4 ms·

This is not a tenable argument if you are a materialist. Either you accept the idea that human sentience is attributable to a soul (and therefore cease to be a

This is not a tenable argument if you are a materialist. Either you accept the idea that human sentience is attributable to a soul (and therefore cease to be a materialist), you decide that the outward appearance of sentience is sufficient (which you might find to be troubling due to its potential implications for the insane), or you say that sentience is unknowable. There's no fourth option.

greysphere 4y ago
Isn't the gp post claiming (2), and AI isn't close to appearing sentient? To do so it would have to demonstrate qualities x,y,z...
- antibasilisk 4y ago
  They can't be claiming 2 because will, intention etc are not empirically verifiable, they're opaque internal processes. What we can verify is whether it demonstrates reason and understanding, and it inarguably does both, which is why the researcher in question takes the position he does.
andrei_says_ 4y ago
How about sentience being product of the body, its nervous system, evolutionary mechanisms and complexity? As in - an emergent phenomenon made possible by the presence of all three? Nothing we’ve seen in the realm of ai touches even a speck of the brain’s complexity and ongoing re-configuration in the presence of change and in the context of survival. I’ve been reading George Lakoff’s work - turns out our ability to choose is based on an emotional component. Any rational decision is based on a preference, and all preferences are detected via the mechanism of emotion. People with certain brain traumas resulting in damaged emotional centers become unable to make choices.
- antibasilisk 4y ago
  Hmm I guess saying that it's an emergent property of certain things would work as a fourth option, but why would it have to be an emergent phenomenon of body, nervous system, evolutionary mechanism and complexity, as opposed to for example a nervous system and complexity? You'd also have to drill down into the definitions of body, nervous system, etc. Is a keyboard part of a nervous system? Is a computer tower a body? Why or why not? Would your criteria necessarily exclude the possibility of artificial sentience in the first place (although I'm not necessarily saying that's a flaw)?
- chpatrick 4y ago
  I have absolutely no idea how your nervous system works or how complex you are but I presume you're sentient because of how you communicate. Why is that different with an AI?
lamontcg 4y ago
I think I'm more than a language model which is just an algorithmically complicated madlib generator with a steepest descent algorithm. If that's all you think you are, then I can't help you.
- antibasilisk 4y ago
  I was merely outlining the positions that are taken regarding sentience. My personal position is that I'm Muslim, so a priori I do not believe machines can be sentient, regardless how intelligent they are, because divine revelation informs us of the existence of a soul, and of what does and does not have a soul.
snowwrestler 4y ago
The current argument is not “is artificial sentience generally impossible,” it is “are the machine learning systems we are currently building and operating sentient.” Arguing that they are not sentient is not in conflict with materialism. And in fact it is supported by tons of material evidence: for example they don’t act sentient at all if you decline to provide a prompt.
- antibasilisk 4y ago
  > The current argument is not “is artificial sentience generally impossible,” it is “are the machine learning systems we are currently building and operating sentient.” The argument is 'what is sentience?' >And in fact it is supported by tons of material evidence: for example they don’t act sentient at all if you decline to provide a prompt This falls into the category of 'appearance of sentience'. Based on this logic all that would be required would be a perpetually self-prompting InstructGPT instance.
  - snowwrestler 4y ago
    No, you’re trying to make the argument about “what is sentience?” The comment at the top of this thread, to which you replied, states “Chatbots are not sentient.” Which is a narrower conversation.
    - maxdoop 4y ago
      And how can that claim be made with confidence? We can’t even define our own sentience and consciousness (are those things the same?) and yet folks are constantly making a strong claim that LLMs are definitely , definitely not sentient. I have not seen one reason why that confidence should exist. People seem to think we are special because we are made of meat, and we are “more complex”. But I still ask — what is so different about the human brain than an LLM? And further, what proof is there that we know what sentience truly is, and that LLMs don’t have it?
andrewflnr 4y ago
False dichotomy. It's consistent to claim that sentience is an aspect of physical but not readily observable processes, where "not readily observable" could in this case be as trivial as "requires access to mindstate, so fMRI in humans" and still distinguish sentience from a thin appearance of it.
- antibasilisk 4y ago
  Not so, since without a living non-sentient brain for comparison, it would be impossible to distinguish the mindstate that produces sentience from random noise, and since we're presumably accepting that humans are sentient no such brain could exist in the first place.
hnfong 4y ago
The True Materialist will tell you that it's biological cells and DNA that gives rise to sentience. AIs made from artificial neural networks are actually not quite the materialistic thing. It's some spooky emergent property that doesn't depend on material substrates, you can run an AI on a Nvidia card for sure, but you can also run it in your head (theoretically, if you live forever). Or at least with the assistance of (a lot of) pen and paper. Where does the sentience live then, in your mind? That doesn't sound very materialistic to me :D (Note: I think the GP is full of it, but I just thought I'd set the record straight about materialism here :P )
- pksebben 4y ago
  FWIG one materialist view of sentience is that it's an emergent property - it is a way of describing a state of a complex system. The question, then, is: what are the properties of the system that the term describes? In this case, it looks kinda like: memory, self-awareness, and understanding. Memory isn't a question - it's possible for these things to remember (even if sometimes that memory is in the form of access to data on the internet). Self-awareness is trickier - but by most measurable schema, the machines seem to be doing as well as humans do in this regard (they can describe themselves and talk about their place in the world in context). Understanding is a topic of hot debate - but I think it tends to fall to the same problems as that of sentience generally; that is, we have this concept of what understanding is, and we know enough about how LLMs work to say that they're not the same, but if you start poking at how to measure it, we're back to square one vis a vis what's apparent might not be what is.

