 Hello to everybody. I'm Jeff Sachs. Welcome to Book Club with Jeff Sachs, and I'm absolutely thrilled to be speaking today with Professor Anil Seth, who is a Professor of Cognitive and Computational Neuroscience at the University of Sussex. And he's also Co-Director of the Sackler Center for Consciousness Science. And he's the author of an absolutely scintillating, wonderful, exciting book to read, which I could not recommend more strongly, Being You, A New Science of Consciousness. And Anil, thank you so much for joining today. And I can't wait to speak to you about this book because when I was reading it, I was so excited on every page. And it's great just to have a chance to chat with you about it. Jeff, that's really nice to hear. Thank you very much for the invitation. I'm delighted to be here too. Why A New Science? Let's start with that. It is a new science. The book describes that. But for people listening in consciousness in some way or another, we must have been aware of it for a long time. But what's new about our understanding of consciousness? The boring answer to that question is it's always hard to find a subtitle for a book that hasn't been already used. There's a lot of new science, obviously, and that's what I want you to tell us about. Yeah, there is truth to it. There's a story behind the subtitle, Consciousness, this central mystery of life. Why do we have conscious experiences? How does this massive electrochemical machinery that we call a brain, how does this physical stuff give rise to any kind of subjective experience, the sensation of seeing red or the feeling of pain? This has been, of course, it's not new, this question. This has been a question that's challenged philosophers for thousands of years and scientists from the beginning of science. But it's often been slightly, at least in most of the 20th century, it was quite a disreputable thing to study. And this is partly because it doesn't fit neatly into the normal scientific method. For most other areas of science, we can be objective about the data, you know, we can make measurements, we can make measurements about how hot something is, we can collect data about the climate, we can look at cells under a microscope. But the target of a science of consciousness is, by definition, subjective. You know, I have my conscious experience, you have yours, we can't put a conscious experience on the table and look at it. In other words, even though we're talking together, I can't be absolutely sure that you're conscious what's going on in your mind or you of me, because I can't directly observe your consciousness. And in fact, you could be a very sophisticated artificial intelligence bot that I'm speaking to. Is that the point? That's kind of the point. I mean, maybe a slightly unsophisticated AI would do the job for me. But so there's this point of you, of course, solipsism, which would say that I only know that I can be conscious. And that's the only thing I can be really sure of. I don't even know that anybody else is you, anybody in my family, anybody. But really, I think most people don't take that as a serious possibility. And it doesn't really affect our ability to do science. The problem that's more relevant is that I don't have direct access to what you are conscious of. I assume that you are conscious, but your conscious experiences are yours alone. You can tell me about them. You can write stuff down about them. I can ask you questions about them. But you alone have them. The philosopher Tom Nagel, who uses the definition of consciousness that I quite like, he said for a conscious organism, there is something it is like to be that organism. There's something going on to be a conscious creature in the world. He used that definition of paper called what it is, what is it like to be a bat? As a question. And his answer to that was, of course, only a bat can know. We can describe how bats sense their environments. They have echolocation, they have a kind of biological sonar. And we can imagine what it might be like to have that. Is it something a little bit like touch, something else? But only a bat really knows what it's like to be a bat. And this is a challenge for doing a science of consciousness, because we can't really be fully objective about the data. And so for a long time, it was considered, okay, this is mainly a philosophical question, or it's so mysterious that it's the realm of just some sort of speculation, maybe consciousness is everywhere, maybe it's due to some as yet unknown principle of physics. And when I was doing my undergrad degree, you know, starting out in the 90s, you know, advised very strongly, don't touch that. Don't go there. But the tide was beginning to turn because it is so central, it is so interesting. And it's actually very relevant. We think about the millions of people with psychiatric conditions of one sort or another, people who suffer brain injury, and lose consciousness, or it's not have alterations to their consciousness. It really matters. And of course, it is something that science can study. We now have brain imaging devices, so we can look inside the living brain and see what happens when people have different kinds of experiences. And we can manipulate consciousness as well, we can see what happens when people go under anesthesia. We can design experiments that carefully control the things they might experience. So over the last 30 years or so, there's been a resurgence within the science of consciousness, where it's once again taken much more seriously by neuroscientists, psychologists, computer scientists, and so on. And I feel very lucky to have been working at a time where my interestedness, which I think was there from a very early age, suddenly became available as a career path again. Yeah, it's so phenomenally interesting. And I want to actually use a different sense of the phenomenology being incredibly interesting. You talk about three different ways you might approach this topic of consciousness. And one is a big word which we sometimes hear and for most of us, what exactly does that mean? You say you could consider consciousness as phenomenology, you could consider consciousness from a functional point of view, you could consider consciousness from a behavioral point of view. What do you mean by that, those distinctions? Yeah, there's a good three words to put out. I have to say when I was writing a book, phenomenology is a bit of a technical term and I sort of used it thinking there's no real alternative. But when I was recording the audiobook, I started to really regret it because when you have to say phenomenology over and over again, you make some mistakes. But I use the word because it picks out what for me is really the central feature, the central thing to explain about consciousness. Phenomenology is a description of the experiential nature of a perceptual experience of a perception. So when I open my eyes and I see the room around me and I see through the window to the house across the road, I experience colors, I experience shapes in different positions, I experience objects in relation to each other. And there's a characteristic, that's the phenomenology of it. It's not just that my eyes are telling my brain, there is a house there, or the table is brown or something like that. I have experiences that are that manifest in their characteristics of having color, having shape. You feel it. In contrast, something like an emotional experience. You feel it and you feel it and you feel it in a particular way. But then other kinds of experiences like a smell or an emotional experience, you feel it in a different way. So an emotional experience, I feel it in