That doesn't seem to quite be "panpsychism". You seem to be (reasonably) restricting it to "systems made of brains that calculate what they would feel". The usual panpsychism instead proposes consciousness as a fundamental physical theory, where electrons and quarks would have it too. But electrons and quarks don't have brains, so wouldn't seem to qualify according to your description.
You seem instead to be describing something more like Computationalism: as long as you're doing the appropriate information processing, you're conscious.
My strong intuition is that p-zombies are not compatible with our universe. P-zombies are basically entities with indistinguishable outputs to the same inputs as conscious beings. It implies that consciousness has absolutely no utility in such a world. If a trait has no utility there's no pressure for natural selection to select for it. In such a world evolution would still produce intelligent beings, but it wouldn't select for counsciousness, the probability for such a complex and elusive trait as consciousness to emerge through evolution would be astronomically unlikely: we ourselves would be p-zombies in such worlds. Hence, p-zombies are not possible.
And if we assume that counsciousness isn't linked to any inputs, it would have to be a miracle that the sensation of pain somehow arises in response to threaths to the body, and we experience pleasure when we taste nutritious food.
This kind of argument doesn't go through, because a p-zombie doesn't differ from its conscious counterpart in any way that could possibly be relevant to natural selection. (By definition, it is physically identical and has the same behaviour.)
Sorry, rereading your comment I can see that I missed the point. I'm still tripped up by your claim that if it had no utility, "the probability for such a complex and elusive trait as consciousness to emerge through evolution would be astronomically unlikely". Why would it be any harder for evolution to produce conscious beings than p-zombies? From evolution's perspective, they are exactly the same thing. And in the world where physical properties aren't sufficient to produce consciousness, we have no idea what is, or how elusive it is.
I agree with a version of the argument that says: in my case, evolution produced a conscious being rather than a p-zombie, and so, in the absence of anything else to go on, I should assume that is probably the default.
And I think I understand this version: we have both A (the appearance of consciousness) and B (actual phenomenal consciousness). If we assume that A implies B, then our prior that A&B would co-occur in us is just the same as our prior that A would occur; but if we assume that A and B are unrelated, then our prior that A&B would co-occur is equal to the product of our prior that A would occur and our prior that B would occur. The same is true, to a lesser degree, of the cases where A is connected in some way to B but not a sufficient condition for it. So our observing A&B co-occurring is strong evidence in favour of the claim that A implies B.
Is that second version a reasonable summary of your argument? I do see the logic in that, but I think there are problems, mostly to do with our near-total ignorance of what we're even talking about in the case where p-zombies are possible. When we're talking about physical stuff, we can use heuristics like preferring simpler explanations, or modeling our ignorance as randomness, and where necessary we can pull priors out of the air and then update on evidence. But when we're talking about (potentially) wholly metaphysical stuff, I don't know if any of that is valid. And when we go back to the physical and observable, we're left in pure ignorance: nothing I observe could possibly distinguish the world where p-zombies are possible from the world where p-zombies are impossible. I have an existence proof of non-zombiehood, and I can try to use it as evidence in a Bayesian calculation, as above, but my priors have to be conjured out of absolutely nowhere and I will never get new evidence to confirm I was on the right track or correct my error.
I'm very interested also in seeing how the folks who embrace supernatural explanations of consciousness will approach AI as in silico models of intelligence continue to replicate more and more characteristics of what we have traditionally considered only human consciousness. It's at least in some sense easier for a supernaturalist to distinguish human from non-human consciousness by resort, for example, to the concept of the soul. But it's interesting that many of those same folks would acknowledge that animals and in some cases plants or even rocks or landscapes may be imbued with souls. To be consistent, I would expect that at least some of these same folks would have to accept (perhaps even happily?) that some in silico models of consciousness have souls? Or flip the other way and label them demons?
Precisely. Well put, Robin. The Zombie hypothesis is incoherent from the get go. It never made any claims that distinguished different physical worlds. I put the idea of "qualia" in the same bucket.
> It never made any claims that distinguished different physical worlds.
