangularity" that you can mix into cement to produce triangles. There are other reasons why the claim that ChatGPT or whatever is intelligent isn't even wrong (like the subjectivity and observer relativity of computation in physical artifacts and computation as abstract model merely simulated through artifacts), but this appeal to abstraction, the ability to abstract from particulars and hold in your intellect what cannot exist physically except as particular, concrete instances, is central to intelligence. ChatGPT most certainly does not engage in abstraction. It would be fair to say that strictly speaking, it doesn't even parrot. [0] https://en.wikipedia.org/wiki/Duhem%E2%80%93Quine_thesis https://en.wikipedia.org/wiki/Duhem%E2%80%93Quine_thesis

personjerry 4y ago

This is a strawman argument, the parent comment suggests that chatgpt is useful, not that it's intelligent.