my body, it has valence. It feels good or bad. So these are the ways, these are the things we're trying to explain about consciousness, why an experience has the character that it does, why it has the phenomenology that it does. And the challenge is back to what we previously said that the phenomenology for me is mine alone. I can describe it, but nobody else can experience it. Whereas for these other aspects of consciousness, function and behavior, then those are more shareable. They're a bit more available to the normal methods of science. So behavior is when I'm conscious of something, then I can do certain things in virtue of being conscious. If I see a glass of water on the table in front of me consciously, I can pick it up and drink from it. I can move it away. There are many things I can do in virtue of being conscious. And then the functional aspects of consciousness is sort of that just more broadly. What does consciousness give to the brain? And typically the answer is when some mental state like a thought or a perception is conscious, it means that it can affect many different things that I do. I can remember it, I can talk about it, I can use it to guide my behavior. But things that remain unconscious don't generally have that flexibility. So there are these differences and these are very important in how we study consciousness. But fundamentally, at least what I want to get at is the phenomenology. Why do experiences fuel the way they do for the subject? So just to continue on that, and if I could go back to the analogy of the machine, the smart machine, which I think for us these days is inevitably part of our reflection, a self-driving vehicle perceives a pedestrian in front. Functionally, it is programmed to stop before hitting that pedestrian, otherwise we wouldn't let it on the road. And behaviorally, it applies to brakes. So that self-driving vehicle, which is a perceptive machine, has two attributes of consciousness, but we presume that the car doesn't feel that feeling, oh, there's a pedestrian in front. So that is the part of the phenomenology that is the difference of consciousness from pattern recognition, for example, which we think the car does. The sensor on the car is able to see the pedestrian, it's able to functionally respond to that by stopping the car, but we don't think that it has the feeling, oh, I'm a car, I'm on the road and I better stop before I hit that person. Is that right? I'd go most of the way with you on that, but there's quite a difference of opinion. So there will be some people, I think some of my colleagues would be maybe a little bit wondering whether the car has a modicum of consciousness, because there are some views in this field that say, no, consciousness is only a matter of function of the right kind. You get all the functional relations right, you program a computer in the right way or you build a machine in the right way, then it's conscious. That's all there is to it. That's consciousness only a matter of cognitive function. I am not convinced by that. I think it's unsafe to assume that. And we can talk about why later on. But the other interesting point about the self-driving car is that, yes, a self-driving car is very good at certain things. It's very good at driving or almost very good at driving. They're still not quite as good as they need to be yet. But they're surprisingly good at driving. But they're not good at anything else. They're very specialized. You can't ask a self-driving car that you can't just suddenly start having a conversation with it about philosophy. It's not going to make you breakfast. So the level of its function is not the same as sort of functions that we humans have in virtue of being conscious. When we're conscious of things, we're very, very able to deal with new situations, environments, generalize very quickly to new challenges. So some people would say that when we have an artificial system that reaches some threshold of functional competence, and they usually set this at the level of a human being, which I think is a bit anthropocentric. It's like treating us as the gold standard for everything. Then consciousness happens. Then the lights come on, the inner lights come on. Now, I don't think that's a safe assumption because just have, for me, just having functional competence is not the same as having phenomenology. And in fact, you don't have to have that much functional competence to have phenomenology. We, you know, we imagine or I might imagine that young babies are aware on the animals that are not human, but still behave in their world. They probably have conscious experiences without having much in the way of human. We do think they feel that they have, they feel the phenomenon, they feel, and another word to add to this, which, which I like, I don't know if the neuroscientists like you refer to it, qualia, the idea of that feeling of something. I doubt the car has it. I would wonder whether your colleagues that think the car is conscious has the qualia of being a car or the presence of a pedestrian. But this question of feeling comes to the statement you made, you said that there's the feeling that the light goes on inside. And I just want to ask about this basic problem. You describe yourself as a physicalist that there's a material world and follows material laws and that the purpose of science is to understand. And in fact, you say that the purpose is to explain, predict, and control. So three purposes, but of a physical system. And that sounds right to me. And I share that, share that view. And so the question is just even stepping back to the most basic rudiment, which I think is why it's called the hard problem. How could a physical system of carbon and oxygen and nitrogen and other atoms feel anything? In other words, you could imagine that they interact with the environment, that photons hit them. You could even understand, because we do even make artificial neural networks in our smart machines, how our wetware, as you call the brain, would have a neural network that could figure out that I'm speaking to you or that's a pedestrian in front of the car. Because we could understand how information could be processed that way. But how in the end could a physical system that we believe is physically a system of underlying atoms, basically, how could it have feeling in the end? How could it say, I'm here? And gee, isn't the universe interesting? And I'm so lucky I'm speaking with Professor Anil Seth about all these things. How do I feel that even if I'm just a physical system that's in that sense no different from the computer or the self-driving vehicle? And that really is the heart of the matter. It is why it's called a hard problem. So that phrase is due to the philosopher David Chalmers, who famously, again about 30 years ago, distinguished the hard and easy problems. And the easy problems for Chalmers were all the problems about how the brain works, how it transforms inputs into outputs, how it guides behavior, that don't involve consciousness, though, for which you can set aside consciousness. And the hard problem is exactly this. How and why does any physical system, whether it's made of neurons and carbon or nuts and bolts and silicon, give rise to any kind of subjective feeling whatsoever? And people have often stumbled against this. And one response to this hard problem is to say, oh, we need some super radical solution. So some philosophers would say, actually, consciousness doesn't exist. We're so mistaken about there even being a mystery here. All we have to explain is function and behavior. That's it. I don't buy this because it's for me denying the central thing we're