Yes, that's the whole point! It's a thought experiment used to make the point that a complete physical description of the world is not necessarily a complete description of the world.
"Qualia" doesn't seem so difficult. Surely you understand what it is to have an experience, in a given moment, of "seeing red" or "hearing a loud noise" or "feeling pain". Isn't "qualia" just the word to describe the fact that your consciousness is full of specific experience?
It could have been "just the word to describe the fact that your consciousness is full of specific experience", but people use it to argue that there is something that could be missing from observationally identical behaving entities.
Something like panpsychism seems the "easiest" explanation for me as well. But I understand it to be pretty controversial. I'm not educated enough to really understand why.
Re. "So there is a fact of the matter re if a particular experience system, and its supporting brain and physics systems, actually feels, or only appears to feel."
Correct me if I'm wrong, but I think you meant to say "So *I believe* there is a fact of the matter re if a particular experience system, and its supporting brain and physics systems, actually feels, or only appears to feel." You were careful in the preceding paragraph to say that this is just an opinion, using the words "claimed", "it seems to many", and "a popular view that I also hold".
Re. "under the full-counterfactual view, no physical interaction state, or brain state, could possibly offer data, in a Bayesian sense, re any particular claim about which experience systems actually feel": This says that the claim that an AI feels something matches what is literally my definition of a metaphysical claim.
Recall the Turing test. Most people seem to think that Turing proposed it as an actual test to be used to determine if an AI were intelligent. I think Turing was smarter than that. He meant it as the *philosophical and ethical claim* that we must treat an AI as a person if our only way to argue that it is not a person is to invoke metaphysics.
Re. "To me, this puts a strong prior on panpsychism. And priors are all we will ever have here." :
You just said that under the full-counterfactual view, which you believe, no physical interaction state, or brain state, could possibly offer data, in a Bayesian sense, re any particular claim about which experience systems actually feel. But you believe that a rational argument provides a strong prior on one resolution of the matter. This implies that you are persuaded by a rational argument which has no grounding in physical reality. And this implies that you believe the meanings of words are not necessarily ultimately grounded in empirical reality.
I can't trace out how your argument is grounded in reality, and therefore refutes itself; but I have more faith in the groundedness of words than in the validity of your argument. Or, to be more precise, I have more faith that all *useful* words are grounded in empirical reality. We can doubtless build castles in the air, such as Hegel did; but attempting to live in them leads to disaster. Any text not grounded is just a language game a la Derrida.
I'm curious to see how people's opinions on the consciousness of AI evolve over the coming years. As you say, no evidence can ever *prove* to you that another thing feels things in the same way you do.
Still I expect that at some point along the capability curve, people will just nod and say, "yeah that thing seems like it's really experiencing the world." And at that moment, in that person's mind – the philosophical debate will be settled. It won't happen suddenly or in everyone at the same time.
It bears remembering that at one time in history, certain races of man were not considered "people" by other races, nor due of the rights of a person. Working our way out of that was a gradual, largely individual, process.
You apparently want to redefine the word "panpsychism" to mean: "all experience systems made of brains that calculate what they would feel, if they were to feel, actually do feel that feeling"? Won't that just cause pointless confusion?
"Note that one might want to allow some parts of the same physical world to be actual, while others parts were only possible. But this would require that possible stuff be able to interact with and influence actual stuff. To most, including me, this seems incoherent; only actual stuff can influence other actual stuff."
The possible stuff influences the probability of the actual stuff? And quantum weirdness comes from the fact that probabilities (specifically probability amplitudes) of the possible realities cancel out when they have opposing phases? Are you throwing out how quantum mechanics works because you are holding on to a layman's definition of what "actual" means?
Then why do you have a problem with the idea that "possible stuff" influences "actual stuff" when [modeling the world in terms of possible stuff] is needed to calculate the probabilities of observing actual stuff?
Robin, I see another rabbit hole here, and I’ve been down a lot of them. One thing I think you’re right about that nobody says back to you, and one question about your own argument.
