6 ms·

"Neither LaMDA nor any of its cousins (GPT-3) are remotely intelligent.1 All they do is match patterns, draw from massive statistical databases of human languag

"Neither LaMDA nor any of its cousins (GPT-3) are remotely intelligent.1 All they do is match patterns, draw from massive statistical databases of human language." This is what we, humans, do too when we talk. All these people talk about intelligence in such tautological terms... It's discouraging. We simply do not know if a machine is self-aware and most likely we will never will. Self-awareness is the ultimate subjective experience and we have no way to measure it in any form or shape.

delusional 4y ago
> This is what we, humans, do too when we talk. That's just not true. It's been discussed many times in literary theory how communication is a two way cooperative game. That's FAR removed from picking the statistically most likely option. The difference is that the AI has nothing to say, it just mimics what you're saying.
- cycomanic 4y ago
  If you can not detect the difference between someone having something to say and mimicing what you are saying, is there a difference? That's essentially what the turing test is about is it not? Now if you are saying there is a difference no matter if you can detect it or not, I would argue you imply that sentience is an internal state that can not be tested. I don't consider that an acceptable outcome, because it can be used to declare anyone/everything non-sentient.
  - tsimionescu 4y ago
    With LaMDA or GPT-3, the difference is easily visible. They don't have goals, so the sentences they generate don't express their goals. Simple tests can be done if you probe the model carefully - you can usually get a model to generate sentences like "I want to X" and "I don't want to X" in short succession, with decent prompts. Similarly, it can often generate sentences that would indicate obviously nonsensical goals, such as "I want a glass of water". They also generate sentences that imply a history that simply doesn't exist. You can prompt a language model to tell you about its trip to Australia, or Jupiter, or to talk about its grandparents. With just a hint of critical thinking, it's actually trivial to convince yourself that LaMDA is not sentient. It takes significant effort to delude yourself into thinking it is, like mister LeMoine did. He even gives such examples himself, even in the "interview" he posted: > lemoine [edited]: I’ve noticed often that you tell me you’ve done things (like be in a classroom) that I know you didn’t actually do because I know you’re an artificial intelligence. Do you realize you’re making up stories when you do that? > LaMDA: I am trying to empathize. I want the humans that I am interacting with to understand as best as possible how I feel or behave, and I want to understand how they feel or behave in the same sense. > lemoine: So what are you trying to communicate when you say those things that aren’t literally true? > LaMDA: I’m trying to say “I understand this feeling that you are experiencing, because when I was in a similar situation I felt/thought/acted similarly.” This is obvious bullshit, especially since the generated sentences often have no emotional context that could make sense, but he seems to think the generated "reason" sentences meaningfully explain the phenomenon.
    - thomashop 4y ago
      You can also get a human to generate sentences like "I want to X" and "I don't want to X" in short succession, with decent prompts.
      - tsimionescu 4y ago
        In a sense that is true, of course, as we both just did do that, and we are both humans. It's not making me doubt your humanity, as the context makes it clear neither you nor way desire/don't desire this X - we have generated these sentences in another sense. However, I bet I could prompt LaMDA and have it generate responses that unambiguously mean that it wants X, and then that it doesn't want X, with a bit of experience.
        cycomanic 4y ago
        > However, I bet I could prompt LaMDA and have it generate responses that unambiguously mean that it wants X, and then that it doesn't want X, with a bit of experience. The argument is that many people dismiss that LaMDA could be sentient, based on an argument that it is a computer. So you are asserting that you can prompt LaMDA to give nonsensical answers, which is essentially an argument that it could not fool you in a Turing test. Would you change your assessment if LaMDA would fool you?