interested in that, in fact, is something it is like to be conscious. The other way of avoiding the hard problem is to say, ah, consciousness is everywhere. There's this view in philosophy called panpsychism that says, okay, since we can't explain how it comes about, let's just sort of assume it's everywhere. You know, it's a fundamental part of nature, like mass or energy. This, again, is just an easy get out for me. It doesn't explain anything, even if it happens to be true. I don't think it is true, but it doesn't even then get us anywhere. So what do we do? Well, first thing is to admit, I don't yet know, and I think anybody does, how this hard problem is solved. But not all hard problems need to be solved in the sense of finding the special source that magic experience from mechanism. Sometimes the history of science proceeds differently. So I always use as an imperfect analogy what happened in the study of life. And by the way, before you get into that, I just wanted to say I loved this part of the book. I mean, I love so many parts of the book. But if you could take it a little bit in detail for readers, because they may not know at the 19th century, vitalism was the theory of life. And we have a completely different view now. And so I found that very compelling from an idea of how science proceeds. And it's a wonderful story. So sorry to interrupt. I'm very glad. No, not at all. So absolutely. So there was this mystery, which is still to some extent, we don't understand everything about life and how a cell works and what the real differences between the living and the nonliving. But 150 years ago, let's say, there was this deep sense of mystery. And it was almost exactly phrased in the way that you put the hard problem that how could it be that anything made of physical stuff could have the property of being alive? You can see some things that are alive and some things aren't. And when an organism dies, something happens. But what? Nothing seems to leave the system. There are all these weird experiments that were done at one point where people would weigh animals at the point of them, they're dying to see if they could measure how much life or the soul weighed. And I think it was supposed to be 49 grams or something was the idea, but it was just a badly done experiment. But the concern at that time was just this was a problem that could not be addressed using the normal methods of science. There had to be something over and above. There had to be something different. And that was the Ellen Vitao, the spark of life, this, this essence of life that explained how somehow a collection of nonliving parts could be a living thing. So you would take all the atoms together and then add the Ilan Vitao and then it could become alive. Exactly. Yeah. The atoms or the molecules or whatever it was, but you'd need this special stuff. And without it, you would not have life. And of course, as biology developed, this idea of vitalism, even though it's still quite intuitive, just became something that biologists no longer needed to reach for. And the interesting thing wasn't that they found an alternative to the Ilan Vitao, it's just that they realized, well, life doesn't work that way. There is no single thing that makes the difference or marks the difference between the living and the nonliving. So life as we think about it now in biology is really a collection of different processes. It's a collection of metabolism, reproduction, homeostasis, the body regulates living things, regulate to remain in a certain range of physiological viability. There are all these different processes. And once you've explained each of those processes, then you find that you've actually explained what it is for a system to be alive. And so the hard problem of life wasn't solved head on. It was gradually dissolved by identifying the features of living systems and explaining each of them in turn. And that's the approach to consciousness that I take in the book and that I've been taking in my career for the last 20 years, is not to solve the hard problem of consciousness, but to dissolve it, to think about the problem, not to treat it as one big scary mystery in search of a Eureka solution, but to identify its different properties and go as far as we can in explaining each of these properties. And the hope is that by doing this, we dissolve the hard problem. I don't know if we actually will, but we might. And I think we're just going to make a lot more progress that way than by bashing our heads against the brick wall of the hard problem. And I think it's right to say that if Craig Ventner or some other scientist one day announces, I have created artificial life by taking the following non-living molecules, combining them, and creating an energy substrate and so forth for them. It would be obviously a big scientific advance, but it would not be a whole new state of belief about science. It would be, yes, that's the path we are on. This is going to happen. In fact, it is happening in certain ways and stepping up the ladder, not the full, not the full step from beginning to end, from soup to nuts, as it were, but it is happening. And the question is, then, will the same step-by-step process turn out to unveil consciousness and perhaps create artificial consciousness as well, which of course I want to come back to in our discussion. But as you take this piece, this step-by-step approach, the book goes through different components that you argue are vital and they're fascinating. One is the level of consciousness, one is the content of consciousness, and one is the idea of self-consciousness. If you could take us through each of these, because all of them open up remarkable facets of what you were trying to unpeel as you take apart this overall phenomenon. So maybe to start with the level of consciousness. Level of consciousness is one of the ways of carving up the whole big problem of consciousness, this question of intuitively how conscious we are. And I mean this in the sense that we lose consciousness when we fall into a dreamless sleep and certainly when do we go under general anesthesia and then we regain consciousness when we come back again. So there's a sense in which the level of consciousness varies in these different conditions. And there are a couple of things I think that we've learned over recent years. The first is, is consciousness really something you can measure along one scale? There are some things in science like heat, which do turn out to be measurable along a single scale, temperature. And once you've done that, you've done everything about heat that there is. But consciousness I think is probably a bit different. It probably doesn't turn out to be something that you can arrange along a single scale. It's probably more multidimensional. But even so, there's been quite a lot of progress in coming up with measures of conscious level, even if they're not the whole story. And I think this is exciting just from one point of view, because the ability to make measurements, even if they're imperfect is always a transformational part of coming to a scientific understanding of something. If you want to understand spaces, then you need to be able to make measurements. These developments are exciting just for what they promise. And even for where they are already, they're surprisingly, at least for me, surprisingly informative. So one of the challenges of measuring how conscious you are is to do it in a way that doesn't rely on overt behavior. It doesn't rely on me telling somebody whether I'm conscious or not. But just by looking at the brain dynamics, what's going on in the brain and work that's been pioneered by neuroscientists