Latest / The Smart Cookies podcast / "Is Consciousness a Controlled Hallucination?" With Professor Anil Seth

Transcript

"Is Consciousness a Controlled Hallucination?" With Professor Anil Seth · The Smart Cookies podcast

• Zaza: Anil Seth is a professor of cognitive and computational neuroscience at the University of Sussex, where he is also co-director of the Sackler for consciousness science. He's also co-director of the Canadian Institute for Advanced Research Program on Brain, Mind and Consciousness and of the Liver Hulmet Doctoral Scholarship Program from Sensation and Perception to Awareness. His latest book, Being New, Excuse Me, A New Science of Consciousness, which I have right here because I just read it, is a Sunday Times top 10 bestseller, an economist book of the year and a guardian and financial times science book of the year. His TED talk on consciousness has more than 13 million views and is one of Ted's most popular science talks. Professor Seth, welcome to Eurotrash. Anil Seth: Thank you for having me. Zaza: an absolute pleasure. The release of chat GPT-4 and the ensuing state of utter frenzy forces us to begin with AI, of course. At least on the internet people are behaving as if Skynet is literally around the corner and as per usual we seem very optimistic about our worst fears. What's been most interesting to me though is that there's these systems, engineers, and they're kind of not denying that we are really close to building conscious machines. However, the peculiar thing is that we don't even fully understand what consciousness is, and yet we are confidently Anil Seth: And yet they're helping Zaza: predicting Anil Seth: us to predict that Zaza: that Anil Seth: we're Zaza: we're Anil Seth: disconnected. Zaza: just about to create one from scratch. What I'm missing in these debates is people like you, researchers who actually have an idea of what consciousness might be. first question to you has to be this. Do you think it's possible Anil Seth: Thanks for watching! Zaza: to create something so complex without really understanding it in the first place? Can consciousness just arise spontaneously on its own? Like let's say we release chat GPT-11 and it's so smart it's suddenly aware of itself. Anil Seth: Well, there are actually two questions there. I think for the first one, it's indeed possible for us to build things that we don't fully understand. I think these chat GPT algorithms are examples of that. People don't fully understand how they do what they do and why they sometimes do quite strange things. We can get onto that. The second question about whether this means GPT-11 or GPT-600, that's a different thing. And I personally, I don't see any prospect for that at all. And the reason is that there's a distinction which is surprisingly little made in all the debate about artificial intelligence and about these large language models, which is that intelligence is not the same thing as consciousness. And for Zaza: Right. Anil Seth: us humans, we tend to think it goes together. And I think that's a residue human exceptionalism, this tendency to put ourselves at the center of things and at the top of every pile. We think we're smart and we know we're conscious. So there's this idea kind of baked in there that as artificial systems get smarter and smarter, that at some point the lights come on and consciousness just comes along for the ride. I just don't think this is likely at all. It makes a lot of assumptions. Secondly, and I think most importantly, that consciousness is the kind of thing that computers could have. And this is just not known. I mean, there are some things that computers do, which actually generates a thing in question. A computer that's playing chess actually is playing chess. But a computer that's simulating, let's say a weather system, it doesn't actually ever get wet or windy inside that computer. just a simulation. And so the question that I think is still open is consciousness more like chess or is it more like the weather? And the prevailing view has been that it's more like chess, that consciousness is some kind of information processing that, that if you program a computer right, it will happen. But that's a massive assumption. And my research and my way of thinking has tended to suggest the opposite, that consciousness is a, is It's fundamentally related to our nature as living, embodied, flesh and blood creatures. And from that perspective, in order to get conscious AI, one first has to have living machines. And even GPT-6000 is not going to be alive. Zaza: What about like a replicant from Blade Runner? A machine that has a body and some sort of maybe organic matter or something along Anil Seth: That's Zaza: those lines. Anil Seth: yeah, Zaza: Could Anil Seth: that's Zaza: that have Anil Seth: much more Zaza: consciousness? Anil Seth: interesting. Well, yes. And once we go into the realm of science fiction, then a lot more things come on the table and actually, of course, science fiction has done a much better job than most professional commentators on trying to tease apart these things. I mean, Zaza: Oof, Anil Seth: Blade Zaza: ouch. Anil Seth: Runner is, Zaza: Nice. Anil Seth: um, is an extraordinary examination of the role of emotion and embodiment in consciousness and, and selfhood. So it prioritizes emotion in a way that other perspectives have tended to there's this, you know, we're so familiar with the Turing test and there's all this, these ideas now that chat GPT makes the Turing test redundant. It passes it because the Turing test being this test of machine intelligence. Right. So can you as a human tell a difference between interacting with a human or interacting with a machine? I think most people still can when it comes to chat GPT, but you know, maybe GPT six few down the road that that'll change Blade Runner had this thing called the VoigtKampf test. And this is a test about emotional reactivity. Does an organ, does a person, whether you, if you're trying to decide whether such and such is a replicant or a person, do they have the right emotional responses? And I think that's much closer to the nub of things when it comes to consciousness. So maybe, you know, that's, I think, much more plausible, both embodied so it actually interacts with its environment. That's also something that's lacking from large language models, which is why they don't understand anything, let alone be conscious of anything. Understanding requires, in my view, interaction with an environment. It's that grounding of symbols in sensory motor flow that gives our words meaning. a large language model that is embedded in a robot that actually interacts with an environment or that is learning during the process of interacting with an environment that could maybe understand, but it still might not be conscious. But then bringing in physiology, bringing in the capacity for emotion, bringing in basically the imperative that the system has to keep itself alive. that makes a