Roger Penrose Discusses Consciousness (2017)
nautil.us
nautil.us
Personally I changed my mind quite recently and the following quote from the article now resonates very much.
Minsky called consciousness a “suitcase word” that lacks the rigor of a scientific concept. “We have to replace it by ‘reflection’ and ‘decisions’ and about a dozen other things,” he said. “So instead of talking about the mystery of consciousness, let’s talk about the 20 or 30 really important mental processes that are involved.
Just sit down and try to nail down what you mean when you say, you are conscious or you feel pain. Try to make precise what you think distinguishes you from a philosophical zombie. When the voice in your head says that this was not the best decision and next time you should do something else, is that really any different from one system monitoring and controlling another system? When you are experiencing pain, when you try to get away from its source, when you maybe scream, is that really any different from you car turning on the low oil pressure warning light?
Of course it is! Or is it? Maybe your brain just turned the I-am-special indicator light on...
This is actually quite subtle. For example, I don't see how you can "simulate" bell non-locality. Is it even possible to define the word "simulate" here? The whole status of quantum randomness is unsettled, I don't think it can be settled, but anyway it is not as straightforward as you suggest.
Scott Aaronson has a nice paper where he discusses related ideas: https://arxiv.org/abs/1306.0159
Quantum non-locality can not be simulated classically if you distribute your computations just as the simulated system is distributed and if you do not allow for communication between the computers. This is however not an issue if you perform the simulation in one place or allow for communication between the computers. An example may help. Take the classical experiment of producing two spins with opposite orientations in one place, then separating them by a large distance, and finally measuring them at those distant locations. You can easily simulate that with one classical computer, but you can not use two classical computers - one to simulate each spin and the measurement of that spin at the distant location - without allowing the two computers to communicate even after the time the spins have been separated [in the simulation].
This especially highlights that quantum mechanics can not be described by a local hidden variable theory, or maybe it is better to think about it the other way round, such an simulation is not possible because quantum mechanics can not be explained by a local hidden variable theory as predicted by the Bell inequality and demonstrated by Bell test experiments. You can not write an algorithm that generates some hidden state while the two spins respectively the two computers are not yet separated and can therefore still communicate with each other and that then reproduces the results of all possible measurement experiments after the separation and without further communication.
This tends to get cleared up when a keyword like "qualia" gets mentioned, but it's so common to see lots of words wasted before that happens, if it even does. It's interesting how impactful the breadth of a shared vocabulary can be.
Well, qualia is exactly one of the things I would try to avoid. What do we really mean with that word, what are objective properties of qualia? As I said, I am increasingly entertaining the idea that we are just getting fooled by ourselves when our brains tell us that we obviously and undeniably have consciousness and experience qualia. What if this is just a hardcoded, meaningless phrase our brain spits out every time we or somebody else asks us whether we are conscious or experience qualia? Brain: Am I conscious? Brain: Yes. No more than five lines of code.
That can be useful depending on what you want to achieve in the discussion, but ignoring the qualia question altogether won't make it go away :)
what are objective properties of qualia?
If we could answer that, we would have solved most of the problem. The closest thing to an objective property is that we indeed talk about them i.e. they do objectively affect the external world. This is true whether they're illusory by nature or not.
I've tried to understand the illusory explanation, but I really can't get over the sheer undeniability of actually having a subjective experience, that being (as Descartes put it) the only thing whose existence I can be really truly certain of. The illusory explanation feels too much like saying "if it doesn't fit our current scientific worldview, it's not real". How did you get over this hurdle?
I didn't. Not too long ago I was just thinking about what a sophisticated artificial intelligence based on essentially run of the mill algorithms would be missing compared to a conscious human and I tried to formulate it without using any words with fuzzy meaning. And then I just failed. Every aspect I considered could be more or less easily replicated with normal algorithms. It seems that only my brain is telling me that there is something more going on in humans but there is no evidence for that, at least none that I am aware of. And then I just tried to do the rational thing - if you are convinced of something but without having any evidence for it, then at least consider the possibility, that this conviction might be wrong. And last but not least I realized how incredible easy it would be to add just a few lines of code to an artificial intelligence so that it always argues with great conviction that it is of course self-evident, obvious, and undeniable that there is more going on inside of an artificial intelligence than inside a human. In some sense this would be a very simple and elegant solution for a very hard problem, wouldn't it?
In other words, knowing how to simulate a phenomenon is not for me a sufficient explanation of the phenomenon, nor even necessarily a very strong indication of what the phenomenon's actual implementation is like.
It seems that only my brain is telling me that there is something more going on in humans but there is no evidence for that, at least none that I am aware of.