in Madison, Wisconsin, Giulio Tononi and his colleagues, Michele Massimini and others have really advanced this. And the technique they use is quite simple. And it just involves injecting a pulse of energy, electromagnetic energy into the brain just using an electromagnet. And you then use electroencephalography, EEG, to record the echo. You can think of it a little bit like throwing a stone into a pond and seeing how the ripples spread out. If the pond is very still, you just get a very simple pattern of ripples. And that would correspond to an unconscious state. You still get a response, but it's quite a simple response. But if you throw the stone into a pond where there's some wind on it and some other activity, maybe some fish swimming about, the ripples become a lot more complicated. They bounce off each other, they refract and so on. And if you measure the complexity of the ripples here, that would correspond to the brain being in a conscious state. And this research has reached a stage where you can now indeed put a number. And for a particular person, that number roughly corresponds to how conscious they are. Now this is already quite incredible. As you point out, it has direct and crucial clinical applications because you could have a patient after an accident in a vegetative state, not conscious, but also in what's called a locked-in condition, where it appears that the person is completely unresponsive but actually conscious, even hearing what people are saying in the hospital room. And as you point out through really interesting, rather bizarre kinds of experiments, able to communicate by what their brain shows on a brain scan. That's right. It's actually the vegetative state that's the biggest challenge here for people with locked-in syndrome. It's a very, very tragic condition. But there is actually very little ambiguity that these people remain conscious. They can often still move their eyes in response to command. And if you put them in a brain scanner, you'll see that most things are quite normal. The real challenging condition is vegetative state. And this is again a condition that people might enter after brain damage of one sort or another. And the characteristic of the vegetative state, and there are tens of thousands of people, if not more in this condition across the world, is that people will still go through cycles of waking and sleeping. Their eyes will open. They might look around. They even might move their bodies a bit. So they're not paralyzed. But that just doesn't seem to be anybody home. They don't respond to commands. They don't interact. They don't engage. They don't focus where you're showing something. So the impression is that they're awake, but not aware. They have wakefulness, but they don't have consciousness. But how do we really know this? Because it may be that they are conscious. They're just not able to express it through behavior. And this is where techniques like this way of measuring conscious level and other methods have really come into their own. Because these are ways of getting around the inferring just on the basis of behavior, looking at what's happening in the brain, and using that information to help reach a diagnosis or a prognosis about how this person is. Are they actually conscious, or are they more likely to recover? And indeed, there are other elaborations of this sort of method that have been used for communication. So my colleague and friend, Adrian Oh, and pioneered these studies where instead of zapping the brain with some energy and listening to the echo, you do something which sounds rather strange. You just ask them a question. And the question is something like, imagine playing tennis. And of course, you know, this is strange. This person is not responding to anything. So why would you ask them to imagine playing tennis? Well, it's because if you imagine playing tennis, parts of your brain have to do with planning smooth movements, fluid movements, are reliably activated. It's not just your auditory cortex, because you're hearing something or even your language areas, because there are words, you will engage areas to do with playing tennis. And then if you ask the same person to imagine walking around the rooms of their house, well, there's auditory and language involved, but there's also other areas that are involved in navigating around complex spaces. And these parts of the brain are very different. So if somebody who is who you're not sure whether they are conscious, if they actually, if their brain responds in this, in this same way with these different areas lighting up under these different tasks, well, that's pretty good. That's a pretty good indication that they are conscious. And then you can indeed go the next step and say, okay, well, now imagine playing tennis for yes, and imagine walking around your house for no. And you've got a very rudimentary means of communication. But frankly, you know, that's that's very low bandwidth, but it makes all the difference in the world. And it is a very strange situation, though, I was involved in developing a radio place some years ago that took that as a, as a premise, you know, what, what could you do if your communication was restricted to that, to that level, you can go so far. But, you know, there are certain difficult questions. Can you ask a person whether they want to remain alive or not? That's a very ethically challenging area around this whole field of research. But it just shows that the science of consciousness is not an abstract philosophical armchair adventure. It's making a real difference in the clinic and in the world. And, and Neil, is it this may be so crudely put and, and therefore, absolutely not right, but can we think of consciousness in this way as certainly relating to the complex networking of neuronal firings. So, consciousness at some level, and this five function of, that you refer to as well, says that consciousness is a, is a brain function in which the neurons are connected with a certain level of complexity. And that's what you're measuring. And that somehow turns out to be the correlate of the conscious experience. Yes, mostly. So, again, there's this challenge of, of, we can say these things are going on in the brain and they correlate with being conscious in a certain way. But did he really explain it? That's, that's the challenge, right? And so some people, you know, Giulio Tenoni, who we mentioned a minute ago, and he has this measure of Phi, which is indeed this measure of the complexity of the interactions between these parts of the brain, will make the case that no, that is consciousness. That really is in the sense that, that's, you know, the heat really is the mean molecular kinetic energy that there's this sort of, in philosophy would call this relationship of identity. And consciousness just is this. Now, I don't go that far. But the reason I still think it's very interesting, not, not only because it works, which is great, but this property of, of complexity of the brain being complex in a particular way, which involves being able to go into many different states, but all these states being connected, being unified. To me, that explains something about the phenomenal phenomenology of consciousness, the experiential character of consciousness. Every experience we have is both highly informative. It's different from every other experience and unified. So you can draw a link, you can draw a bridge, you can say, okay, conscious experiences are this particular way. They're unified and integrated and informative. And so the underlying brain dynamics should also have that property and the extent to which they have that property, things