difference. Zaza: I think you reference Daniel Dennett in the book when he says these are intelligent tools, not our colleagues. Right? At least so Anil Seth: I Zaza: far. Anil Seth: think his comment, it's a beautiful phrase. He advises us years ago. I mean, this is not a new thing. He's not just responding to the current wave of frenzy, as you nicely put it. He advises us to always remember that we're building tools, not colleagues, when we're thinking about these artificial systems. And I think that's never been more relevant than now. Part of the problem in how we're grappling is this idea that they might be very similar to us, that we, how are we going to coexist with them? How are they going to contribute rather than be dangerous for society? And it is important to remember that just like a search engine is a tool, just like a piece of paper and a pen is a tool, large language model can be a tool. It can be very good for some things and horrendously bad at other things. that we keep projecting mind into these things. We're very anthropomorphic people. We can't resist often projecting mind and consciousness into things that seem similar to us in some way. And actually I think that's the much more likely scenario. It's not that large language models will actually be conscious. It's that we will be unable to resist contributing them with conscious minds in much the same way that certain visual illusions once, even though we know what's going on, we can't help seeing the illusion. We'd be susceptible to a kind of cognitive illusion that large language models have minds when in fact they don't. And it's recognizing that, that will allow us to better treat these things as tools rather than colleagues. Zaza: Speaking of chat GPT, I have a little confession to make. I typed, what should I ask Professor Anil Seth about AI into chat GPT? Of Anil Seth: Hehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehe Zaza: course I couldn't help myself. I kind of thought it would be only fair that I let Anil Seth: Of course. Zaza: it ask you a question about itself. It came up with a dozen questions, most of them pretty boring, to be honest, but I thought this one was kind of interesting. So I'll have a go. in which humans and machines merge into a single entity. What is your perspective on this possibility? Are there any ethical or philosophical implications that we should consider here? Anil Seth: Well, well done chat GPT. That's a good, by the way, did that come Zaza: Right. Anil Seth: from GPT four? Which, which version of GPT Zaza: Yes, Anil Seth: did that Zaza: GPT-4. Anil Seth: come from? Zaza: That was literally Anil Seth: Okay. Cause Zaza: today. Anil Seth: that was today. I mean, this is one of the interesting things that I, the rate of progress in these large language models is very impressive. GPT four is so much better than previous iterations. Zaza: Yeah, I did have to give another command because at the beginning he was asking really generic, boring questions Anil Seth: Right. Zaza: and then I think I said, can you ask something a little bit more daring? And then this came out. Anil Seth: Yes, that's one of the great things you can do. You can push it in different directions. I noticed you said he rather than it. I keep thinking of them as it Zaza: It Anil Seth: to try Zaza: of course Anil Seth: and Zaza: of course. Yeah. Anil Seth: again, remember Zaza: Yeah Anil Seth: it's a tool, not a colleague. Zaza: Right, right, thank you. I need that. Anil Seth: But for the question itself, this idea of a kind of post human era, when we merge with machines, in a sense, that's already happened. You know, I think that's been happening for a long time. My, my colleague and friend, Andy Clark, a philosopher talks a lot about the extended mind. Now the idea that the boundary of our mind is not at the skull. It takes in all the things, not all the things, but many of the things we interact with, like pen and paper, but certainly a lot of my memory is on my smartphone already. cognitive ability, I would say, is not in my head anymore. It's in search engines, it's in all sorts of other things. So there's a sense in which this trajectory of merging with machines is already underway. But it hasn't reached the point where we become a single entity. I'm not entirely sure what that would mean. I do think it's not just of increasingly useful and potentially dangerous machine learning systems. There are of course more direct ways in which we are entering this so-called post-human realm. People can have neural implants. Elon Musk is working on Neuralink and things like this, which can directly introduce non-biological components into our brains. Of course, outside our brains, this is no longer a controversial thing. We have artificial limbs. things that help us. But at what point does this constitute becoming post-human? I'm not really sure what that would mean. I don't see that as particularly imminent prospects, but I do see sort of increasing integration between humankind and various forms of technology that undoubtedly will pose a lot of ethical There are lots of just, I think, quite presake, but very important ethical issues about accessibility and equity. What about lifespan? Are we able to use technologies to extend life? Until now, the fact that most people die within a few decades of each other is a great equalizer. And if that's suddenly off the table, then I think society is in a lot of trouble. By the way, I have to say, I also, like everybody who've been playing with GPT And I had asked GPT-4 to write a biography of myself and some other friends. And what I was struck by there was how it combined generic things that were true with very specific things that were wrong. So it got where I did my PhD wrong. It got my birthday and place wrong. because these facts that it was suggesting were not online anywhere else. They were just wrong. And a lot of people say that large language models hallucinate. I rather think they confabulate Zaza: Hmm. Anil Seth: and hallucination is a kind of false perception. Zaza: Right. Anil Seth: Confabulation is something we see in neurology is certain in certain conditions, psychiatric conditions and neurological conditions. People just seamlessly make stuff up, fill in gaps in a way they're not even aware of doing. And for me, that's much more redolent of what large language models do. They just make stuff up and they don't know the difference between making stuff up and saying stuff that's correct because they don't actually know anything. I don't think there's any understanding there at all. And what really betrayed that for me was when I asked the GPT-4, a bit like you, saying, you know, ask something a bit more daring. I said, write the bio again. but this time more accurately without so many errors in dates and places. And I thought that the system might be able to respond either by fixing the errors or by becoming more generic in those places where it had made errors, like saying