Why such distrust in your own brain? :) I think (the existence of) your own immediate subjective experience is evidence, and compelling evidence at that. It's just not the kind of evidence that's currently easily accessible by science. Over the course of history, many things have been inaccessible to science, and then science has progressed.
Science has at least one plausible angle of attack too. Qualia seem to be associated with only some processes in the brain (others are entirely subconscious), some of the time (attention can be redirected). This gives some hope of finding correlations between qualia and brain activity, which AIUI is a major topic in neuroscience currently.
I totally agree, that would be an invalid conclusion. But that is not my point, my point is that we might not actually experience qualia, that our believe in us experiencing qualia is a confusion caused by our brain. If that were the case, there would be nothing an artificial would or even could be lacking, its experience of qualia - or rather the lack of it - would be identical to ours. The only difference would be that humans would claim to experience qualia - while they actually don't - while an artificial intelligence would not experience qualia either but also not claim to do so, at least unless the artificial intelligence also replicates the false human claim and claimed to experience qualia, which would be trivial to do.
If we only care about externally observable behaviour, then yes, they can be equivalent, but I don't think that's all we should care about.
I can only repeat my challenge, not sure if I wrote it to you before or whether it was in another thread. Try to make precise what this non-external thing is or provide evidence for its existence. Just come up with one sentence that says something about it that is neither trivial nor so fuzzy that it could mean anything and nothing. If you can do that, I will immediately switch to your side of the argument. If you do not want to reply to all my points, reply to this one, it's the central one. But be warned, you will have to pass quite a high bar with respect to non-triviality and non-fuzziness to convince me.
Why such distrust in your own brain? :)
Because my brain is telling itself and every other brain asking about it, that it is conscious and experiences qualia while the very same brain is unable to explain what those words even mean.
I think (the existence of) your own immediate subjective experience is evidence, and compelling evidence at that.
I totally disagree here, my brain telling my brain that my brain has some special thing going on is kind of the worst evidence possible. On an intuitive level I am as convinced as you are that I am conscious and experience qualia but once I start looking at it rationally this conviction starts fading.
Over the course of history, many things have been inaccessible to science, and then science has progressed.
Can you think of an example where we were not only unable to understand something but were even unable to describe the thing we could not understand? I have the feeling that there might have been a few such things but ad hoc I can not name any.
Qualia seem to be associated with only some processes in the brain (others are entirely subconscious), some of the time (attention can be redirected). This gives some hope of finding correlations between qualia and brain activity, which AIUI is a major topic in neuroscience currently.
Does that mean that you would be willing to attribute full consciousness to a CPU with some RAM if it ran some software creating the same patterns in memory as they occur in our brains?
If there were a nice answer to this, the "hard problem" would be a lot less hard. But I'll give it a shot.
Along the lines of Descartes, assuming maximal paranoia about everything, it's the thing whose existence you can doubt the least.
That is, let's follow Descartes's line of thinking about how the external world and even all the ideas and concepts in your mind could, in principle, all be an elaborate lie. Then the only thing you absolutely cannot doubt is the existence of the experience of your own thinking. That experience is qualia.
Checking this definition against your criteria:
- Could it mean anything or nothing? It may be ambiguous in that there might be multiple things whose existence you can doubt equally little, but I'd say they're all probably in the ballpark.
- Trivial or fuzzy? Maybe? Admittedly I invoke the somewhat fuzzy concept of "doubt", which I haven't defined properly, and probably can't. But hey, all reasoning builds on some axioms :)
By the way, are you sure your challenge places the burden of proof in the right place? I feel like your position of "our brains are fooling us" is quite a bit stronger and "less default" than my fairly agnostic position. I feel your position at least owes an explanation of why our brains might be fooling us in such and oddly specific and seemingly unnecessary way.
Can you think of an example where we were not only unable to understand something but were even unable to describe the thing we could not understand?
Vitalism (https://en.wikipedia.org/wiki/Vitalism), and the progression from Newtonian physics, are common examples given for this.
Whether we have been unable to describe the things we could not yet understand, I suppose it depends on what counts as a description. We have always been able to point at the objects and phenomena that didn't fit our theories, sure. Qualia on a subjective level might be fundamentally more ineffable for now, but if we insist on immediate objectivity (and I don't think we necessarily should), we can point at qualia too at least on the behavioural level of "we don't know how or why human brains produce this peculiar behaviour of contemplating qualia".
Just like physicists needed to ask the right questions to improve upon Newtonian physics, I think finding correlations between brain activity and (reported) qualia will help us find the right questions, probably at first along the lines of "what's so special about this specific brain region / process". I doubt we'll manage to answer the behavioural question in satisfactory detail without getting at least some new ideas about how to approach the subjective question.