will, will be conscious. But it's not the same as saying that is consciousness. It's, it's drawing, it's building an expansionary bridge from the one to the other, just as with life, you would say, okay, living things have this property of metabolism and these, these circuits within metabolic circuits within the body have this property. So that's part of what being a life is all about. So it's one part of the puzzle, but not, not the whole story yet. And so that brings me to the second part of your puzzle, this wonderful intermediate part of the book where you talk about perception and our sense of what we see and feel as being in some way the hallucination, the hallucinatory. And this is a wonderful unfolding. And I love for you to describe how could it be that conscious perception is controlled hallucinations is how you describe it. It's true. And the word controlled is super important here. I think we can get a little bit off, off piece if we, if we focus too much on the hallucination. But yeah, this is the sort of second aspect of consciousness. This, we have a conscious level, but then when we are conscious, we're conscious of a world around us. We're conscious of the objects in our immediate environment. We're conscious of just, yeah, being in a world. And how does this happen? What explains the nature of the our experiences of the world? And I like this part of the book because I think it connects most closely to our everyday experience. You open your eyes in the morning and there's a world there. How does that everyday miracle happen? Now it seems intuitive for us that when we perceive the world, it's largely a process of the brain just reading out the sensory information as if the world just pours itself into the mind. The light comes into the eyes, the light comes into the eyes, the sound comes into the ears, the molecules come into, into the nose. And, and you just process that is, is what you're saying. That's, that's the intuition. That's the intuition. That's the common sense of you and it's processed in complicated ways, but basically it's all, the heavy lifting is all done in this direction of the world coming into the brain. And maybe the self is there at the center, which is receiving all this information and doing the perceiving. But another way to think of it, and this again, this is an idea with a very long history, but just gathered momentum over the last few years, is that perception is much more from the inside out, all the top down in terms of the brain than it is from the outside in or the bottom up. In philosophy, this goes right back to Plato and his allegory of the cave where prisoners see shadows on the wall cast by the firelight and they take the shadows to be the real world because that's all they, they know. And then Kant much later talked about the idea of the Newman and that there's this whatever reality really is, we never have direct access to it. We can only ever see it, perceive it through a sensory veil. And much more recently, this idea is exemplified by the notion that perception is a process of inference. So instead of it being a direct readout of sensory information, the brain is always having to interpret this sensory information. Sensory signals, light waves coming into the eyes, sound waves coming into the ear are always ambiguous. They don't come with labels about where they're from, what object they're from. They're just patterns of energy. And they may differ. You take a piece of white paper from inside to outside, the light waves hitting your eye change dramatically, but the paper doesn't seem to change color. The brain is always figuring out what the sensory information means. And one idea for how this works is that the brain is always generating predictions about the causes of its sensory input. So it's making predictions about what's actually out there in the world, like a piece of paper or a dock. And the sensory signals, instead of being read out to give us our perceptual universe, they just serve to correct the predictions, update the predictions, calibrate them, keep them closely tied to the world. But what we actually perceive is the content of the top down predictions. Our brains are always guessing what's there, making best guesses. And that's what we perceive. And this is why I call it a kind of controlled hallucination. It's a term that I heard from Chris Fritha, one of my mentors and other people. But I kind of like it because we typically use the word hallucination to describe perceiving something that isn't there, that other people don't. And I'm not saying that this happens all the time. No, our perceptions are closely related to what's there, but there's a continuity here. So this is why our perceptions are hallucinations, but they're controlled by the sensory signals, which is why they're not arbitrary. They serve a clear purpose. I was really taken by your description of the brain and the analogy with the Plato's cave, as you mentioned, where Plato said that human beings are always in the darkness and we see reflections on the wall. But the brain, as you say, is completely in the dark. It's completely in a case in the skull. It has no direct access to the world. It has to figure out everything. We think the ideas are somehow emerging in the brain, but the brain's guessing, okay, my eyes are telling me something, but what's going on here? And that you're saying is the inevitable normal process of everything that we perceive. And then as you illustrate with these powerful examples, because we're making guesses, we can also fall prey rather easily to tricks because the brain uses certain heuristics to guess it's this or that. And by playing on those tricks, we get optical illusions and mental distortions that we otherwise wouldn't understand. That's right. Pretty much every visual illusion, optical illusion, magic trick, or as you say, psychiatric condition can be understood this way. The really tricky thing for me to actually recognize in myself and convey is that this applies to normal perception as well. It's not that if only we could see through our illusions, we'd see the world as it really is. If only we could just banish all these distortions, we'd have direct access to the world. This is the slight issue with illusions. They always suggest there's a correct way to perceive things that we're being tricked. And for illusions, that's the case. Two lines might look different lengths, but they are the same length. There's an objective truth to that. But where that metaphor misses the mark is that for our normal perceptual experience, evolution didn't design our brains to perceive the world as it really is. And in fact, that seems to be not sure what that would even mean. So all our perceptions are constructed. And some of them, when they deviate too much from what's out there, then they become very problematic for us. Then we might start hallucinating as the word is normally used. We'll see things that really aren't there that other people don't see. But all of our perceptual experiences are inside out constructions that are calibrated by sensory data. And so this is why illusion is a bit problematic. And it's also why that the really tricky thing to get heads around here is that it doesn't seem like that to us. For an illusion, we can see it because we can look at it two ways. We can think, oh yeah, the lines look the same length or these two patches of color are the same color. But when we look around in our daily lives, it just seems as though the car really is red and that that redness is a mind independent property of the car. We don't experience it as being the brain based construction that it really is. I don't