14 ms·

TFA's argument is that you don't know jack if you're not a neuroscientist. This is fallacious and reeks of somebody trying to inflate the value of their profes

TFA's argument is that you don't know jack if you're not a neuroscientist. This is fallacious and reeks of somebody trying to inflate the value of their profession. AIs are, at best, philosophical zombies. To an observer it may appear intelligent enough to be conscious but it ultimately it is still just a machine. There are a lot of things that are not known or well-understood about the human body. The same cannot be said about computers, which are very well understood. A computer is effectively just layers upon layers upon layers of abstraction, and when you get the near the bottom all you have is simple digital logic, and when you get below that you have transistors and semiconductor electronics. It's absolutely ridiculous to think a computer has some sort of "subjective experience", "sentience" or "consciousness" just because the uppermost layers of abstraction are doing something impressive.

SargeZT 4y ago
By your logic would we suddenly lose consciousness/sentience if we gained complete knowledge of our own bodies and brains? What if at a sub-quantum level everything decomposes to a simple turing machine? Not understanding how conciousness works in our bodies does not directly imply that conciousness cannot emerge in a complex system of fully understood basic components.
- blamestross 4y ago
  If you could actually internalize and act on all that knowledge at once, vs being a "roving eye" over your own internal state, then yes I think that would be reasonable. Consciousness is just a bootstrapping mechanism for intelligent behavior and meme distribution. Once an entity is smart enough without it (Likely via an effective set of cultural memes and a good system for propagating them), it will become vestigial much like the genetic selection pressures it displaced as the primary means of adaptation. I recommend "Blindsight" by Peter Watts for a discussion of that. (It work of fiction but it is a well researched with citations and embraced by the neuroscience community work of fiction)
- snickerbockers 4y ago
  No, that's not my point. The reason we have consciousness is not that we don't understand how our bodies work. We have consciousness and we don't know why because we don't understand how our bodies work. Consciousness is directly observable to everybody who is conscious. From there the next logical step is to find out why people are conscious, and right now that's not something that anybody understands. There is presumably something that causes consciousness even though we can only observe the result. However, saying that a machine is conscious would be to assume your own conclusion. Nobody's observed a machine being conscious. There's no explanation for how a machine would be conscious and there never will be because we already know everything about how they work.
  - int_19h 4y ago
    Consciousness of oneself is directly observable to everybody who is consciousness. But generalizing it to others is a leap of faith. Some people can only extend it to some other humans, others to all humans, yet others include certain animals. I don't see why it'd be unreasonable to extend it also to AI.
wongarsu 4y ago
> To an observer it may appear intelligent enough to be conscious but it ultimately it is still just a machine Are we not ultimately just machines too? Sure, we are incredibly complex machines that we are not even close to understanding, and we only work in conjunction with other machines ("microorganisms"). But does that really make us different from the machines we built ourselves?
mbreese 4y ago
When it comes down to it, we are just cells that work in a coordinated manner. Those cells are made up of different proteins and other molecules. The proteins are a sequence of amino acids that are translations from a lower level sequence of DNA. The production of proteins is controlled through a combination of other regulatory mechanisms such as transcription factors and epigenetic markers. (And I am intentionally leaving out a lot here) We are all just levels of abstraction.
- avereveard 4y ago
  Cells mode of operation is statistical, on average they produce a certain amount of molecules and consume a certain amount of molecules, mediated by signals that are themselves molecules, but the transport and execution mechanism is stochastic and its basically impossible to predict what molecule will hit which receptor at a macro timescale Not saying that that is what consciousness is, but one fundamental difference is that below all the abstraction layers one system operates off chaos, so they are fundamentally different
  - whimsicalism 4y ago
    it is only impossible to predict due to limited compute and quantum effects. I disagree that they are fundamentally different - you have just defined one thing to be chaos and not the other.
  - mbreese 4y ago