just born in the UK rather than born in London, which was wrong, I wasn't born in London, I was born in Oxford. But instead it said born in Hammersmith, London, which is even more specific even more likely to be wrong. And it invented a whole new career stage for me in Germany where I've never worked. So it's just, I think this, you know, whenever you see something that says, oh my gosh, these systems, they can theory of mind just emerges when you train them on enough data and they get into this weird so-called interpolation regime where we don't quite know what's happening. These things emerge like theory of mind. There'll be some other domain where they're still revealed you know, the unkind phrase as being a kind of stochastic parrot. They can make errors which betray a total lack of what we think needs to be going on in order to do this kind of thing. They don't have in this case any kind of metacognition, the ability to know when they are right or wrong and do something about it. So yes, there's a long Zaza: Yeah, Anil Seth: way Zaza: I Anil Seth: to go Zaza: asked it. Anil Seth: even for GPT-4. Zaza: Right, right. I asked it to tell me about myself, and it said, I don't know this person. So my vanity was immediately offended. And I was like, that's enough of you. Anil Seth: Hehehe Zaza: You mentioned Elon Musk. I have to preface this by saying that I'm not a huge fan of the man. But I wanted to ask, what do you think of the open letter that he signed, among other people, and that says, or it calls on all AI labs paused for at least six months the training of AI systems more powerful than GPT-4. It was also signed by Steve Wozniak, Andrew Yang and thousands of other people and researchers. What do you think of this petition? Anil Seth: Well, I was one of the thousands of others. I also put my name to that. Zaza: Oh, Anil Seth: And, um, Zaza: right. Anil Seth: so I think it's important and I think there are a number of reasons for, I think it's first thing to say is with so many signatories on it, it's, you can't expect that every signatory agrees with every part of the letter and I don't, I wouldn't write the same letter. I don't agree with every part, but I think it's trying to make an important point. And it's a point that needs to be made. Right. And I, there are a few reasons. Probably the most significant reason is that the acceleration in the power and potential of these systems is taking everybody by surprise. It's taking by surprise those Zaza: Yeah. Anil Seth: people in the business as well as those people on the outside of it. It is changing things. It can be used to generate enormous amounts of misinformation, disinformation, biased information. These are problems that are already here. And then there are a whole raft of problems. I think the more sort of existential problems about what happens if these systems are given the ability to make decisions, to control things. It is concerning because as we've just been discussing with these biographies, they're kind of amusing, but they get things very wrong. understanding of why they get things wrong, why they sometimes get things right. They're very unreliable. And yeah, in my kind of most of my research, let's say in psychology, neuroscience, if you're going to do an experiment, you've got to go through a pretty rigorous ethics application process before you do your experiment on people. These large language models of vast experiments on human society. And it's not really asking all that much that there should be some regulation of how these things are done because this already happens in so many other areas. And so we're not, I mean, the six month pause, I think that's, for me, that's more just an attention grabbing thing. I don't think anybody seriously expects that to happen. But the point of the letter is not for that specific thing to be the case. But for me anyway, it's to raise this issue the news agenda and have it discussed not only by scientists and not only by people in tech, but by people in the humanities, by people in sociology, by people in politics, because these are issues for which I'm not sure the scientists and technocrats will have the best answer. I mean, it's something that affects the whole of society and the debate therefore needs to go and the ivory tower. Zaza: All right, leaving a potential emergence of hell 9,000 behind us, I think it's time I stop beating around the bush and just go for it. Professor Seth, what is consciousness? Anil Seth: Well, we've already Zaza: Yup. Anil Seth: established that it's not something that GPT-4 has, even though it might seem like it has it. So what is it? Well, Zaza: All right, Anil Seth: I go Zaza: good beginning. Anil Seth: right back to the definition of Thomas Nagel. He's a philosopher of mind, a very renowned philosopher of mind. And in a famous paper from decades ago now, he put it like this. He said, for a conscious organism, there is something it is like to be that organism. It feels like something to be me. It feels like something to be you. It doesn't feel like anything to be GPT-4. It doesn't feel like anything to be a table or a chair or an iPhone. Really, consciousness is any kind of experience whatsoever. When you open your eyes in the morning, it's not just that your brain is processing information that signals are coming in and eventually you get up. happening too. There's visual qualia as the philosophers say, you know, there's the colors that enter your mind, experiences of emotion. We all know what it is. Again, it's what we lose when we fall into a dreamless sleep or go into anesthesia and comes back when we come around again. Without consciousness, we are merely complicated objects. So I think that's another makes us different from being just a very complicated object. Zaza: In being you, you're right, you could even say that we're hallucinating all the Anil Seth: Mm-hmm. Zaza: time. It's just that when we agree about our hallucinations, that's what we call reality. Could you explain this a little bit more? What did you mean by that? Anil Seth: I would try. So what I mean by that is it's an exploration of what I think is the central aspect of consciousness for us, which is perceptual experience. So let's go back to that scenario again. You open your eyes in the morning and it seems as though you experience the world just as it is. Let's stick to vision. Zaza: Yeah. Anil Seth: You open your eyes and there's a world full of objects and people and places and they have colors and shapes and they might be moving and doing all sorts of things and it seems as though that world is just there and we receive it. It just pours itself transparently into our minds and the self might be perched within the skull doing the perceiving now. This might be how things seem, but it is not how things actually are and a very long tradition of I mean, it goes right back to Plato, but there are many more recent landmarks as well. Suggest that perception isn't a passive readout of an external