Does that mean that you would be willing to attribute full consciousness to a CPU with some RAM if it ran some software creating the same patterns in memory as they occur in our brains?
The closer the simulation's implementation is to a one-to-one to physical brain processes, the more likely I would think it is to have qualia. If a sufficiently one-to-one implementation of whatever mechanism turns out to be responsible for qualia is possible with CPU and RAM as they exist today, then it'd follow that qualia (and the behaviour of talking about them) are a "by-product" of certain kinds of information processing, rather than some funky physics shenanigans. I'm entirely open to that possibility.
And yes, my sciencey brain also pattern-matches what I just said with "actually having qualia is unnecessary in that statement!", and I'd be inclined to agree with that if qualia weren't so seemingly unnecessary and elaborate for a mere illusion, and if I didn't actually have qualia :)
The same question can be asked. What are objective properties of "being fooled"?
Can you see something? Yes, you can. Can you reflect on your ability to see? Yes. Can you think? Can you reflect on what you are thinking now? Can you reflect on your ability to think? Yes. Yes. Yes. What you are being fooled about?
Getting convinced [by your brain] that an objectively wrong fact is true. For a example an optical illusion where your brain decides that two objectively identical colors are different colors. And we can find out about this by just asking you about your perception. And you can stop getting fooled by learning about that optical illusion so that in the future some part of your brain might raise doubt about the difference the other part of your brain detects which will in turn change your answer when asked to something expressing less confidence about the two colors.
Can you see something? Yes, you can. Can you reflect on your ability to see? Yes. Can you think? Can you reflect on what you are thinking now? Can you reflect on your ability to think? Yes. Yes. Yes. What you are being fooled about?
I am too lazy to work through all of them, but I would tend to argue along the following line. The fact that I can see can be verified from the outside - you show me different things and then ask me about them. We do not even need something as complex as you, maybe just an LED changing brightness or color from time to time and a button I am supposed to press every time the visible state of the LED changes. After controlling for things like me sensing the current through the LED, the amount of correlation between the signals to the LED and from the button will be a pretty good indicator of my ability for some basic vision capability.
But what when you claim to feel pain? How would you try to convince me that there is more to this than the pain bit getting set and in turn causing and affecting your actions? And you can of course not just appeal to me also being a human and therefore just knowing that there is more to the experience of pain.
Decrease of performance on tasks unrelated to damaged area shows that something is going on beyond physical damage and usual stimulus processing.
Worker ants, for example, continue to carry on their work even after fatal damage to their gaster.
You will not continue to perform mental arithmetic with the same efficiency after your little toe has been struck by a hammer, despite it serves no purpose in doing mental arithmetic (or any purpose at all, for that matter). And if you will do, then you don't feel pain.
But more relevant to the topic, while generally reduced performance is strongly related to pain, I have no problem imaging a robot exhibiting degraded performance due to damage in somewhat unrelated functions for the purpose of resource conservation or something along that line. It could even be a kind of overreaction - if the damage and its consequences are not well known, then be conservative and reduce performance in general. But while building something like that into a robot makes it behave more like a human, I am pretty sure that this is not what people usually think of when they say that they feel pain, i.e. most people would still say that the robot still does not really feel pain.
"Being fooled" would arise from the model being incorrect of itself, the "conscious" self influencing the model to not faithfully represent reality, and so on.
I'd make an analogy to "Trusted Execution Technology"[1] on computers these days. It's easy to overlook except when it goes wrong.
[1]https://en.wikipedia.org/wiki/Trusted_Execution_Technology
Isn't that what a object recognition algorithm would do? Identify regions of an image and recognize them as instances of some class of things, maybe also build up relations between them - a person walking on the side walk vs another person riding a bike that has been recognized.
[...] which is not subject to free will [...]
I am not sure what you mean with that.
"Being fooled" would arise from the model being incorrect of itself [...]
Exactly, that's my point. We are so sure that our experience of »seeing« a tree is fundamentally different from a camera »seeing« a tree and a object recognition algorithm labeling a region of pixels as a tree...but where is the evidence for that? No human has ever been a camera attached to a computer and could compare the two »experiences«. It is just our brain believing that a human is different from a camera plus computer and algorithms, so maybe our brain is just wrong in this regard, fooling us into believing that there is something more going on inside of us.
I'd make an analogy to "Trusted Execution Technology"[1] on computers these days. It's easy to overlook except when it goes wrong.
I was really convinced for my entire life that me »seeing« a tree involves something that a camera does not have, it always just seemed so obvious to me. But once I started really thinking about it, I noticed that I neither have evidence for that, nor that I am even able to make precise what this missing thing is supposed to be. So it is not only that I may have missed the TPM module, but I am even unable to find it when I am actively looking for it. Maybe it is just hidden very well, but my natural confidence in its existence paired with zero evidence for that, makes me at least somewhat suspicious nowadays.