know. We froze again. Shall we just try to push through? Let's try. And if it freezes again, we'll have to refresh. So Anil, you introduced a key word, which actually I don't think arose in our conversation yet, but it's central for the next part of the book. And that was the word evolve, that our consciousness evolved because we are species that evolved and the consciousness evolved not to give us precise readouts of nature, but evolved for our survival. Evolution is guided not by how accurate your perceptions are necessarily, they're guided by how successful you are at surviving and propagating the genes that give you your being, your being you. And so this is really the theme of the self part of the book that consciousness is an evolutionary construct for the survival of the organism. And we're therefore conscious of all sorts of things. And our emotions are part of that consciousness as well. They're not just feeling good or bad. They evolved because they are readouts for us of how we're doing on this struggle for survival. Could you describe that vantage point for us? Yeah, definitely. And this brings us on to the third of the major divisions of consciousness that I talk about, the self. And again, there's a fundamental intuition to be challenged here, which is that the self is the thing that does the perceiving. That's reading out all these information from the sensory signals. It sure feels that way, but it does feel that way. But I think the same lesson applies that the self is not the thing that does the perceiving. There's no mysterious essence of me lurking somewhere inside my skull. The experience of being a self of being me or of being you or of being anyone is just another collection of perceptual experiences. There are other kinds of controlled hallucinations, but they have a different character. So whereas my experiences of what I see in the world around me are objects and shapes with colors and so on. The experience of being a self has a very different character. It's composed of, as you mentioned, it's composed of emotions and moods. But there are also other aspects. There's the experience of identifying with a particular body. So part of the universe is my body and the rest of it isn't. That's a fundamental aspect of being myself. Then there's experiences of volition and agency, of intending to do things, what people often call free will. And then only above that, there's experiences of being a continuous person over time with a name and a set of memories and plans for the future, this narrative self. But all of these things we tend to experience as bundled together as the nature of what it is to be a human self. And the view that I take is that all of these elements of selfhood are, well for the first thing, they're potentially separable. And we notice from various neurological and psychiatric conditions as well as increasing numbers of experiments where we can tease apart these different elements of self while others remain. And they're fundamentally guided, as we said, by this evolutionary constraint to stay alive. So just as we don't perceive the world as it really is, we perceive it in ways that are helpful to guide our behavior. The whole idea of perceiving a self, I think, gives our conscious experiences a scaffold on which our behaviors are arranged so that they keep us alive. The most fundamental experience aspect of being a self is not, for me, the higher levels of personal identity. It is this fundamental sense of being a living organism. Things are immediately bad or good. Things are likely to be bad or good in the future. And this core level of experience of selfhood is what emotions and moods are all about. And my view, for which there's some evidence now, is that these aspects of perception are forms of controlled hallucination. The brain is still making predictions about sensory signals. But when it comes to emotions and moods, the relevant sensory signals come from within the body, not from the world outside. By the way, what a brilliant insight of William James, who is one of the founders of modern scientific psychology, to realize that emotions and moods are in some sense these readouts of internal states that then are interpreted in certain ways exactly as you're describing. But that was not the view of emotions before William James, really, as a kind of perception, but in internal perception. That's right. So like all these ideas, there is a long history and William James and Carl Lange were among the first more than 100 years ago to come up with this idea of emotion as being a readout of changes in the physiological state of the body. And this view has continued to evolve. And there's still a lot of debate even today about whether, for instance, there are some basic categories of emotion that are shared among all creatures, among all humans, but also other creatures or whether emotions are always constructed or always acts of interpretation. And I think it's probably neither quite one or the other. But for me, the important thing is that there's a great commonality between emotions and other forms of perception. It's, I think, natural for most people to think of emotions as being completely different to visual experiences. We might see something and feel joy, but they seem to be different classes of experience altogether. And I think there's much more continuity than is typically appreciated. They're all based on the same common principles of the brain sending out predictions and updating these predictions with sensory data. As I was reading this part of the book, I was thinking, I was hoping you'd get to, but now I know it will be in your next book or in a forthcoming book. You make a very convincing point that because of the evolution of consciousness, because of what we know about brain circuitry, because of what we know about behavioral observation that consciousness is something that is common with other animals, maybe even very simple animals, but certainly mammals and probably a much broader class of animals. But I wanted the next part to be what is distinctive about human consciousness compared with other animals. And for me, my guess, and I'd like your view about it, the self-awareness part or the self-conscious part or what you call metacognition, thinking about our thinking, or the fact that we not only have a self, we feel that we have an organism, but we know or we feel we know we're here and there's a universe and I can even look out at the universe or I can even look inside myself, though it's hard to do. That level of higher level consciousness, I would guess is distinctive to humans in some way and I would just love your thought about it. I've assumed that the additional prefrontal cortex that was added in human evolution plays some distinctive role in that added piece. Maybe that's completely wrong, but is it right to think about this self-consciousness, self-awareness as probably being distinctive at least to the higher primates and maybe to human beings as a special feature of us? The first thing is you make a very important point here, which is that there is a distinction, a critical distinction between consciousness in general and self-consciousness in particular and it's often too easy to conflate the two things and we think, sometimes say being conscious is about being self-conscious, that being consciousness is all about yes, it's because I know I'm conscious, that's what consciousness is, but no in fact it's possible to imagine organisms, other creatures, even infant humans or even people with very advanced Alzheimer's or dementia, you can think of them as possibly being conscious but without having much self-consciousness going on, so