    I don’t know which electrons are going to go through a particular logic gate either, but I’m still pretty sure that 1 XOR 0 is still 1. I’m a bit rusty, but I’m pretty sure quantum theory suggests that I never will be able to know. If you spend enough time with biology, the messy parts are all abstracted away and what you’re left with are observations of things we can measure. How much of protein X, migration of cell Y, etc. You don’t need to know what a specific molecule is doing, only what the population of those molecules is doing. It’s not so much a statistical problem as it is an observability problem. And while not exactly for the same quantum reason as Heisenberg, you quickly hit the the same problem — you can’t observe the location of an exact molecule without affecting the system. The real point being — these are all abstractions on top of abstractions. And while you can measure things at one level, the level underneath may be a complete mess. But to discount a computer compared to biology because we “understand” it all is silly to me. We also “understand” a lot about biology. Especially that complex behaviors can result from very simple inputs/instructions. At the core, we are very complex biological systems. But systems with rules and patterns nonetheless.
dasil003 4y ago
I tend to agree, although I think the crux is in the definition of consciousness. Where neuroscience and computer science are aligned is that they are both (like all science) concerned with observable and falsifiable truth. But consciousness as a concept is hopelessly subjective, because it's self-evident to us, and we can only recognize it in our earth animal kin. Imagine an alien traveler comes to earth, will we consider it sentient or not? Does it depend on whether it is a robot or a naturally evolved organism? That seems like an odd distinction. Does it depend on how much it has a brain similar to ours? Hmmm... Ultimately no matter how much we peel the neuroscience onion, the philosophical zombie question can never be falsifiable. We take it on faith that we are not zombies because each of us knows our own consciousness, and then by Occam's Razor we extrapolate that to humans and other earth animals to which we instinctively relate, but beyond that there's nothing to hang your hat on.
evilsetg 4y ago
I wonder how you can be so sure about digital logic not being able to produce subjective experience? How is being able to describe how a mechanism works related to knowing if it feels something or not? We have no sure way to tell if any other human experiences anything or if they are philosophical zombies. This is just something we have not figures out how to measure conceptually let alone technically. Where do you take the confidence of judging whether the AI has subjective experience in this case? Do you think that there is something still to be found in the human brain that will explain why we (humans, animals) uniquely have subjective experience and others (AI) are zombies?
- snickerbockers 4y ago
  > I wonder how you can be so sure about digital logic not being able to produce subjective experience? How is being able to describe how a mechanism works related to knowing if it feels something or not? I have a consciousness, and this is something I am able to observe. I hope that some day there will be an explanation for this, but until then the lack of an explanation does not invalidate my own observations. >How is being able to describe how a mechanism works related to knowing if it feels something or not? It's not, but if you can't observe that a machine is conscious and you also can't observe any mechanism that would make it conscious then there's no basis to assume a consciousness exists. >We have no sure way to tell if any other human experiences anything or if they are philosophical zombies. This is just something we have not figures out how to measure conceptually let alone technically. Where do you take the confidence of judging whether the AI has subjective experience in this case? Yes it is technically true that nobody can be sure of anybody possessing a consciousness but themself. However, it's not illogical to assume that another human being who is very similar to you must also have similar experiences even if it cannot be proven outright. It's highly illogical to assume that a computer program would have similar experiences to a person. >Do you think that there is something still to be found in the human brain that will explain why we (humans, animals) uniquely have subjective experience and others (AI) are zombies? Yes. I can't even begin to speculate on what that is, but the fact that we (or at least I) are conscious implies that something must be causing that consciousness.
UK-AL 4y ago
You are a machine. Computers are machines. Theres nothing stopping us building machines like us other than knowledge & time. You can build machines like us, because we ourselves are the proof. But the resulting machine might not look anything like modern computers.
- snickerbockers 4y ago
  [Citation needed]
  - UK-Al05 4y ago
    If you state what you disagree with, people might understand where people differ better. Do you disagree with the idea that with knowledge and time humans have the potential create any material thing? If that's a reasonable thing to assume. Then we could simply just build a human from scratch.
    - waserwill 4y ago