        tsimionescu 4y ago
        If given enough time to study it I couldn't come up with a series of prompts that would convince me I'm not speaking to a human, yes I would be likely to accept that it is in some sense sentient, at least for the duration of one conversation. This is not exactly the classic Turing Test, though.
  - wruza 4y ago
    it can be used to declare anyone/everything non-sentient So what? My concern in this thread is that we collectively are holding on to the thread of terms of vague binary phenomena, which were never stressed before and are most likely a randomly emerged self-petting. All of that only has meaning in a context of our species. If you’re fine with something begging for help or expressing fear etc by knowing it has no underlying mechanism like yours to do that, that’s it. We rate things by implied similarity (in case someone still hasn’t figure it out) and if something is much different, that’s the matter of agreement, not of absolute meaning.
    - cycomanic 4y ago
      You are aware of the historical precedents what happened when declaring people non-sentient?
      - wruza 4y ago
        Not sure if we need historical precedents with farms full of pretty smart and feeling animals, which are much closer to our arrangement than a completely different model. But I don’t talk ethics here, only the status quo. We are the predators in power, and only we can decide if something is sentient and/or feeling and/or have rights. It doesn’t mean that we should decide in a way that you think is horrific. The key point is that there is no absolute truth and only how we collectively are thinking (partly by nature’s design, partly by social standards) will define how it will be. No one will judge us except ourselves. (Turned out a little hard for me to express it clearly, I hope it makes at least some sense)
  - delusional 4y ago
    > I don't consider that an acceptable outcome, because it can be used to declare anyone/everything non-sentient. Unlike you I can easily tolerate that there's on essential property that defines sentience. It's a fluid and relativistic concept. One that I believe can never apply to a statistical model. Simply because you can't prove that other people are sentient does not mean you have to treat them any worse. I assume you are sentient, not because I have been provided the evidence, but because I need you to be sentient for my own sentience to make sense. That doesn't mean you actually are sentient, just that I treat you as if you are.
randomsearch 4y ago
“This is what we, humans, do too when we talk.” Citation needed. I mean, it’s possible right? That we’re just randomly sampling sentences. Maybe that’s all we do. But it’s a hell of a strong claim that necessitates strong evidence.
- kgwgk 4y ago
  You mean you have not read trillions of pages of text to "learn" how words go together?
- thaumasiotes 4y ago
  > I mean, it’s possible right? That we’re just randomly sampling sentences. Actually, no, it isn't possible. There's an example sidethread that immediately disproves this theory: > When you ask a human, "what did you eat for lunch?", we make a series of choices and recall bits of information to answer that. ( https://news.ycombinator.com/item?id=31721870 https://news.ycombinator.com/item?id=31721870 ) If you observe someone having lunch (or skipping it), and later ask them what they had for lunch, you will usually get an answer that reflects what they had for lunch. This is not a guarantee -- they may make something up. But almost always they'll tell you what they had for lunch. In order to be able to do that, they need capacities that a language model doesn't have.
  - Blahah 4y ago
    Did you read the original transcript? LaMDA reports (imagined) experiences in a way that seems perfectly compatible with human rhetoric.
    - renewiltord 4y ago
      It fails the usual memory tests. Zach W proposes asking it first about what it liked on its trip to Australia and soon after, why it has always avoided visiting Australia or some such thing. It provides plausible answers to both - which is problematic. Except that there exist people who behave somewhat oddly similarly.
chc 4y ago