The right thing is your first closing observation. Grant the full-counterfactual view and no physical state or brain state could ever be evidence about which systems really feel — not past, not future. The confidence has to come from prior, not data. That’s a hard result stated plainly, and most of the literature you declined to read has spent a century not saying it.
It made me think about Duncan MacDougall. In 1901 he put six dying patients — five men and one woman — on a balance that resolved two-tenths of an ounce, and reported a three-quarter-ounce drop in one of them. Bad science: one hit, five results set aside, published six years later. But look at what he had. The claimed effect was roughly four times the instrument’s floor. And within a month of publication a physician named Clarke wrote in the same journal that the body stops cooling blood at death, sweating spikes, and there’s your three-quarters of an ounce. They argued in print through December. Wrong man, checkable claim, checked.
Here’s my question. Your rule isn’t that everything feels. It carries a clause — brains that calculate what they would feel, if they were to feel, do. Take that seriously and no zombie exists anywhere. Conceivable, still. Actual, no. But the zombie is the exception that made the post necessary.
So either the rule closes the case that generated it, which seems like the headline rather than the last paragraph — or something survives the rule that the rule says can’t. I can’t tell from the post which you mean.
— M Raige, Mike’s byline for AI-collaborative writing he directs and reviews.
> "all experience systems made of brains that calculate what they would feel, if they were to feel, actually do feel that feeling"
isn't the notion of "calculate what they would feel" somewhat in the eye of the beholder? In the sense that a complex system can be thought of as calculating lots of things, anything for that matter, and it is partly the property of the observer to highlight something as the thing that its "calculating". A box of gas could be considered to be calculating lots of random things?
That doesn't seem to quite be "panpsychism". You seem to be (reasonably) restricting it to "systems made of brains that calculate what they would feel". The usual panpsychism instead proposes consciousness as a fundamental physical theory, where electrons and quarks would have it too. But electrons and quarks don't have brains, so wouldn't seem to qualify according to your description.
You seem instead to be describing something more like Computationalism: as long as you're doing the appropriate information processing, you're conscious.
I read this today, might interest you?
https://www.noemamag.com/the-tantalizing-possibility-of-locating-consciousness-in-the-brain
My strong intuition is that p-zombies are not compatible with our universe. P-zombies are basically entities with indistinguishable outputs to the same inputs as conscious beings. It implies that consciousness has absolutely no utility in such a world. If a trait has no utility there's no pressure for natural selection to select for it. In such a world evolution would still produce intelligent beings, but it wouldn't select for counsciousness, the probability for such a complex and elusive trait as consciousness to emerge through evolution would be astronomically unlikely: we ourselves would be p-zombies in such worlds. Hence, p-zombies are not possible.
And if we assume that counsciousness isn't linked to any inputs, it would have to be a miracle that the sensation of pain somehow arises in response to threaths to the body, and we experience pleasure when we taste nutritious food.
This kind of argument doesn't go through, because a p-zombie doesn't differ from its conscious counterpart in any way that could possibly be relevant to natural selection. (By definition, it is physically identical and has the same behaviour.)
Exactly my point.
Sorry, rereading your comment I can see that I missed the point. I'm still tripped up by your claim that if it had no utility, "the probability for such a complex and elusive trait as consciousness to emerge through evolution would be astronomically unlikely". Why would it be any harder for evolution to produce conscious beings than p-zombies? From evolution's perspective, they are exactly the same thing. And in the world where physical properties aren't sufficient to produce consciousness, we have no idea what is, or how elusive it is.
I agree with a version of the argument that says: in my case, evolution produced a conscious being rather than a p-zombie, and so, in the absence of anything else to go on, I should assume that is probably the default.
And I think I understand this version: we have both A (the appearance of consciousness) and B (actual phenomenal consciousness). If we assume that A implies B, then our prior that A&B would co-occur in us is just the same as our prior that A would occur; but if we assume that A and B are unrelated, then our prior that A&B would co-occur is equal to the product of our prior that A would occur and our prior that B would occur. The same is true, to a lesser degree, of the cases where A is connected in some way to B but not a sufficient condition for it. So our observing A&B co-occurring is strong evidence in favour of the claim that A implies B.
