This is not the first time it seems to me that Yudkowsky is hinting for us to understand this instead of writing directly. But on the other hand, he seems to write as if he doesn’t really know the answer. Specifically in this case he says about qualia, that we should ask the question “what makes us think we have qualia?” At first I got that wrong, in the sense that qualia is just an illusion. But then I did a mental experiment about a neural network. What if we created a neural network that could similarly answer questions about the picture it recognized from its cameras. We ask it what it sees in the right corner, it answers. And so on. And all the answers will be as if she really SEES the whole picture from qualia, just like people do. As if the retina or its counterpart is already enough to have qualia, the only difference is that not all retina wearers can answer us in detail about its contents. On this basis, it seems that qualia is not just some “consciousness field” type property of the universe, it is an inherent logical/mathematical phenomenon, its absence not possible in any alternate world, it is just confusion/nonsense. Another thought experiment also hints at this. Imagine a universe without qualia. Wait, but how? There is no one observer there, no one point of view. No one can ever see what’s going on in that universe. And it just seems… Wrong. It’s as if the existence of a universe in which nothing can be observed, even if there are philosophical zombies there, just doesn’t make sense. In other words, qualia are strictly retinal cells, animals have them as well, neural networks, and probably in a very simplified form even bacteria and sensors. The ability to speak only allows us to “cash in” on them, to inform people of their existence, though in fact they have always been there, everywhere.
Next comes the question of self-consciousness. It doesn’t seem to be some conceptual (epiphenomenal) thing that makes you valuable. It just seems to be something that is missing when we dream.
I can even imagine people with no self-awareness, their difference would be that they would be incapable of self-reflection. Because of this, they would even be more efficient workers, not distracted, not engaged in procrastination and useless hesitation and so on. They will be able to answer questions about their condition, they can be made to go one level deeper, but they will not do it on their own. Because of this, they will be sillier. New thoughts will visit them only on the basis of external stimuli. They will invent fewer new ideas in a lifetime and have worse experience unless someone outside forces them to think until they say it seems the task is over. It seems their difference will be that they don’t have wandering chaotic signals and closed loops of neurons in their brains to repeatedly pass backward error propagation. They might also not have a task stack, they might not be able to remember something at the right moment or just remember, they would necessarily need an alarm clock. Lack of ability to hold more than one object in short-term memory? And hence the only reason why their emotions are less valuable is because they are shorter at the same stimuli, no self-reflection loop, just a single run of the pulse of negative encouragement. There will be no shame for a week, just a single unpleasant sensation and a momentary change in behavior, probably to a lesser degree than in repeatedly reflective beings. The problem here, as with pain, is that we cannot stop our unpleasant emotions, even if we have already learned a lesson from them.
Perhaps I’m wrong. Or have found some other aspect and given it a description and it’s not self-consciousness. But what appeals to me about this explanation is that it seems to remove any mystery from the concept of consciousness and make it just another of the understandable properties of the human brain.
Another aspect of consciousness may be the ability to separate one’s self from the outside world, to distinguish one’s actions from external ones. And it seems again to be one of the functions that is disconnected in sleep, there you don’t distinguish between your body moving somewhere and the change in the external environment and the change in your thoughts. Then it also seems to hurt your learning and be like that robot that just paints over the blue pixels and doesn’t try to take the blue glass off your visor.
The point is that this again deprives self-consciousness of the properties of an icon. It does not magically draw the line between valued minds and unvalued minds. The difference between a conscious being and an unconscious one would not be like the difference between a person and an ironman, but like the difference between a sober person and a drunk. We value them less because they are less controllable and more dangerous, less intelligent and therefore worse at fulfilling deals. It’s not a question of sanity and not sanity, of value and not, but of Lawful and Chaotic Worldview (hmm, does Yudkowsky write about this somewhere else in The Mad Investor of Chaos? Because I didn’t finish reading it). It’s not that we value them less because of mere consciousness, it’s that we can rely on them less and prioritize actors we can rely on.
At the same time, judging from the human experience, it seems that while we cannot build anything without qualia, we can build something without happiness and unhappiness, without boredom and pain. It seems that people sometimes just won’t experience emotion, won’t agonize, but can still learn from their mistakes. This is what we might call conscious thinking, plan-building, just remembering, not reinforcement mechanisms. When we can change our actions simply because we made a decision. But so far I don’t have any ideas about that. Except that emotion is basically just a property of neural signals running in a loop.