world. It's always and everywhere an active construction. So imagine being your brain, you know, your brain is stuck inside this bony vault of a skull and it's trying to figure out what's out there in the world. are noisy and ambiguous sensory signals, which are only indirectly related to things out there. So in order to make sense of these sensory signals, the brain has to make use of its prior expectations or beliefs about what's going on in the world. And use those to make to interpret these unlabeled electrical signals that are coming into the brain. And what this means is that perception, this process of figuring out what's there, is best thought of as a kind of perceptual best guess. The brain is trying to figure out the most likely cause of its sensory inputs. And under the hood, the way it's doing this, or the way I think, and many others think it's doing this, is that it's continually throwing out predictions from a sort of inside out or top down direction, predictions about the sensory signals. And then the sensory information coming into the brain is just reporting the error, the discrepancy between what the brain predicts and what it actually gets from the world. And by trying to minimize this prediction error all the time, the brain can settle on a best guess of what's there. And what I think is happening is that is what we perceive. is not just a transparent representation of the world. It's the brain's top-down, inside-out, best guess of the world, a kind of controlled hallucination. So this is where this word comes from. Now, it's a controlled hallucination because it comes from within. And we typically think of hallucinations as false perceptions that come from within. But essentially, it's a controlled hallucination. So one that is tied to the world this process of prediction error minimization in a way that makes it very useful for us. So, yes, our perceptual experience is kind of individual and unique. It comes from within, but we all have largely similar brains. So, you know, if you and I are looking at a red car, we'll both say, yeah, I see a red car. And that kind of social consensus, I think that like we are all experiencing the world objectively even though we're not. So that's the sense in which I mean that when we agree about our hallucinations, that's what we call reality. Reality of course Zaza: Mm. Anil Seth: itself as people from Emmanuel Kant onwards have said, we can never experience that directly. It's always hidden behind a sensory veil. We will only ever experience it indirectly. Zaza: So there's billions of brains out there and yet somehow we're still all having kind of matching shared hallucinations. We're very different people. How come we don't have very different hallucinations then? I'm kind of stuck Anil Seth: Actually, Zaza: here Anil Seth: yeah, Zaza: a little bit, I suppose. Anil Seth: well, I think we kind of do. So there's two ways to look at this. You can either focus on the shared aspect. And I think that that's important. We don't experience the world in radically different ways. And there's good reasons for that. I mean, our brains are tuned by evolution to keep us alive, very broadly put. And there are things therefore that all human brains will have in common. all from the same species. So there's going to be probably more similarity than difference in how we experience the world. I mean, we all have in our brains assumptions like how light behaves, whether this light usually comes from above and how light reflects off surfaces. So we all see colors. So lots of things will be the same. Emotions, we will experience emotions in roughly, in similar ways because we all have bodies that have basically the same components. side of the coin, but I actually think the other side of the coin is more interesting because It's not that we need to have identical experiences, and I don't think we do. I think we all live in subjectively unique worlds. Sometimes the differences can be quite large. I mean, this is typically when we will say somebody is hallucinating. Or we might use the term Zaza: Yeah. Anil Seth: neurodiversity, which appeals to the idea that there are some people who might have quite different experiences of things, things like that. But I think there's a hidden landscape of what I've been calling perceptual diversity where we all differ but we might not recognize these differences precisely because we use the same words if we speak the same language and also because our perceptual experience has the character of being real. You know if I look at outside and see a red car it just seems to me that see it differently and the fact that they might is kind of hidden and sometimes it can surface it. Do you remember a few years ago there was that photo of a dress that half the world saw as blue and black and half the world saw as white Zaza: Right, Anil Seth: and gold? Zaza: yeah. Yeah, I do. I do. I remember Anil Seth: And which Zaza: that. Anil Seth: way Zaza: I Anil Seth: did Zaza: think Anil Seth: you Zaza: I saw Anil Seth: see Zaza: it Anil Seth: it? Zaza: as blue. Anil Seth: Well, Zaza: I think I saw it as blue, dark blue, Anil Seth: good. Zaza: yeah. Anil Seth: Well, then, you know, Zaza: And Anil Seth: you Zaza: I Anil Seth: and Zaza: couldn't Anil Seth: I are Zaza: understand Anil Seth: similar like Zaza: why Anil Seth: that. Zaza: people aren't seeing it the way that I'm seeing it. I was like, what's wrong with you? Anil Seth: Exactly, exactly. That's Zaza: It's obviously Anil Seth: the thing. So Zaza: blue. Anil Seth: that was one example where it's revealed. I see the interesting thing for me about that was not just that people saw it differently. It was how hard it was for people to understand that other people could have a different experience. Zaza: Tiedän, että ihmiset on liian sanoneet minua, jotenkin täysin. Anil Seth: And yeah, that's the thing. But the thing is it's not just the dress, right? I mean, it's everywhere and all the time. We're all seeing things slightly differently, but the difference doesn't usually this into language in that way. So over the last year or so, a couple of years, I've been involved with this project called the Perception Census. And this is actually an attempt to try and measure, map out this hidden landscape of perceptual diversity, because we really don't know much about it. Psychologists tend to study maybe one or two different aspects of perception. But what we're trying to do with the Perception Census at lots of different aspects of perception in all together, like color, sound, music, time, emotion, expectations, all sorts of things, and figure out how we all differ in different countries and different ages, different cultural backgrounds. And so this is actually something that's still going on. We're still collecting data. If anybody listening to this conversation would like to take part, it would be fantastic. That would really help. It's all you need is your own computer. It's a series of little