If you want evidence, read Kafka, read Incompleteness (about Kurt Godel), read Schreber's "Memoirs of my Nervous Illness".
No, ad hoc and without having thought too much about it, I would say subjectivity is a form of relativity. The temperature on a sunny day being 25 °C is an objective fact, depending on where you usually live that might be a subjectively cold or subjectively hot day to you.
Maybe the people who talks about qualia means that the subject has some characteristic which is not objective or measurable.
If you cut your finger, you will detect that, you will see the wound and nerves in your skin will sense the wound in it. That are objective things. There will also be objective reactions, blood clotting and other low-level biological reactions starting, you might take actions like looking for a plaster or saying that this hurts. Up to this point, I would argue, most people would agree that a sophisticated robot could behave identical to you.
But then you might claim that there is a qualia to this event of cutting your finger, it is painful for you. You feel pain and this is more than you simply detecting the wound and saying it hurts. And we all know what this is supposed to mean, we all know what pain feels like. But now try to nail this down, what is that you claim to have or experience? What is it, that a robot is missing when its cameras see a cut in its skin and some kind of sensors in the skin signal damage, flipping some damage bit in turn starting repair processes and maybe even saying »I cut my finger, it hurts!« What do we have to add to make the robot feel pain?
For a long time I was convinced that the robot is missing something, I just could not say what exactly it is. Now I am pretty seriously considering that nothing is missing, that me feeling pain is not much more than flipping the pain and the set of reaction triggered or influenced by that.
When we see, the eye picks signals, when we touch something some cells pick the signals that inform about pressure or temperature. These signals are sent to the brain. They may be coded in binary form. The brain gets the signals and processes them and produces some reaction.
Up to this point we are dealing with signals, and the signals apparently would be enough to transmit the information which would be processed, and the reaction would also consist of signals, which may be binary-like. But why, if signals are enough, we "see" an image, we feel the temperature? I mean, there's something else apart from the binary signals. How do the signals become the experienced perceptions?
That is the million dollar question. I have never been a camera attached to a computer running some object recognition software, but I am willing to at least consider the possibility that having an array of pixels and a list of regions with labels for the objects in memory is the same thing as me »seeing« a scene.
I mean, what does it really mean that I »see« a tree? I am aware that light with some specific frequencies - mostly shades of green and maybe brownish bark - is coming from some direction in some specific pattern of a trunk and branches and leaves and I associate labels with the whole thing and parts of it. Maybe let's add that I reconstruct a kind of 3D model of the tree. What more does it mean to »see« a tree?
Had we discussed this just four weeks ago, I would have totally agreed with you. And to some extend I still find it hard to believe myself.
I'm more likely to believe that there's an unexplained gap between the signals and the subjective experience.
Now try to tell me more about this gap. What is part of this gap? And what is the evidence for that or the existence of this gap in general? And try to avoid using any word with fuzzy or ambiguous meaning like experience or qualia and don't appeal to the fact that I am also a human and therefore know what you mean.
My own failure to come up with anything I can say about this gap and my failure to find any evidence for the existence of this gap is what made me consider the possibility that the gap does not exist, that it is an illusion my brain tricks me into believing in.
Had we discussed this just four weeks ago, I would have totally agreed with you. And to some extend I still find it hard to believe myself.
I'm more likely to believe that there's an unexplained gap between the signals and the subjective experience.
Now try to tell me more about this gap. What is part of this gap? And what is the evidence for that or the existence of this gap in general? And try hard to avoid using any word with fuzzy or ambiguous meaning like experience or
There is just one you that does things by whimsy all the time, and at the same time tries to make a coherent story about your immediate behavior in a way that is consistent with your past. A naive stoic might say that one should not become too upset about things outside your control. But in truth the best you can do is take the emotions and thoughts you feel in-stride because even those are outside your control—the stoic’s decision to follow stoicism in the first place is after all ineffable.
I'm sorry, but because you're using the terms, I have to interject and inform the wider audience that this claim is not at all an accurate reading of Freud. Freud makes it extremely clear in his text The Ego and The Id that Ego/Id is one thing. The perceptible boundary is whether a mental process is available to being represented in language.
While I would agree that in the end the brain is just one thing doing its thing, I would still say that it is also composed out of several interacting components with different functions and responsibilities. For example one part of your brain might have an impulsive response in a certain situation and another part might try to overrule that with a more rational thought. Some third component will of course just pick a winner of the struggle without anything you can do about it. Well, you are of course doing all those thing, but you know what I mean.