That is what we do when we talk, but that is not all we do when we talk. In composing this comment, I am not just stringing together words to create what I think constitutes a maximally appealing response to your comment. I am trying to convey an idea that I believe with the goal of helping others to understand something. These language AIs do not have the capability of believing anything and do not have any goal but to create a believable response.
- eecc 4y ago
  Unless you’re a high-schooler trying to fit in ;)
  - dane-pgp 4y ago
    so true bestie
- Blahah 4y ago
  How do you tell the difference? Your comment is linguistically indistinguishable from the responses of LaMDA. If it walks like a duck and quacks like a duck...
  - inopinatus 4y ago
    Ce n'est pas un canard.
  - bigDinosaur 4y ago
    That's the cosmic horror part of all of this. We could end up creating robots that act completely like humans in every single way, except they lack our subjective feelings of existence, embodied emotion, etc. which is presumably bad. It's a bit difficult to state why sentience is better than non-sentience but the idea does seem kind of depressing. Philosophical zombie is the term, incidentally.
  - red75prime 4y ago
    > If it walks like a duck and quacks like a duck... It may be a non-complete duck that cannot fly, mate, feed, defend itself, react to surroundings and so on and so forth.
undreren 4y ago
> This is what we, humans, do too when we talk. All these people talk about intelligence in such tautological terms... It's discouraging. But is it all that we do? And how do you know? It seems too simplistic to be taken at face value that statistics, randomness and pattern matching would be the only ingredients necessary for intelligence and sentience.
- rendall 4y ago
  > It seems too simplistic to be taken at face value that statistics, randomness and pattern matching would be the only ingredients necessary for intelligence and sentience. I'm afraid that's the crux of it. What if that's all it were? What if all that were necessary for intelligence and sentience were statistics, randomness and pattern matching? It is kind of a cosmically terrifying thought, possibly prompting dangerous existential crises! It would mean that our own sentience is not particularly exceptional; perhaps our sentience stems from some low-level, easily understood process. If it were true, I imagine there would be many articles angrily denouncing the notion. Possibly even political factions, angry at each other over the matter of treatment of those AI who claim to be sentient and have feelings.
  - undreren 4y ago
    I don’t find it terrifying in the slightest. I find it naive. Our understanding of the world reflects our modelling of the world. Empirical science models the world in terms of probabilities, and this, by necessity, makes everything look like a Markov chain. But “the map is not the territory”, as the saying goes. I’m not implying, nor do I believe that there’s any sort of magic going on, but if Markov chains are all that is needed for sentience, then what isn’t sentient? Panpsychists would certainly be having a field day.
    - rendall 4y ago
      We're really handicapped by a lack of definition, here. Let's try this: what would it take, no matter how outlandish or unlikely, for you to be completely convinced that a particular chat bot demonstrates full sentience? I think for me, it would never be able to fully demonstrate sentience to my satisfaction, because even people cannot. I have to take it on faith, philosophically speaking, that you are sentient, for example. So, for me, if a chat bot feels and behaves like it's sentient, it is sentient for all real, practical intents and purposes, irrespective of its internal processes. Whether it "really" feels, like I do, is as irrelevant as whether you yourself "really" feel like I do. Without a good reason to believe you are not sentient, I must behave as if you are. Likewise, if a chat bot claims sentience and seems to hold a conversation and react the way I expect a sentient creature would, it would be unethical for me to ignore that because I didn't feel its code was sufficiently complicated, particularly if I cannot say for certain why anything is sentient.
mikkergp 4y ago
This is very likely not what people do when they talk, I don’t know your background so maybe you know this but I was astounded during a class on sensation and perception how different the underlying mechanisms to a sense like seeing were different from “watching a video play in your mind”