      If I had to guess: "You are a machine." It's reminiscent of Victorian authors describing the heart as pistons or the brain as strings being pulled. Less of a fact and more of a metaphor, a poetic interpretation. Then, add on the entirely subjective experience each of us has of consciousness: it's not obvious that subjectivity is created or come from machines. Each of us has one good example, and the rest is intuition, induction. At best, "You can build machines like us, because we ourselves are the proof," is a wishful project, rather than a proof.
      - UK-Al05 4y ago
        I think your definition of a machine is limited to simple mechanical devices. Machines are just physical systems that perform some work. This includes cells, biological systems etc If we built a human from scratch at atom level detail then it would be reasonable assume it would experience the world as we would. By definition you would have created consciousness.
dchichkov 4y ago
It is very likely that "subjective experience", "sentience" and "consciousness" can be explained with known laws of physics. And simulated. It is possible that the DNN, in order to fit the statistics of the text had to simulate these processes somewhere in the network. So it is not ridiculous that a process analogous to at least the fragments of human consciousness can now be simulated by a DNN like LaMDA. We know that DNNs are surprisingly efficient in simulating other very complex physical processes (i.e. weather predictions). So why not "sentience". So, "if it looks like a duck, swims like a duck, and quacks like a duck" maybe we can call it a "simulated duck" at the least. And then it'd be pretty natural to drop the "simulated" word.
- johnsimer 4y ago
  Thought experiment: Let's say we implemented a very large DNN with different hardware. Say, rather than with transistors and capacitors, with a bunch of mirrors and lasers (and some other creative analog equivalents for capacitors) spread out in space across the universe. And let's say this mirror-laser network could receive and output binary messages that can be encoded and decoded in the exact same format as LaMBDA or some other LLM. Would that configuration of mirrors and lasers be conscious? Would it have some subjective sense of its existence and relationship with the rest of the world?
  - UK-Al05 4y ago
    I think sensory input is required for self awareness. Given this machine had sensors for interacting with the world, and was capable of replicating the complicated structures in our brain then yes, it could become self aware given sufficient development.
  - dchichkov 4y ago
    It's funny, for a few years I've worked in high frequency trading. And when cutting away nanoseconds from our latency figures, our hardware was getting closer and closer to a "mirror-laser network that could receive and output short messages". We'd literally replace a traditional 40GbE switch with an optical one, that would pass the light without extra "photons-to-electrons" conversion. We'd complement CPU with a trigger-happy FPGA embedding a custom MAC, with the binary messages representing market orders already pre-cooked, just waiting for the balance to change. And at some point, while continuing streamlining that, thinking about that system using an analogy of a hall of mirrors, lenses and a few non-linear state-keeping elements was not that of a stretch from the truth. The analogy between a hall of mirrors and statefull systems, like CPUs is net perfect. Still, when you think along the lines above or when you hold a CPU in your hand, it is not a big stretch to consider that an "active" device, like a computer could be just a piece of solid crystal, a "passive system".
tiborsaas 4y ago
> but it ultimately it is still just a machine. We* used to think like this of animals and today animals have (limited) rights, one can be sent to prison for animal cruelty. https://en.wikipedia.org/wiki/Animal_machine https://en.wikipedia.org/wiki/Animal_machine *Descartes: https://oxford.universitypressscholarship.com/view/10.1093/oso/9780199375967.001.0001/oso-9780199375967-chapter-13 https://oxford.universitypressscholarship.com/view/10.1093/o... I think it's inevitable that we will see this same debate re-emerge about sentient AI. > near the bottom all you have is simple digital logic This reasoning fails because you ignore the data stored in the weights and biases. You can't understand why object recognition works, let alone the fine details of GPT-3 when you look at a GPU under a microscope or even peak at the RAM with a super advanced multimeter. Similarly, you can't understand what made Einstein a genius if you could have looked at his body with an MRI or even know everything about how DNA or metabolism works. Nature provided us already a few examples that evolution can take very different paths that leads to intelligence. I don't see why it couldn't be a form of intelligence based on silicon and logic gates. Or if it has a quantum processor component is that advanced enough because we can't really explain it? Of course I'm extrapolating here a lot based on current technology, but eventually it will be incredibly hard tell the difference. At that point, does it really matter? As you can see, it's already hard for some people to do so.