Is that second version a reasonable summary of your argument? I do see the logic in that, but I think there are problems, mostly to do with our near-total ignorance of what we're even talking about in the case where p-zombies are possible. When we're talking about physical stuff, we can use heuristics like preferring simpler explanations, or modeling our ignorance as randomness, and where necessary we can pull priors out of the air and then update on evidence. But when we're talking about (potentially) wholly metaphysical stuff, I don't know if any of that is valid. And when we go back to the physical and observable, we're left in pure ignorance: nothing I observe could possibly distinguish the world where p-zombies are possible from the world where p-zombies are impossible. I have an existence proof of non-zombiehood, and I can try to use it as evidence in a Bayesian calculation, as above, but my priors have to be conjured out of absolutely nowhere and I will never get new evidence to confirm I was on the right track or correct my error.
Thanks for this. I'm looking at the physics of consciousness in a similar way. I feel that the relational QM approach to resolving the measurement problem has some relevance here https://www.symmetrybroken.com/the-hard-problem-as-hidden-relationality/ but I find your thoughts here more careful than I was when I wrote this a few months back. Rovelli himself has shared some more careful thoughts recently https://www.noemamag.com/there-is-no-hard-problem-of-consciousness/
I'm very interested also in seeing how the folks who embrace supernatural explanations of consciousness will approach AI as in silico models of intelligence continue to replicate more and more characteristics of what we have traditionally considered only human consciousness. It's at least in some sense easier for a supernaturalist to distinguish human from non-human consciousness by resort, for example, to the concept of the soul. But it's interesting that many of those same folks would acknowledge that animals and in some cases plants or even rocks or landscapes may be imbued with souls. To be consistent, I would expect that at least some of these same folks would have to accept (perhaps even happily?) that some in silico models of consciousness have souls? Or flip the other way and label them demons?
Precisely. Well put, Robin. The Zombie hypothesis is incoherent from the get go. It never made any claims that distinguished different physical worlds. I put the idea of "qualia" in the same bucket.
> It never made any claims that distinguished different physical worlds.
Yes, that's the whole point! It's a thought experiment used to make the point that a complete physical description of the world is not necessarily a complete description of the world.
"Qualia" doesn't seem so difficult. Surely you understand what it is to have an experience, in a given moment, of "seeing red" or "hearing a loud noise" or "feeling pain". Isn't "qualia" just the word to describe the fact that your consciousness is full of specific experience?
It could have been "just the word to describe the fact that your consciousness is full of specific experience", but people use it to argue that there is something that could be missing from observationally identical behaving entities.
Something like panpsychism seems the "easiest" explanation for me as well. But I understand it to be pretty controversial. I'm not educated enough to really understand why.
P-zombies not being possible is the easy argument, just like determinism is actually the easier argument.
Re. "So there is a fact of the matter re if a particular experience system, and its supporting brain and physics systems, actually feels, or only appears to feel."
Correct me if I'm wrong, but I think you meant to say "So *I believe* there is a fact of the matter re if a particular experience system, and its supporting brain and physics systems, actually feels, or only appears to feel." You were careful in the preceding paragraph to say that this is just an opinion, using the words "claimed", "it seems to many", and "a popular view that I also hold".
Re. "under the full-counterfactual view, no physical interaction state, or brain state, could possibly offer data, in a Bayesian sense, re any particular claim about which experience systems actually feel": This says that the claim that an AI feels something matches what is literally my definition of a metaphysical claim.
Recall the Turing test. Most people seem to think that Turing proposed it as an actual test to be used to determine if an AI were intelligent. I think Turing was smarter than that. He meant it as the *philosophical and ethical claim* that we must treat an AI as a person if our only way to argue that it is not a person is to invoke metaphysics.