online experiments and illusions and interactive games that we've carefully developed to try and tease apart how we each experience things differently. And by taking part, you learn a little bit about perception. You learn about your own distinctive way of perceiving things. you get this this warm glow of contributing to to science. Zaza: All right, Anil Seth: So Zaza: we'll put the link on YouTube under the video. Anil Seth: Fantastic. Zaza: Yeah, absolutely. Anil Seth: Yeah, we've already had 20, I think 25,000 people already. So it's gonna be a big study. So yeah, if people take part, we're hoping to make this a real landmark study and it's fun to do. So yes, looking forward to that. Zaza: OK, very cool. You touched on it already. We have this weird looking kind of wet wired lump of neurons sitting in the darkness of our skull, as you said, receiving electrical impulses from our sensors, such as retinas or our skin, et cetera. So there is no direct contact with the outside world, really. I think you call it that the reality is always behind a sensory veil. Reading your book though, I kept wondering what would it feel like to actually have a direct window into the world out there. Is objective reality knowable to us at all with the brains that we have? Anil Seth: I think as Zaza: Or is Anil Seth: going Zaza: that Anil Seth: back to Immanuel Kant, Zaza: an insane Anil Seth: I don't think Zaza: question? Anil Seth: it makes sense. Yeah, I don't think it makes sense. It's interesting to Zaza: Yeah. Anil Seth: wonder why it doesn't make sense because the phenomenology, which is to say the way experience seems to us to be, is that it's direct, right? We seem to have this direct experiential access to the world. But Zaza: It Anil Seth: I don't Zaza: seems Anil Seth: even Zaza: to me Anil Seth: think that's Zaza: like you talked about waking up in the morning and Anil Seth: Yeah. Zaza: opening your eyes. They seem like a window into the world Anil Seth: Yeah. Zaza: out there. Straight Anil Seth: And there are two Zaza: up. Anil Seth: things to say about this. There's like, the first is what would it even mean for that to be possible? Like, I don't think it even means anything for that to be possible. What is out there really, or in here in the brain and the mind, cause the brain is part of reality too. It's not something apart from it. Zaza: Yeah. Anil Seth: And here it depends on, well, you know, ask whatever, ask your preferred physicist. It's, is it strings? Is it quarks? Is it a bubble of quantum foam? You know, what is going on there? What would it mean to experience that? It's really not clear. Even take something as straightforward as color and light. We've known since Newton how color works in terms of electromagnetic radiation. Electromagnetic radiation is not colored. It's just radiation. So what would it mean to directly experience just electromagnetic radiation? It's really not clear. I don't think experience is the kind of thing can directly reflect reality. It is always, I think, by definition, a construction and interpretation. So that's one thing. The flip side of that, I think, is equally interesting and also not often discussed. So this sense that we have of direct experiential access to the world, that itself is a is part of perception, that it's a property of perception we seem to experience the world as it is. That too is a kind of perception. And the reason I like to think of it that way is because there are examples where that's not the case. So we have things like if you look at the sun and then look away, you can have a visual after image and that's a visual experience, but it doesn't feel like it's part of the world. There are some people who have synesthesia where they might see colors text. That's fascinating. What's often not emphasized there is that they never experience the colors as being part of the world. They don't experience them as being real. Zaza: so they know that this is a glitch. Anil Seth: They know it's a kind of glitch and well, or a feature, it's also, it could be a good thing. I Zaza: A Anil Seth: mean, Zaza: feature, Anil Seth: for many people it's a good thing, but Zaza: of course. Anil Seth: it's not experienced as part of the, where there's different. And then there are people with condition called depersonalization, derealization. It's a very difficult to diagnose condition. But the, one of the overriding features in this condition is the experience that one's experience is not quite real. people with depersonalization, derealization have this kind of as-if-ness, looking at the world as if through a pane of transparent glass, as if watching themselves on TV. There's something that takes away this feeling that most of us have most of the time, that we're directly, experientially connected to the world. And the fact is, that itself seems to be, not something we can take for granted. And we should, we certainly shouldn't take it as evidence that there actually is a direct connection for the reasons we've already discussed that that wouldn't even make sense. So yeah, it's a really, it's a really interesting area. I think it's gonna become more important, not only for these clinical reasons about understanding and helping people with depersonalization, but as technologies like virtual reality and augmented reality develop, be playing more in this this weird liminal zone where experiences may seem real in a way they haven't done before in media. Like in a cinema you can be very immersed but there's always a level at which you know that you're in a cinema and not actually, you know, wherever the film is being set. These borders are going to be eroded. So understanding how it is that reality I think is going to be a new frontier in consciousness research Zaza: Speaking of cinema, I just saw The Matrix recently. I know this is a complete digression, but what's a deja vu? Do you have any thoughts on that? Anil Seth: Hehehehe Zaza: Because sometimes I have really strong deja vus and I'm like, what's... In The Matrix, they say this is a glitch in The Matrix. Anil Seth: Yeah, I don't think it's a glitch. I mean, I love the matrix, such a good film, but I don't think Zaza: Yeah, Anil Seth: it is a, Zaza: same. Anil Seth: is a glitch in the matrix, but I, I think all these Zaza: Right, Anil Seth: experiences Zaza: settled. Anil Seth: that I don't, I think they are fascinating. I won't go on for very long on this because I actually don't really have a particularly compelling story about it. But with all experiences like this, whether it's deja vu or out of body experience or things like this, there are two ways they can be interpreted. One is literally like, okay. actually happened before and I'm having a kind of memory echo of it now. Or for an out of body experience, people can take it as evidence that yes, my consciousness can leave my body and go moving around in a way that suggests that it is in fact immaterial, doesn't depend on the brain and the body. That's one way to