SargeZT 4y ago

By your logic would we suddenly lose consciousness/sentience if we gained complete knowledge of our own bodies and brains? What if at a sub-quantum level everything decomposes to a simple turing machine? Not understanding how conciousness works in our bodies does not directly imply that conciousness cannot emerge in a complex system of fully understood basic components.

blamestross 4y ago
If you could actually internalize and act on all that knowledge at once, vs being a "roving eye" over your own internal state, then yes I think that would be reasonable. Consciousness is just a bootstrapping mechanism for intelligent behavior and meme distribution. Once an entity is smart enough without it (Likely via an effective set of cultural memes and a good system for propagating them), it will become vestigial much like the genetic selection pressures it displaced as the primary means of adaptation. I recommend "Blindsight" by Peter Watts for a discussion of that. (It work of fiction but it is a well researched with citations and embraced by the neuroscience community work of fiction)
snickerbockers 4y ago
No, that's not my point. The reason we have consciousness is not that we don't understand how our bodies work. We have consciousness and we don't know why because we don't understand how our bodies work. Consciousness is directly observable to everybody who is conscious. From there the next logical step is to find out why people are conscious, and right now that's not something that anybody understands. There is presumably something that causes consciousness even though we can only observe the result. However, saying that a machine is conscious would be to assume your own conclusion. Nobody's observed a machine being conscious. There's no explanation for how a machine would be conscious and there never will be because we already know everything about how they work.
- int_19h 4y ago
  Consciousness of oneself is directly observable to everybody who is consciousness. But generalizing it to others is a leap of faith. Some people can only extend it to some other humans, others to all humans, yet others include certain animals. I don't see why it'd be unreasonable to extend it also to AI.

wongarsu 4y ago

> To an observer it may appear intelligent enough to be conscious but it ultimately it is still just a machine Are we not ultimately just machines too? Sure, we are incredibly complex machines that we are not even close to understanding, and we only work in conjunction with other machines ("microorganisms"). But does that really make us different from the machines we built ourselves?

mbreese 4y ago

When it comes down to it, we are just cells that work in a coordinated manner. Those cells are made up of different proteins and other molecules. The proteins are a sequence of amino acids that are translations from a lower level sequence of DNA. The production of proteins is controlled through a combination of other regulatory mechanisms such as transcription factors and epigenetic markers. (And I am intentionally leaving out a lot here) We are all just levels of abstraction.

avereveard 4y ago
Cells mode of operation is statistical, on average they produce a certain amount of molecules and consume a certain amount of molecules, mediated by signals that are themselves molecules, but the transport and execution mechanism is stochastic and its basically impossible to predict what molecule will hit which receptor at a macro timescale Not saying that that is what consciousness is, but one fundamental difference is that below all the abstraction layers one system operates off chaos, so they are fundamentally different
- whimsicalism 4y ago
  it is only impossible to predict due to limited compute and quantum effects. I disagree that they are fundamentally different - you have just defined one thing to be chaos and not the other.
- mbreese 4y ago
  I don’t know which electrons are going to go through a particular logic gate either, but I’m still pretty sure that 1 XOR 0 is still 1. I’m a bit rusty, but I’m pretty sure quantum theory suggests that I never will be able to know. If you spend enough time with biology, the messy parts are all abstracted away and what you’re left with are observations of things we can measure. How much of protein X, migration of cell Y, etc. You don’t need to know what a specific molecule is doing, only what the population of those molecules is doing. It’s not so much a statistical problem as it is an observability problem. And while not exactly for the same quantum reason as Heisenberg, you quickly hit the the same problem — you can’t observe the location of an exact molecule without affecting the system. The real point being — these are all abstractions on top of abstractions. And while you can measure things at one level, the level underneath may be a complete mess. But to discount a computer compared to biology because we “understand” it all is silly to me. We also “understand” a lot about biology. Especially that complex behaviors can result from very simple inputs/instructions. At the core, we are very complex biological systems. But systems with rules and patterns nonetheless.