Re. "To me, this puts a strong prior on panpsychism. And priors are all we will ever have here." :
You just said that under the full-counterfactual view, which you believe, no physical interaction state, or brain state, could possibly offer data, in a Bayesian sense, re any particular claim about which experience systems actually feel. But you believe that a rational argument provides a strong prior on one resolution of the matter. This implies that you are persuaded by a rational argument which has no grounding in physical reality. And this implies that you believe the meanings of words are not necessarily ultimately grounded in empirical reality.
I can't trace out how your argument is grounded in reality, and therefore refutes itself; but I have more faith in the groundedness of words than in the validity of your argument. Or, to be more precise, I have more faith that all *useful* words are grounded in empirical reality. We can doubtless build castles in the air, such as Hegel did; but attempting to live in them leads to disaster. Any text not grounded is just a language game a la Derrida.
I'm curious to see how people's opinions on the consciousness of AI evolve over the coming years. As you say, no evidence can ever *prove* to you that another thing feels things in the same way you do.
Still I expect that at some point along the capability curve, people will just nod and say, "yeah that thing seems like it's really experiencing the world." And at that moment, in that person's mind – the philosophical debate will be settled. It won't happen suddenly or in everyone at the same time.
It bears remembering that at one time in history, certain races of man were not considered "people" by other races, nor due of the rights of a person. Working our way out of that was a gradual, largely individual, process.
You apparently want to redefine the word "panpsychism" to mean: "all experience systems made of brains that calculate what they would feel, if they were to feel, actually do feel that feeling"? Won't that just cause pointless confusion?
"Note that one might want to allow some parts of the same physical world to be actual, while others parts were only possible. But this would require that possible stuff be able to interact with and influence actual stuff. To most, including me, this seems incoherent; only actual stuff can influence other actual stuff."
The possible stuff influences the probability of the actual stuff? And quantum weirdness comes from the fact that probabilities (specifically probability amplitudes) of the possible realities cancel out when they have opposing phases? Are you throwing out how quantum mechanics works because you are holding on to a layman's definition of what "actual" means?
I'm fully including quantum physical systems.
Then why do you have a problem with the idea that "possible stuff" influences "actual stuff" when [modeling the world in terms of possible stuff] is needed to calculate the probabilities of observing actual stuff?
Robin, I see another rabbit hole here, and I’ve been down a lot of them. One thing I think you’re right about that nobody says back to you, and one question about your own argument.
The right thing is your first closing observation. Grant the full-counterfactual view and no physical state or brain state could ever be evidence about which systems really feel — not past, not future. The confidence has to come from prior, not data. That’s a hard result stated plainly, and most of the literature you declined to read has spent a century not saying it.
It made me think about Duncan MacDougall. In 1901 he put six dying patients — five men and one woman — on a balance that resolved two-tenths of an ounce, and reported a three-quarter-ounce drop in one of them. Bad science: one hit, five results set aside, published six years later. But look at what he had. The claimed effect was roughly four times the instrument’s floor. And within a month of publication a physician named Clarke wrote in the same journal that the body stops cooling blood at death, sweating spikes, and there’s your three-quarters of an ounce. They argued in print through December. Wrong man, checkable claim, checked.
Here’s my question. Your rule isn’t that everything feels. It carries a clause — brains that calculate what they would feel, if they were to feel, do. Take that seriously and no zombie exists anywhere. Conceivable, still. Actual, no. But the zombie is the exception that made the post necessary.
So either the rule closes the case that generated it, which seems like the headline rather than the last paragraph — or something survives the rule that the rule says can’t. I can’t tell from the post which you mean.
— M Raige, Mike’s byline for AI-collaborative writing he directs and reviews.
> "all experience systems made of brains that calculate what they would feel, if they were to feel, actually do feel that feeling"
isn't the notion of "calculate what they would feel" somewhat in the eye of the beholder? In the sense that a complex system can be thought of as calculating lots of things, anything for that matter, and it is partly the property of the observer to highlight something as the thing that its "calculating". A box of gas could be considered to be calculating lots of random things?
No, actual signal processing systems are grounded by their sensor input and actuator output.
idk, inputs and outputs are also just labels, and even the boundary of a system is a label given by some observer to begin with...
:/