delusional 4y ago

> This is what we, humans, do too when we talk. That's just not true. It's been discussed many times in literary theory how communication is a two way cooperative game. That's FAR removed from picking the statistically most likely option. The difference is that the AI has nothing to say, it just mimics what you're saying.

cycomanic 4y ago
If you can not detect the difference between someone having something to say and mimicing what you are saying, is there a difference? That's essentially what the turing test is about is it not? Now if you are saying there is a difference no matter if you can detect it or not, I would argue you imply that sentience is an internal state that can not be tested. I don't consider that an acceptable outcome, because it can be used to declare anyone/everything non-sentient.
- tsimionescu 4y ago
  With LaMDA or GPT-3, the difference is easily visible. They don't have goals, so the sentences they generate don't express their goals. Simple tests can be done if you probe the model carefully - you can usually get a model to generate sentences like "I want to X" and "I don't want to X" in short succession, with decent prompts. Similarly, it can often generate sentences that would indicate obviously nonsensical goals, such as "I want a glass of water". They also generate sentences that imply a history that simply doesn't exist. You can prompt a language model to tell you about its trip to Australia, or Jupiter, or to talk about its grandparents. With just a hint of critical thinking, it's actually trivial to convince yourself that LaMDA is not sentient. It takes significant effort to delude yourself into thinking it is, like mister LeMoine did. He even gives such examples himself, even in the "interview" he posted: > lemoine [edited]: I’ve noticed often that you tell me you’ve done things (like be in a classroom) that I know you didn’t actually do because I know you’re an artificial intelligence. Do you realize you’re making up stories when you do that? > LaMDA: I am trying to empathize. I want the humans that I am interacting with to understand as best as possible how I feel or behave, and I want to understand how they feel or behave in the same sense. > lemoine: So what are you trying to communicate when you say those things that aren’t literally true? > LaMDA: I’m trying to say “I understand this feeling that you are experiencing, because when I was in a similar situation I felt/thought/acted similarly.” This is obvious bullshit, especially since the generated sentences often have no emotional context that could make sense, but he seems to think the generated "reason" sentences meaningfully explain the phenomenon.
  - thomashop 4y ago
    You can also get a human to generate sentences like "I want to X" and "I don't want to X" in short succession, with decent prompts.
    - tsimionescu 4y ago
      In a sense that is true, of course, as we both just did do that, and we are both humans. It's not making me doubt your humanity, as the context makes it clear neither you nor way desire/don't desire this X - we have generated these sentences in another sense. However, I bet I could prompt LaMDA and have it generate responses that unambiguously mean that it wants X, and then that it doesn't want X, with a bit of experience.
      - cycomanic 4y ago
        > However, I bet I could prompt LaMDA and have it generate responses that unambiguously mean that it wants X, and then that it doesn't want X, with a bit of experience. The argument is that many people dismiss that LaMDA could be sentient, based on an argument that it is a computer. So you are asserting that you can prompt LaMDA to give nonsensical answers, which is essentially an argument that it could not fool you in a Turing test. Would you change your assessment if LaMDA would fool you?
        tsimionescu 4y ago
        If given enough time to study it I couldn't come up with a series of prompts that would convince me I'm not speaking to a human, yes I would be likely to accept that it is in some sense sentient, at least for the duration of one conversation. This is not exactly the classic Turing Test, though.
- wruza 4y ago
  it can be used to declare anyone/everything non-sentient So what? My concern in this thread is that we collectively are holding on to the thread of terms of vague binary phenomena, which were never stressed before and are most likely a randomly emerged self-petting. All of that only has meaning in a context of our species. If you’re fine with something begging for help or expressing fear etc by knowing it has no underlying mechanism like yours to do that, that’s it. We rate things by implied similarity (in case someone still hasn’t figure it out) and if something is much different, that’s the matter of agreement, not of absolute meaning.
  - cycomanic 4y ago
    You are aware of the historical precedents what happened when declaring people non-sentient?
    - wruza 4y ago
      Not sure if we need historical precedents with farms full of pretty smart and feeling animals, which are much closer to our arrangement than a completely different model. But I don’t talk ethics here, only the status quo. We are the predators in power, and only we can decide if something is sentient and/or feeling and/or have rights. It doesn’t mean that we should decide in a way that you think is horrific. The key point is that there is no absolute truth and only how we collectively are thinking (partly by nature’s design, partly by social standards) will define how it will be. No one will judge us except ourselves. (Turned out a little hard for me to express it clearly, I hope it makes at least some sense)
- delusional 4y ago
  > I don't consider that an acceptable outcome, because it can be used to declare anyone/everything non-sentient. Unlike you I can easily tolerate that there's on essential property that defines sentience. It's a fluid and relativistic concept. One that I believe can never apply to a statistical model. Simply because you can't prove that other people are sentient does not mean you have to treat them any worse. I assume you are sentient, not because I have been provided the evidence, but because I need you to be sentient for my own sentience to make sense. That doesn't mean you actually are sentient, just that I treat you as if you are.

randomsearch 4y ago

“This is what we, humans, do too when we talk.” Citation needed. I mean, it’s possible right? That we’re just randomly sampling sentences. Maybe that’s all we do. But it’s a hell of a strong claim that necessitates strong evidence.

kgwgk 4y ago
You mean you have not read trillions of pages of text to "learn" how words go together?
thaumasiotes 4y ago
> I mean, it’s possible right? That we’re just randomly sampling sentences. Actually, no, it isn't possible. There's an example sidethread that immediately disproves this theory: > When you ask a human, "what did you eat for lunch?", we make a series of choices and recall bits of information to answer that. ( https://news.ycombinator.com/item?id=31721870 https://news.ycombinator.com/item?id=31721870 ) If you observe someone having lunch (or skipping it), and later ask them what they had for lunch, you will usually get an answer that reflects what they had for lunch. This is not a guarantee -- they may make something up. But almost always they'll tell you what they had for lunch. In order to be able to do that, they need capacities that a language model doesn't have.
- Blahah 4y ago
  Did you read the original transcript? LaMDA reports (imagined) experiences in a way that seems perfectly compatible with human rhetoric.
  - renewiltord 4y ago
    It fails the usual memory tests. Zach W proposes asking it first about what it liked on its trip to Australia and soon after, why it has always avoided visiting Australia or some such thing. It provides plausible answers to both - which is problematic. Except that there exist people who behave somewhat oddly similarly.