Also, I don’t think there’s anything mysterious about the redness of red, it seems to be just a result of you starting to reason about the intermediate layers of neurons as well as the input layers, kind of like the unsolvable question of “but is Pluto a planet after all?” And so I am a materialist, but accordingly I expect that qualia are not some mental substance, but fully physical chains of neurons. And I also predict that with enough technology it will be possible to connect one person’s neurons to another person’s neurons in the right way, and then that person will see someone else’s redness and be able to see if it is the same or different. This is a falsifiable scientific hypothesis. We can test whether they are the same. We can test the hypothesis that the same chains of neurons, when connected to different brains, will give the same sensations. This is not just a belief in materialism and the use of Occam’s razor, it is a testable question. Here’s if, after trying different neural circuits, people say the redness is the same, and philosophers talk about how suddenly they see different redness, even though they say it’s the same, then that would be a multiplication of entities.
Finally, the sense of Pluto as a planet or redness of redness is ineffable because these are internal layers of neurons, not inputs. In the case of the latter, we can point to an external object to change their state, compare and synchronize. In the case of the former we cannot. Just as you cannot explain to a colorblind person the inexpressibility of color and to a blind person the inexpressibility of vision. But these are not some fundamental obstacles, but, as I have written before, only a consequence of the fact that people do not have mechanisms for object serialization and conscious reflexive reprogramming of the source code/neuronal circuits. If there were, you could simply look into your brain, find out exactly how your neurons are connected, transform your neurons’ territory into their map, describe that map with language, have another person transform them back into territory, forcing your brain to form new neurons according to their description. And after that, you could, without “non-defeatability,” let a color-blind person see color whenever they want, even if their eyes have no cones, and they are unable to detect color by sight. You could also describe to a blind person what neurons you feel your visual cortex has, so that he would start the process of forming the same combination of them in his brain, after which you could tell the blind person which neurons need to be activated for him to feel what you feel, even if he is unable to determine for himself what the picture really is.
(I have a concern that this might be a dangerous topic to discuss, but it seems to be discussed elsewhere on lessvrong with no problem, including Yudkowsky himself, when he dispels the mystery/black box around different brain abilities. So I’m just not going to write specifically about the ways of abuse)
(This turned out to be longer than I expected, I don’t know if I should create a full top-level post)
This is not the first time it seems to me that Yudkowsky is hinting for us to understand this instead of writing directly. But on the other hand, he seems to write as if he doesn’t really know the answer. Specifically in this case he says about qualia, that we should ask the question “what makes us think we have qualia?” At first I got that wrong, in the sense that qualia is just an illusion. But then I did a mental experiment about a neural network. What if we created a neural network that could similarly answer questions about the picture it recognized from its cameras. We ask it what it sees in the right corner, it answers. And so on. And all the answers will be as if she really SEES the whole picture from qualia, just like people do. As if the retina or its counterpart is already enough to have qualia, the only difference is that not all retina wearers can answer us in detail about its contents. On this basis, it seems that qualia is not just some “consciousness field” type property of the universe, it is an inherent logical/mathematical phenomenon, its absence not possible in any alternate world, it is just confusion/nonsense. Another thought experiment also hints at this. Imagine a universe without qualia. Wait, but how? There is no one observer there, no one point of view. No one can ever see what’s going on in that universe. And it just seems… Wrong. It’s as if the existence of a universe in which nothing can be observed, even if there are philosophical zombies there, just doesn’t make sense. In other words, qualia are strictly retinal cells, animals have them as well, neural networks, and probably in a very simplified form even bacteria and sensors. The ability to speak only allows us to “cash in” on them, to inform people of their existence, though in fact they have always been there, everywhere. Next comes the question of self-consciousness. It doesn’t seem to be some conceptual (epiphenomenal) thing that makes you valuable. It just seems to be something that is missing when we dream. I can even imagine people with no self-awareness, their difference would be that they would be incapable of self-reflection. Because of this, they would even be more efficient workers, not distracted, not engaged in procrastination and useless hesitation and so on. They will be able to answer questions about their condition, they can be made to go one level deeper, but they will not do it on their own. Because of this, they will be sillier. New thoughts will visit them only on the basis of external stimuli. They will invent fewer new ideas in a lifetime and have worse experience unless someone outside forces them to think until they say it seems the task is over. It seems their difference will be that they don’t have wandering chaotic signals and closed loops of neurons in their brains to repeatedly pass backward error propagation. They might also not have a task stack, they might not be able to remember