greysphere 4y ago

Isn't the gp post claiming (2), and AI isn't close to appearing sentient? To do so it would have to demonstrate qualities x,y,z...

antibasilisk 4y ago
They can't be claiming 2 because will, intention etc are not empirically verifiable, they're opaque internal processes. What we can verify is whether it demonstrates reason and understanding, and it inarguably does both, which is why the researcher in question takes the position he does.

andrei_says_ 4y ago

How about sentience being product of the body, its nervous system, evolutionary mechanisms and complexity? As in - an emergent phenomenon made possible by the presence of all three? Nothing we’ve seen in the realm of ai touches even a speck of the brain’s complexity and ongoing re-configuration in the presence of change and in the context of survival. I’ve been reading George Lakoff’s work - turns out our ability to choose is based on an emotional component. Any rational decision is based on a preference, and all preferences are detected via the mechanism of emotion. People with certain brain traumas resulting in damaged emotional centers become unable to make choices.

antibasilisk 4y ago
Hmm I guess saying that it's an emergent property of certain things would work as a fourth option, but why would it have to be an emergent phenomenon of body, nervous system, evolutionary mechanism and complexity, as opposed to for example a nervous system and complexity? You'd also have to drill down into the definitions of body, nervous system, etc. Is a keyboard part of a nervous system? Is a computer tower a body? Why or why not? Would your criteria necessarily exclude the possibility of artificial sentience in the first place (although I'm not necessarily saying that's a flaw)?
chpatrick 4y ago
I have absolutely no idea how your nervous system works or how complex you are but I presume you're sentient because of how you communicate. Why is that different with an AI?

lamontcg 4y ago

I think I'm more than a language model which is just an algorithmically complicated madlib generator with a steepest descent algorithm. If that's all you think you are, then I can't help you.

antibasilisk 4y ago
I was merely outlining the positions that are taken regarding sentience. My personal position is that I'm Muslim, so a priori I do not believe machines can be sentient, regardless how intelligent they are, because divine revelation informs us of the existence of a soul, and of what does and does not have a soul.

snowwrestler 4y ago

The current argument is not “is artificial sentience generally impossible,” it is “are the machine learning systems we are currently building and operating sentient.” Arguing that they are not sentient is not in conflict with materialism. And in fact it is supported by tons of material evidence: for example they don’t act sentient at all if you decline to provide a prompt.

antibasilisk 4y ago
> The current argument is not “is artificial sentience generally impossible,” it is “are the machine learning systems we are currently building and operating sentient.” The argument is 'what is sentience?' >And in fact it is supported by tons of material evidence: for example they don’t act sentient at all if you decline to provide a prompt This falls into the category of 'appearance of sentience'. Based on this logic all that would be required would be a perpetually self-prompting InstructGPT instance.
- snowwrestler 4y ago
  No, you’re trying to make the argument about “what is sentience?” The comment at the top of this thread, to which you replied, states “Chatbots are not sentient.” Which is a narrower conversation.
  - maxdoop 4y ago
    And how can that claim be made with confidence? We can’t even define our own sentience and consciousness (are those things the same?) and yet folks are constantly making a strong claim that LLMs are definitely , definitely not sentient. I have not seen one reason why that confidence should exist. People seem to think we are special because we are made of meat, and we are “more complex”. But I still ask — what is so different about the human brain than an LLM? And further, what proof is there that we know what sentience truly is, and that LLMs don’t have it?

andrewflnr 4y ago

False dichotomy. It's consistent to claim that sentience is an aspect of physical but not readily observable processes, where "not readily observable" could in this case be as trivial as "requires access to mindstate, so fMRI in humans" and still distinguish sentience from a thin appearance of it.

antibasilisk 4y ago
Not so, since without a living non-sentient brain for comparison, it would be impossible to distinguish the mindstate that produces sentience from random noise, and since we're presumably accepting that humans are sentient no such brain could exist in the first place.

hnfong 4y ago

The True Materialist will tell you that it's biological cells and DNA that gives rise to sentience. AIs made from artificial neural networks are actually not quite the materialistic thing. It's some spooky emergent property that doesn't depend on material substrates, you can run an AI on a Nvidia card for sure, but you can also run it in your head (theoretically, if you live forever). Or at least with the assistance of (a lot of) pen and paper. Where does the sentience live then, in your mind? That doesn't sound very materialistic to me :D (Note: I think the GP is full of it, but I just thought I'd set the record straight about materialism here :P )

pksebben 4y ago
FWIG one materialist view of sentience is that it's an emergent property - it is a way of describing a state of a complex system. The question, then, is: what are the properties of the system that the term describes? In this case, it looks kinda like: memory, self-awareness, and understanding. Memory isn't a question - it's possible for these things to remember (even if sometimes that memory is in the form of access to data on the internet). Self-awareness is trickier - but by most measurable schema, the machines seem to be doing as well as humans do in this regard (they can describe themselves and talk about their place in the world in context). Understanding is a topic of hot debate - but I think it tends to fall to the same problems as that of sentience generally; that is, we have this concept of what understanding is, and we know enough about how LLMs work to say that they're not the same, but if you start poking at how to measure it, we're back to square one vis a vis what's apparent might not be what is.