the two are indeed separate. They're feeling, I mean people feel things, they experience things, they have moods but they may not be self-aware that they are experiencing it or reflecting on that. That's right and another thing is that ethically that doesn't necessarily matter that much so we have this long pretty atrocious history as human beings of treating animals rather poorly perhaps because we assume they don't have self-consciousness but they may still very well have the capacity to suffer, experience pain, experience pleasure too, all in the absence of a sort of selfhood that we humans have and so we shouldn't use the elaborate selfhood that we have as benchmarks for how we treat other creatures and I do go off in a bit of a diversion in the book, one of my favorite diversions was about the history of criminal animal prosecution. Absolutely amazing, never heard of it before, completely amazing and you have a picture of the pig in the dock actually being charged with the crime and tears coming out of the eyes. I mean it is, it's borderline insane it really is but it just shows that our views of other animals have changed over the centuries in ways that we may not completely have expected or recognized without looking into it but the second, the question you asked specifically was is self-awareness, is that distinctively human, is that really what sets us apart? There's of course a danger to that because again there's probably much more that joins us that we have in common with other creatures than sets us apart and thinking always in terms of what sets us apart, this anthropocentrism has led us astray many times but there is of course something distinctive about being human, there's something distinctive about being anything but there's something particularly distinctive about being human, what is it? It may be aspects of self-awareness, I don't think it's the whole thing but there are some interesting clues, one of the more interesting clues and I still don't really know what to make of this is this task called mirror self-recognition so if you and I look in the mirror we recognize that we're looking at ourselves, that's a very natural thing to do, we assume other people can do it too, everybody can do it but we can only do it after about 18 months of life until we're 18 months old we don't recognize ourselves in the mirror that just comes online about this but we'd never say that infants who are under 18 months aren't conscious or even that they don't have any sense of self, we would just say no their sense of selfhood is not elaborated to the extent that they look in a mirror and recognize that it's me and if you adapt this test to other non-human animals and you do this typically by for instance painting a mark on the forehead of an animal and then presenting them with a mirror and do they use the mirror to investigate the mark or do they sort of behave as if the creature in the mirror is a is a conspecific as another animal then remarkably few other species pass this test so some gorillas do, orangutans do, the odd elephant has been shown to but cats don't, dogs don't, monkeys don't, you know some claims that some fish do but it's very hard to know exactly how to interpret that it's a very different kind of behavior the point is that this is this is a very limited thing very few other creatures do it naturally if at all apart from humans so there is something distinctive going on here and that might be because it might have something to do with our and this is this is now speculation it might have something to do with us being really very highly social creatures I mean it's very very important and we of course there are many other species of social as well but this particular kind of primate sociality that we have I think is not seen so much in other in other creatures so that might be one distinctive element for me though the thing that really sets humans apart is our language people often debate do other species have language and you can certainly see elements of language you can see rich vocabularies in the calls of merma set monkeys and vervet monkeys dolphins and whales certainly communicate but communication is not the same as a rich generative language that allows us to combine elements in new ways to say things that have never been said before I don't think and this is a little out of my wheelhouse but I really don't think there's convincing evidence for the rich kind of language that humans have in any other non-human species now that makes us distinct in a number of ways the first thing is it gives us a kind of culture that that is cumulative it's through language that we can build on the knowledge of previous generations and of course if you just objectively look at what's distinctive about humans it's the cultural legacy that we have for for better or worse in our impact on the planet other creatures don't don't have that at least they don't it's not manifest through culture you could say that the nature of the world is much more impacted by fungi than by humans but not in virtue of language we can see this specifically in music too one of my colleagues and friends Arnie Patel at Tufts has spent his career studying the connection between language and music most of us if we hear a beat we just start tapping along to it naturally and there are some people who who have this condition called amusia where they claim to not have any sense of musicality but most people have a sense of music and and the grammar of music and the grammar of language seem to be very closely tied together it's about constructing patterns over time as well as at a time musical phrases unfold just as linguistic phrases unfold and fascinatingly if you look at other species as well you find species that are otherwise very smart like monkeys but they don't tap naturally along to a beat at all rhythm is just something they have to learn and it's very hard for them to learn but other creatures like songbirds some songbirds naturally entrain to a rhythm so I think there's some fascinating clues here about the connections between language music and the brain and what might make us distinctively human I'm counting on that as I mentioned to you for an upcoming book because I think that that'll be absolutely crucial let me turn to the last the last part of your your book if I might and and that is artificial intelligence and machines and machine consciousness and my part of my reflection is first the the artificial intelligence revolution of of the last 20 years in effect though it dates back already about 75 years in fact is remarkable because who would have predicted that a neural network as a mathematical construct would be able to do such wonderful things as becoming the world's greatest chess player ever or the world's greatest go player or to drive a vehicle in complicated terrain so for me this success of artificial intelligence is profound the luck in a way it seems maybe it's not luck but the insight to use a mathematical construction that we call deep neural networks which are connections of mathematical objects that in a way mimic the connections of brain neurons and how they fire and with different weights of firing like synaptic weights shows to me it has exposed what is intelligence in in a very important way that that knowing how to play chess is a is a matrix of numbers after all it's a set of connections that if you were able to play against yourself 40 million times in four hours because your computer not a human being you can get pretty darn good at at chess but in the end it's just a matrix of numbers that defines the intelligence of chess or similarly a neural network that can perceive and