something at the right moment or just remember, they would necessarily need an alarm clock. Lack of ability to hold more than one object in short-term memory? And hence the only reason why their emotions are less valuable is because they are shorter at the same stimuli, no self-reflection loop, just a single run of the pulse of negative encouragement. There will be no shame for a week, just a single unpleasant sensation and a momentary change in behavior, probably to a lesser degree than in repeatedly reflective beings. The problem here, as with pain, is that we cannot stop our unpleasant emotions, even if we have already learned a lesson from them. Perhaps I’m wrong. Or have found some other aspect and given it a description and it’s not self-consciousness. But what appeals to me about this explanation is that it seems to remove any mystery from the concept of consciousness and make it just another of the understandable properties of the human brain. Another aspect of consciousness may be the ability to separate one’s self from the outside world, to distinguish one’s actions from external ones. And it seems again to be one of the functions that is disconnected in sleep, there you don’t distinguish between your body moving somewhere and the change in the external environment and the change in your thoughts. Then it also seems to hurt your learning and be like that robot that just paints over the blue pixels and doesn’t try to take the blue glass off your visor. The point is that this again deprives self-consciousness of the properties of an icon. It does not magically draw the line between valued minds and unvalued minds. The difference between a conscious being and an unconscious one would not be like the difference between a person and an ironman, but like the difference between a sober person and a drunk. We value them less because they are less controllable and more dangerous, less intelligent and therefore worse at fulfilling deals. It’s not a question of sanity and not sanity, of value and not, but of Lawful and Chaotic Worldview (hmm, does Yudkowsky write about this somewhere else in The Mad Investor of Chaos? Because I didn’t finish reading it). It’s not that we value them less because of mere consciousness, it’s that we can rely on them less and prioritize actors we can rely on. At the same time, judging from the human experience, it seems that while we cannot build anything without qualia, we can build something without happiness and unhappiness, without boredom and pain. It seems that people sometimes just won’t experience emotion, won’t agonize, but can still learn from their mistakes. This is what we might call conscious thinking, plan-building, just remembering, not reinforcement mechanisms. When we can change our actions simply because we made a decision. But so far I don’t have any ideas about that. Except that emotion is basically just a property of neural signals running in a loop. Also, I don’t think there’s anything mysterious about the redness of red, it seems to be just a result of you starting to reason about the intermediate layers of neurons as well as the input layers, kind of like the unsolvable question of “but is Pluto a planet after all?” And so I am a materialist, but accordingly I expect that qualia are not some mental substance, but fully physical chains of neurons. And I also predict that with enough technology it will be possible to connect one person’s neurons to another person’s neurons in the right way, and then that person will see someone else’s redness and be able to see if it is the same or different. This is a falsifiable scientific hypothesis. We can test whether they are the same. We can test the hypothesis that the same chains of neurons, when connected to different brains, will give the same sensations. This is not just a belief in materialism and the use of Occam’s razor, it is a testable question. Here’s if, after trying different neural circuits, people say the redness is the same, and philosophers talk about how suddenly they see different redness, even though they say it’s the same, then that would be a multiplication of entities. Finally, the sense of Pluto as a planet or redness of redness is ineffable because these are internal layers of neurons, not inputs. In the case of the latter, we can point to an external object to change their state, compare and synchronize. In the case of the former we cannot. Just as you cannot explain to a colorblind person the inexpressibility of color and to a blind person the inexpressibility of vision. But these are not some fundamental obstacles, but, as I have written before, only a consequence of the fact that people do not have mechanisms for object serialization and conscious reflexive reprogramming of the source code/neuronal circuits. If there were, you could simply look into your brain, find out exactly how your neurons are connected, transform your neurons’ territory into their map, describe that map with language, have another person transform them back into territory, forcing your brain to form new neurons according to their description. And after that, you could, without “non-defeatability,” let a color-blind person see color whenever they want, even if their eyes have no cones, and they are unable to detect color by sight. You could also describe to a blind person what neurons you feel your visual cortex has, so that he would start the process of forming the same combination of them in his brain, after which you could tell the blind person which neurons need to be activated for him to feel what you feel, even if he is unable to determine for himself what the picture really is. (I have a concern that this might be a dangerous topic to discuss, but it seems to be discussed elsewhere on lessvrong with no problem, including Yudkowsky himself, when he dispels the mystery/black box around different brain abilities. So I’m just not going to write specifically about the ways of abuse) (This turned out to be longer than I expected, I don’t know if I should create a full top-level post)