interpret it. I don't think that's the right way. experiences are always kinds of constructions and we often take for granted about them things that we shouldn't. So in the case of an out of body experience, just because almost all of the time we tend to experience the world from a first person perspective, co-located with our head, that doesn't mean that is always necessarily the case. An out of body experience for me reveals that the first person perspective is something else that the brain is constructing, like the sense of reality that we were just discussing. It's not something that can be taken for granted and sometimes that process can go a bit awry and we have an out of body experience. So a similar thing with Deja Vu. It's not that this situation actually did happen before, but the sense of familiarity that the brain generates and it is not always inevitably tied to whether a situation has actually happened before or not. So I think these experiences are always very useful because they reveal aspects of our experience that otherwise we would not even realize require explanation. Zaza: Okay, so what's the difference then between an actual hallucination as if you have let's say high fever and are seeing things that are not there and a controlled hallucination, a perceptual best guessing, right, which gives rise to our consciousness. Anil Seth: So just Zaza: or is Anil Seth: as Zaza: our consciousness? Anil Seth: normal perception can be thought of as a kind of controlled hallucination, where the brain's best guesses are usefully tied to the world by this process of updating on the basis of sensory information, hallucination as we might typically understand it, this happens in, let's say, schizophrenia or in psychedelics or some other conditions, can controlled perception. So the brain's best guesses in this case are not, you know, they lose their grip on their causes in the world. So it's not a difference of kind. I think under the hood, the same process is going on. The brain is making predictions and these predictions are interacting with sensory information. So everything comes from the inside out in a way that's calibrated coming from the outside in, but the balance is shifted. So in, as our brain's predictions become less. less adequately, less accurately, I guess, I don't know. If they, if they, if this process by which they're updated by sensory signals becomes less constrained, then our perceptual best guesses can start to drift and that's when we will fall into the domain of hallucinations as they're typically understood. Zaza: Is there something, is it similar? I mean, would be like taking psychedelics, we can experience a vastly different reality, at least that's what my friends tell me. And people usually describe it as a more profound one as well. A lot of the times. What is that? Is it similar then to an actual hallucination? What's happening to our consciousness under psilocybin or DMT? Do we tied down to the world out there. Anil Seth: Yeah, I think so. I think Zaza: Or is... Anil Seth: one of the interesting Zaza: Yeah. Anil Seth: things is understanding the differences. So actually this is a project that I'm working on with my group and colleagues, Kesake Suzuki and David Schwartzman here is trying to understand the differences between different kinds of hallucinations. So, you know, an experience under DMT is different from under psilocybin. And of course it's very different from visual hallucinations that people might disease or visual loss. But they're all hallucinations, broadly speaking, because they're all perceptions that differ from the normal or the range of normal perception. So understanding how and why these things differ, I think is really potentially useful. But just sticking with psychedelics for a moment, I think, because they also illustrate one of the themes that that we've been talking a psychedelic experience. Some people will have psychedelics and they'll say, the filters have come off and I see the world more as it is. I have more direct access to underlying Zaza: Yeah. Anil Seth: reality. That I think is, if it's helpful for that person to think of it that way, okay, fine. But I don't think it's a scientifically justifiable way to think about it. I think it's rather that What psychedelics can do is change the way your brain is dealing with sensory information so that you can realize that in fact, all of our experience is a kind of brain-based construction. Now, the kinds of things that happen under psychedelics are precisely the kinds of things, in my view, that you would expect to happen if all our experience was generated and dependent interactions inside inside a skull. So from that perspective, it doesn't give us more direct access to anything. It just underlines the indirectness of all our experience and that too can be an incredibly helpful thing and I think perhaps a more helpful thing because that's the kind of recognition that also can accompany people who have engaged in meditation for many years and meditation can give you insight into into the constructed nature of perceptions of the world and of the self. And we've, you know, people who've gazed in this kind of, I've, I've meditated, but not very diligently. Now it does open up this space between how things seem and how they are. That can be very valuable for people. So that's how I, yeah, that's how I think, but exactly how they work in the break. This is another thing that I think another really exciting area of, of research. collaborating with people to try and unpack how psychedelics really do change the brain. What is going on in a psychedelic state that explains why our perceptual experiences of the world and of the self are as different as they are. I guess one of the reassuring things is because the change in experience is so dramatic, it would be a bit worrying if we didn't see much change in the brain, but there are lots of things. obvious and dramatic as one might expect given how enormous the changes in subjective experience are, but quite what they are is still under, it's still active area of research. Zaza: In the book you write that the self is a bundle of perceptions. Why is it then that we feel we absolutely know in our day to day life that we're having a unified experience of the self, of I? For example, I don't think I remember ever feeling split in different perceptual me's that will make me question a unified Zaza, even under extreme stress, for example. Anil Seth: Well, I'm happy for you. I mean, maybe this is still an interesting experience to come. I've certainly had experiences of the disunity of self in various situations. The term bundle comes from David Hume, the Scottish philosopher. I think he's the first to oppose this idea that the self is not a unified essence, the thing that does the perceiving that might survive after death and so on, but rather collection of related perceptions. There are experiences of the body, there are experiences of free will, experiences