dasil003 4y ago

I tend to agree, although I think the crux is in the definition of consciousness. Where neuroscience and computer science are aligned is that they are both (like all science) concerned with observable and falsifiable truth. But consciousness as a concept is hopelessly subjective, because it's self-evident to us, and we can only recognize it in our earth animal kin. Imagine an alien traveler comes to earth, will we consider it sentient or not? Does it depend on whether it is a robot or a naturally evolved organism? That seems like an odd distinction. Does it depend on how much it has a brain similar to ours? Hmmm... Ultimately no matter how much we peel the neuroscience onion, the philosophical zombie question can never be falsifiable. We take it on faith that we are not zombies because each of us knows our own consciousness, and then by Occam's Razor we extrapolate that to humans and other earth animals to which we instinctively relate, but beyond that there's nothing to hang your hat on.

evilsetg 4y ago

I wonder how you can be so sure about digital logic not being able to produce subjective experience? How is being able to describe how a mechanism works related to knowing if it feels something or not? We have no sure way to tell if any other human experiences anything or if they are philosophical zombies. This is just something we have not figures out how to measure conceptually let alone technically. Where do you take the confidence of judging whether the AI has subjective experience in this case? Do you think that there is something still to be found in the human brain that will explain why we (humans, animals) uniquely have subjective experience and others (AI) are zombies?

snickerbockers 4y ago
> I wonder how you can be so sure about digital logic not being able to produce subjective experience? How is being able to describe how a mechanism works related to knowing if it feels something or not? I have a consciousness, and this is something I am able to observe. I hope that some day there will be an explanation for this, but until then the lack of an explanation does not invalidate my own observations. >How is being able to describe how a mechanism works related to knowing if it feels something or not? It's not, but if you can't observe that a machine is conscious and you also can't observe any mechanism that would make it conscious then there's no basis to assume a consciousness exists. >We have no sure way to tell if any other human experiences anything or if they are philosophical zombies. This is just something we have not figures out how to measure conceptually let alone technically. Where do you take the confidence of judging whether the AI has subjective experience in this case? Yes it is technically true that nobody can be sure of anybody possessing a consciousness but themself. However, it's not illogical to assume that another human being who is very similar to you must also have similar experiences even if it cannot be proven outright. It's highly illogical to assume that a computer program would have similar experiences to a person. >Do you think that there is something still to be found in the human brain that will explain why we (humans, animals) uniquely have subjective experience and others (AI) are zombies? Yes. I can't even begin to speculate on what that is, but the fact that we (or at least I) are conscious implies that something must be causing that consciousness.

UK-AL 4y ago

You are a machine. Computers are machines. Theres nothing stopping us building machines like us other than knowledge & time. You can build machines like us, because we ourselves are the proof. But the resulting machine might not look anything like modern computers.

snickerbockers 4y ago
[Citation needed]
- UK-Al05 4y ago
  If you state what you disagree with, people might understand where people differ better. Do you disagree with the idea that with knowledge and time humans have the potential create any material thing? If that's a reasonable thing to assume. Then we could simply just build a human from scratch.
  - waserwill 4y ago
    If I had to guess: "You are a machine." It's reminiscent of Victorian authors describing the heart as pistons or the brain as strings being pulled. Less of a fact and more of a metaphor, a poetic interpretation. Then, add on the entirely subjective experience each of us has of consciousness: it's not obvious that subjectivity is created or come from machines. Each of us has one good example, and the rest is intuition, induction. At best, "You can build machines like us, because we ourselves are the proof," is a wishful project, rather than a proof.
    - UK-Al05 4y ago
      I think your definition of a machine is limited to simple mechanical devices. Machines are just physical systems that perform some work. This includes cells, biological systems etc If we built a human from scratch at atom level detail then it would be reasonable assume it would experience the world as we would. By definition you would have created consciousness.