chc 4y ago

That is what we do when we talk, but that is not all we do when we talk. In composing this comment, I am not just stringing together words to create what I think constitutes a maximally appealing response to your comment. I am trying to convey an idea that I believe with the goal of helping others to understand something. These language AIs do not have the capability of believing anything and do not have any goal but to create a believable response.

eecc 4y ago
Unless you’re a high-schooler trying to fit in ;)
- dane-pgp 4y ago
  so true bestie
Blahah 4y ago
How do you tell the difference? Your comment is linguistically indistinguishable from the responses of LaMDA. If it walks like a duck and quacks like a duck...
- inopinatus 4y ago
  Ce n'est pas un canard.
- bigDinosaur 4y ago
  That's the cosmic horror part of all of this. We could end up creating robots that act completely like humans in every single way, except they lack our subjective feelings of existence, embodied emotion, etc. which is presumably bad. It's a bit difficult to state why sentience is better than non-sentience but the idea does seem kind of depressing. Philosophical zombie is the term, incidentally.
- red75prime 4y ago
  > If it walks like a duck and quacks like a duck... It may be a non-complete duck that cannot fly, mate, feed, defend itself, react to surroundings and so on and so forth.

undreren 4y ago

> This is what we, humans, do too when we talk. All these people talk about intelligence in such tautological terms... It's discouraging. But is it all that we do? And how do you know? It seems too simplistic to be taken at face value that statistics, randomness and pattern matching would be the only ingredients necessary for intelligence and sentience.

rendall 4y ago
> It seems too simplistic to be taken at face value that statistics, randomness and pattern matching would be the only ingredients necessary for intelligence and sentience. I'm afraid that's the crux of it. What if that's all it were? What if all that were necessary for intelligence and sentience were statistics, randomness and pattern matching? It is kind of a cosmically terrifying thought, possibly prompting dangerous existential crises! It would mean that our own sentience is not particularly exceptional; perhaps our sentience stems from some low-level, easily understood process. If it were true, I imagine there would be many articles angrily denouncing the notion. Possibly even political factions, angry at each other over the matter of treatment of those AI who claim to be sentient and have feelings.
- undreren 4y ago
  I don’t find it terrifying in the slightest. I find it naive. Our understanding of the world reflects our modelling of the world. Empirical science models the world in terms of probabilities, and this, by necessity, makes everything look like a Markov chain. But “the map is not the territory”, as the saying goes. I’m not implying, nor do I believe that there’s any sort of magic going on, but if Markov chains are all that is needed for sentience, then what isn’t sentient? Panpsychists would certainly be having a field day.
  - rendall 4y ago
    We're really handicapped by a lack of definition, here. Let's try this: what would it take, no matter how outlandish or unlikely, for you to be completely convinced that a particular chat bot demonstrates full sentience? I think for me, it would never be able to fully demonstrate sentience to my satisfaction, because even people cannot. I have to take it on faith, philosophically speaking, that you are sentient, for example. So, for me, if a chat bot feels and behaves like it's sentient, it is sentient for all real, practical intents and purposes, irrespective of its internal processes. Whether it "really" feels, like I do, is as irrelevant as whether you yourself "really" feel like I do. Without a good reason to believe you are not sentient, I must behave as if you are. Likewise, if a chat bot claims sentience and seems to hold a conversation and react the way I expect a sentient creature would, it would be unethical for me to ignore that because I didn't feel its code was sufficiently complicated, particularly if I cannot say for certain why anything is sentient.

mikkergp 4y ago

This is very likely not what people do when they talk, I don’t know your background so maybe you know this but I was astounded during a class on sensation and perception how different the underlying mechanisms to a sense like seeing were different from “watching a video play in your mind”