of a first person perspective, of experiences of personal identity. There are many different ways in which we experience being who we are. And in the book, one of the central ideas is to unpack each of these aspects of selfhood perceptual best guess. And the fact that they can be is, you know, one clue about this is there are always conditions where one aspect of self can go away and others can remain. Whether this is in the lab or in psychiatry or neurology clinic, if people lose their memory, so they can no longer lay down We can have out of body experiences yet still experience free will. We can experience our body in different ways yet still know who we are. So these aspects of self actually turn out to be in practice a kind of pick and mix version of self that come together in particular ways for most of us most of the time, but which can come apart and the experience that they have to be unified. inevitably of a single piece. That I think is again another aspect of perception. I think evolution has tuned our self-experience that way because the experience of self basically it's what is it for? Well, it provides an anchor by which we can organize our behavior. The self is this kind of experience of ongoing relevance and salience. that to which we can relate to the other things that we experience in the world so that we do the right thing for the organism. And so it would not be a smart evolutionary solution if we experience the self as always changing and as always somewhat fragmentary. So I think there's a good reason to experience yourself as unified, but that doesn't mean it always is unified. Zaza: Right. That makes me think of that horrible tragic story that you relay in the book about that gentleman. I think he was an editor at BBC3 who lost Anil Seth: Mm-hmm. Zaza: the ability to form new memories or he lost memories, like short-term Anil Seth: No, no. Zaza: memory capacity or was it long-term memory capacity. And then Anil Seth: You Zaza: he Anil Seth: had Zaza: wrote, Anil Seth: it right. Zaza: I woke up, right? Anil Seth: Yeah. Zaza: And then he Anil Seth: Yeah, this guy called Clive Waring, he lost exactly Zaza: Exactly. Anil Seth: this ability to lay down new memories. So his long term Zaza: You right. Anil Seth: memory was intact in the sense that he had good recollection of things that happened before his illness. So he had this brain infection which wipes out the part of his brain needed to lay down new memories. So before that, he could remember what happened before that. But then after that event, which was now, it's 40 years ago now, I believe he's still alive. ability to lay down new memories, certainly for the first 10 years or so, probably more was was almost entirely abolished. His wife used to describe it that he lived in this permanent present tense and his diaries. If you read them, they are extraordinarily moving because he will fill page after page with phrases. I am awake for the first time cross it out. And then below that ignore that now I'm awake for the very first time crossed Now I'm really properly awake crossed out and so on and so on and so on, which just suggests very strongly that his lived experience is of always just waking up and wondering where he is and what's going on. So yes, there's a deeply tragic aspect to it, but it wasn't all tragedy. never forgot the fact that he loved his wife. And so Zaza: Hmm. Anil Seth: they retained this strong bond, even though he couldn't remember their meetings after his illness. And also he was Zaza: piano Anil Seth: heavily Zaza: player, Anil Seth: involved Zaza: right? Anil Seth: in music. I think he was, he may have been, but he was also, he directed choirs. So Zaza: Right. Anil Seth: he was a choral director. And one time his wife took him to his old choir and he was able to lead them in music. And he did this perfectly fluently. and seemed to be totally alive, totally within the moment. Of course, the music was able to take him from moment to moment in a way that in normal life his brain could no longer do. you Zaza: All right, I just have a couple of more questions left. Number one, if consciousness is indeed a controlled hallucination, what's the best way to live our lives in light of this knowledge? Anil Seth: Well, this is a very, very good question. And I'm not, I'm, I certainly don't feel qualified to tell people how to live their lives, but, Zaza: And yet Anil Seth: um, Zaza: I asked you anyway. Anil Seth: and yet I'm going to tell you anyway, so for me, Zaza: Wonderful. Anil Seth: one, one thing, one way it's made a difference is, um, is it just brings a different perspective. So. It's, yeah, of course, when we, when we understand that perception is a kind of construction, it doesn't make us stop seeing colors and hearing music and experiencing the self as unified. You know, how things seem still remains the same way, but it's also changed. There's an old story about Ludwig Wittgenstein and his biographer, and having an argument about the fact that it seems like the sun goes around the earth. Of course, at one point people thought the sun goes around the earth and then Copernicus comes along and shows us no, the earth goes around the sun and the earth rotates on its axis, which is why we have night and day. And when we know this, how do things seem? Well, it still Eastern sets in the West, but it also seems different. Our knowledge of what's going on makes it possible to have two perspectives at once, in a sense. We can lie on our backs on a starry night and understand that it's not just that the sun is going around the earth, but that we're this little speck in a vast abyss. And I think rather the same is true about perception. We can go about our daily lives emotions and highs and lows and frustrations and opportunities and colors and music as I always did. But running alongside that is this recognition that this is all unfolding within my head. It's all a construction. It's not arbitrary. It's related to the world. And for me, that adds a sense of to see myself as part of nature and not apart from it. I can recognize that other people might have different experiences even when we share the same objective reality. And I think that also can be helpful in building new platforms for understanding and empathy and communication. And that's just underlying at all. Yeah, there's just this sense of almost gratitude for this ability that we have to have experience in the first place. Being conscious is a kind of everyday miracle. We should not take it for granted. Doesn't mean that every moment is a beautiful moment, not at all, but that's the texture of life. Zaza: You're right about that. Anil Seth: So Zaza: Yep. Anil Seth: I think people can do with this what they will, but I find it maybe kind of, I think it brings to life something a little bit like meditation might bring as well. It brings this appreciation of what's going on in the flow of experience. It opens up this little space between how things seem and how