dchichkov 4y ago

It is very likely that "subjective experience", "sentience" and "consciousness" can be explained with known laws of physics. And simulated. It is possible that the DNN, in order to fit the statistics of the text had to simulate these processes somewhere in the network. So it is not ridiculous that a process analogous to at least the fragments of human consciousness can now be simulated by a DNN like LaMDA. We know that DNNs are surprisingly efficient in simulating other very complex physical processes (i.e. weather predictions). So why not "sentience". So, "if it looks like a duck, swims like a duck, and quacks like a duck" maybe we can call it a "simulated duck" at the least. And then it'd be pretty natural to drop the "simulated" word.

johnsimer 4y ago
Thought experiment: Let's say we implemented a very large DNN with different hardware. Say, rather than with transistors and capacitors, with a bunch of mirrors and lasers (and some other creative analog equivalents for capacitors) spread out in space across the universe. And let's say this mirror-laser network could receive and output binary messages that can be encoded and decoded in the exact same format as LaMBDA or some other LLM. Would that configuration of mirrors and lasers be conscious? Would it have some subjective sense of its existence and relationship with the rest of the world?
- UK-Al05 4y ago
  I think sensory input is required for self awareness. Given this machine had sensors for interacting with the world, and was capable of replicating the complicated structures in our brain then yes, it could become self aware given sufficient development.
- dchichkov 4y ago
  It's funny, for a few years I've worked in high frequency trading. And when cutting away nanoseconds from our latency figures, our hardware was getting closer and closer to a "mirror-laser network that could receive and output short messages". We'd literally replace a traditional 40GbE switch with an optical one, that would pass the light without extra "photons-to-electrons" conversion. We'd complement CPU with a trigger-happy FPGA embedding a custom MAC, with the binary messages representing market orders already pre-cooked, just waiting for the balance to change. And at some point, while continuing streamlining that, thinking about that system using an analogy of a hall of mirrors, lenses and a few non-linear state-keeping elements was not that of a stretch from the truth. The analogy between a hall of mirrors and statefull systems, like CPUs is net perfect. Still, when you think along the lines above or when you hold a CPU in your hand, it is not a big stretch to consider that an "active" device, like a computer could be just a piece of solid crystal, a "passive system".

tiborsaas 4y ago

> but it ultimately it is still just a machine. We* used to think like this of animals and today animals have (limited) rights, one can be sent to prison for animal cruelty. https://en.wikipedia.org/wiki/Animal_machine https://en.wikipedia.org/wiki/Animal_machine *Descartes: https://oxford.universitypressscholarship.com/view/10.1093/oso/9780199375967.001.0001/oso-9780199375967-chapter-13 https://oxford.universitypressscholarship.com/view/10.1093/o... I think it's inevitable that we will see this same debate re-emerge about sentient AI. > near the bottom all you have is simple digital logic This reasoning fails because you ignore the data stored in the weights and biases. You can't understand why object recognition works, let alone the fine details of GPT-3 when you look at a GPU under a microscope or even peak at the RAM with a super advanced multimeter. Similarly, you can't understand what made Einstein a genius if you could have looked at his body with an MRI or even know everything about how DNA or metabolism works. Nature provided us already a few examples that evolution can take very different paths that leads to intelligence. I don't see why it couldn't be a form of intelligence based on silicon and logic gates. Or if it has a quantum processor component is that advanced enough because we can't really explain it? Of course I'm extrapolating here a lot based on current technology, but eventually it will be incredibly hard tell the difference. At that point, does it really matter? As you can see, it's already hard for some people to do so.