categorize objects shows that perception is is in a way a mathematical process of neural connections so i guess my question to you because you're you're exactly at the interface and i believe in an engineering school also so dealing with the artificial intelligence and the human intelligence for me a number of times scientists say well that's not really intelligence i think what it shows is intelligence is not really what we make of it also in a way it it is a construct not surprisingly of physical objects that if repeated enough actually can do amazing things so i'm in love with neural networks of course we're surrounded by them now in our daily lives artificial intelligence but what do you make of that of of the revolution in terms of comparison with the with human intelligence i think the progress has been extraordinary my phd was actually in ai in artificial intelligence started in the late 90s at what was probably now considered a relative dip in progress a relative flatlining of of activity which picks up again with this with the innovation of deep neural networks and algorithms to train them and now indeed these algorithms are capable of doing superhuman things beating the world's best go player beating the world's best chess player long ago now and object recognition near human or superhuman performance this is this is amazing and i don't want to undermine the progress and it's still it's still continuing having said that there are still i think very significant differences one of the holy grails of research in ai is to achieve what's been called general artificial intelligence general ai or g ai and this is the idea of a system that has the general functional competences of a human being because we not all of us can play chess but if we can play chess we can also do other things we can walk around our house we can open our eyes and classify objects that we see into different categories and we can speak we can do lots of things but a neural network that is trained to play chess can really only play chess a neural network that is trained to drive a car can really only drive a car so the ai systems that we have today are highly accomplished but they're still relatively specialized so that's a big transition that's that's um still to be made and so when we say something like this ai system shows that perception is just you know matrix of numbers being multiplied together in a particular way i think what it really shows is that yes aspects of perception can be implemented that way but maybe humans still do it differently and there's another clue here that in fact we don't need to play ourselves 40 million times in order to get good at chess we only probably need to do it a few hundred times we don't need to be exposed to every single cat on the internet in order to learn to classify cats we just see them a couple of times so we learn things very very quickly we generalize to new situations very rapidly so there's a hint that there are some still quite fundamental principles of ai that remain to be solved and intriguingly and this is an idea that i've been exploring with with actually colleagues who are pioneers in deep learning people like Yoshua Benjio in in canada um we're both involved with this canadian institute for advanced research the things that cutting-edge modern ai systems tend to lack seem to be precisely the things that we humans are good at in virtue of being conscious it's when i'm conscious of something that i'm able to behave very flexibly with respect to it you know move the put the chess board away rather than making another move i can do that in virtue of being conscious of the chess board if i if i perceive something unconsciously and we do you know our perceptual systems can react to things even if they don't surface at a level of conscious experience we can't behave flexibly with respect to them so i think this identifies for me a very exciting nexus of future research is what is it about on the one hand what's missing in current ai that gives it the sort of flexibility that we associate with conscious intelligence and conscious perception in humans and can we use that insight to understand what's going on in human brains that characterizes the difference between being conscious and not being conscious and of course it goes the other way too as we understand a bit more about what consciousness allows us to do as behaving and intelligent human beings can we use those insights to help guide the future development of ai beyond just adding more training data adding more layers to your deep network and so on that big remaining question is that when we do get there does that mean the lights come on for this super new next generation general ai device that can make you breakfast play chess and drive around tokyo and the first time it's ever been there will that be conscious and here we all just rely on our intuitions there's no consensus answer to this those people who think that consciousness just is a matter of wiring something up in the right way getting it to perform the right kind of function they will say yeah for sure or yeah probably but i still have us i still have a suspicion and and this suspicion is born of the previous part of our conversation recognizing the deep roots of human consciousness or animal consciousness in regulating our status as living organisms you know in a computer there's always a sharp division between the hardware and the software you run a program on a computer whether it's a chess playing program or a deep neural network or whatever you can run it on an IBM you can run it on a mac you can run it on a computer made out of tin cans if it's sophisticated enough it's still a computer in human beings in brains and bodies there is no sharp distinction between the the mindware and the wetware there's nowhere where you can just say okay now the substrate no longer matters you could say it's neurons and connections but who knows really now every neuron is a cell that's also doing its best to stay alive and the brain is much more complicated than just a bunch of neurons connected as a network there are chemicals washing around all over the place every cell through the body has an imperative to stay alive so and consciousness itself is fundamentally grounded in our nature as as living machines as beast machines as i use a phrase again surfacing from the from the 18th century um so i am suspicious i think if we i think we will get ai systems that look as if they are conscious in fact we already have them you don't have to actually get to general ai we can have these deep fake videos that give very good impressions of being somebody and they can be they can be powered by these language networks that can make people say things that are convincing so in a virtual world we're almost already there to have things that look as if they're conscious but does that mean the lights actually do come on here i'm still skeptical well anil you say that you we're going to have to rely on intuition but i say we're going to have to rely on you you're at the cutting edge and this idea of marrying consciousness and artificial intelligence sounds absolutely thrilling so i hope we're going to be talking soon and in future years also about the breakthroughs that you're going to be bringing but let me thank you for your wonderful work thank you for this fantastic conversation you really have opened up our eyes and i i know so many readers ahead what it means to be you what it means to be us what it means to be human and we're profoundly grateful to you thank you so much for being with me today i'm the book club with jeff sacks thank you very much it's been an absolute pleasure