things are and we can work with that space however we think best. Zaza: OK, you kind of answered the next question already just now. But if we are a little bit more practical, my absolutely favorite comment under your TED talk, I believe it has almost 14,000 likes, reads like this. Ah, yes, watching this during an existential crisis in the middle of the night was a great idea. Besides being extremely amusing, this statement does ring true in a way to what you just said because your research continuously our knowledge about the thing that is kind of most intimate to us, right? Our consciousness. You basically keep finding out that everything that our intuition tells us about what our minds are and about who we are is wrong. So I guess your knowledge and you said that pretty much right now is in direct contrast to your everyday experience. So I'm wondering, Professor Seth, how are you Is it meditation? Anil Seth: We did talk about it. Zaza: Indeed? Anil Seth: No, Zaza: Or is it something else? Anil Seth: well, maybe going for nice grounding walks by the beach here in Brighton. But, but, but no, I, I, I think we did talk about it really. It's, it's not that it changes my everyday lived experience so fundamentally. I mean, we, I still see the same colors, have the same emotions, get hungry, get tired, enjoy wine. the same, there's just a different layer. Sometimes that layer is useful. Sometimes it may not be useful. I can well understand just as it's a bad idea to start intense meditation when you're going through a very challenging mental health crisis. Let's say if you're having a crisis, this may not be the best time to take on board the indirectness of our perceptual experience. Usually that's something better to have, you know, already installed. I think it's kind of it has prophylactic value, right? If you already have some, you know, a sort of, if you've habituated to thinking that way, then I do think it helps you deal with difficult psychological episodes. But I don't think it's a, it doesn't provide, it's not like a intellectual pill that you can pop to make yourself feel better when things are hard in the moment. I don't know. I go for long walks in the hills and I just continue doing my thing. It doesn't Zaza: Thanks for watching! Anil Seth: disrupt me that much, at least not as far as I would know, but how would I know? I don't have the other me to compare to that did something else. Zaza: Right, okay, the title of this podcast is You Are Trash, so I have to ask you something a bit trashy at the end. Anil Seth: Hehehe Zaza: What is the wackiest theory of consciousness that you ever heard of, but maybe thought contains a grain of truth or has something compelling in it? Anil Seth: Okay, so the thing that comes to mind is a theory or a class of theory has been around for a long time, ever since I first started learning about this stuff in the early 90s as an undergraduate student. The idea that consciousness might have something to do with quantum mechanics. This is one of the most, it's an idea that's never quite gone away. I think most of it has very little basis at all, it's almost like trading on this idea that consciousness is mysterious and quantum mechanics is mysterious, so therefore they must be related, right, which is a very bad reason to motivate a theory. But I do think that we understand not enough about the nature of the material world, because a lot of debate in the philosophy of mind and in consciousness the plausibility of materialism in general, like how is it possible to derive an explanation of consciousness from neurons and neurotransmitters and chemistry and the stuff that we have inside our brains and bodies. We are able to do this for life. Can we do it for consciousness? I am a pragmatic materialist that I think I don't know if we actually will ever be able a lot about consciousness, even if we don't solve the mystery entirely. But it would be nice to solve the mystery entirely. So one possibility is that we will, and that the apparent sense of mystery that still, um, inheres to consciousness will just dissolve in the same way that the mystery of life dissolved. I think that's pretty likely, but we're not there yet. So we can't say for sure. and what we typically think of as materialism beyond neurons and chemicals and so on. And to reach deeper into the nature of matter. And quantum mechanics is our best grip on the nature of matter. There's a theoretical physicist called Carlo Rovelli, who I think is a brilliant writer. Well, he's a brilliant physicist and also a wonderful writer about physics. on quantum mechanics called the relational view, relational interpretation of quantum mechanics, where the idea is, what's really going on in this very mysterious physical realm, the mode of existence, the fundamental mode of existence is relational. Things do not exist as objects with properties, but only relations between things exist. because it relates to some theories about consciousness that stress that information might need to exist in some way. And information is a kind of relational property. So there might be some sense in which 10, 20 years down the line, maybe 50 years, of consciousness that exists now. But things that basically look like materialist theories, but instead of saying it's neural synchrony between different brain regions or whatever, you know, they're phrased at a level of description that is a bit deeper and a bit, you know, that leverages some things that are specifically quantum. I don't know if we'll ever need that. I don't know what in principle it would explain, but that's, possibility. And I think it's an interesting possibility. I think it's always unwise to restrict the possible scope of explanations and materialism has a lot of resources precisely because matter is extremely complicated and we don't understand everything about matter yet, let alone mind. Zaza: We'll follow this quantum physics stuff. I'm severely malnourished in this department. So I think after this conversation, I'll go back to Chad GPT and ask him to help, Anil Seth: Hehehehe Zaza: ask it to help me out a little bit maybe. Professor Seth, thank you so much. This was really illuminating. I enjoyed every second of it. Where can people buy your book, Being You? I think it's available everywhere pretty much. Anil Seth: I hope it would be available in most places, certainly available online and hopefully in all good bookshops as the phrase goes. But yes, you can find it on my webpage as well, annilseth.com and that will give you links both to the book and to the perception census if you want to take part in that. Zaza: Yes. Do you maybe have any social media accounts that people can follow? Anil Seth: Sure, yeah, so I'm still, despite Elon Musk, I'm still on Twitter at Anil K Seth with a K in the middle and on Instagram at Prof Anil Seth. Zaza: All right, again, thank you so much. This was wonderful. Anil Seth: Thank you, Zezza. It's been a real pleasure for me too. Thanks so much.