 For Web_00017.JPG?h=c3919415&itok=jjfFhn5N)
What is consciousness?
Professor Brian Cox and an expert panel explore the elusive science of consciousness – what it is, how it arises, whether it can be observed in the brain, and the most compelling theories explaining it.
In a deep, often mind-bending discussion, the panel tackles the many ways of interpreting consciousness itself, and how fields like psychophysics can be used to quantify imagination and perception.
With questions from the audience, the conversation delves into hallucinations and reality, how memory helps to build identity, and whether AI could ever be conscious too.
Listen and subscribe on podcast platforms
Brian Cox: Hello, I'm Professor Brian Cox. Welcome to A Question of Science recorded here at the Francis Crick Institute in London. This is the podcast where a panel of experts tackle your questions on some of the biggest scientific challenges facing society today, where we'll be asking the world's top scientists your questions about issues that are central to all our lives. Can we cure cancer? Can we live longer? Should we fear AI?
This episode for me is about a subject that I think is one of the most interesting topics in all of science, consciousness. Despite great advances in neuroscience, psychology, and computational modelling, we are nowhere near, I think it's fair to say, having a theory that explains how our experience emerges from physical processes in the brain. So today, we're going to fix that. Are we going to- Actually looking at the panel. As we will be asking, what is consciousness? What processes give rise to our subjective experiences? Where might consciousness be located? How does it arise? And what are some of the most convincing explanations for it?
To answer those questions, I'm joined by four experts involved in the latest discoveries and theories of consciousness. And they are.
Anil Seth: I am Anil Seth. I'm from the University of Sussex, and my research group, we have mathematicians, and psychologists, and neuroscientists, and computer scientists all trying to bash our heads against this question of what is consciousness, and how does it happen for many years now,
Katharina Schmack: I'm Katharina Schmack. I'm a group leader here at The Francis Crick Institute. We study consciousness through the lens of psychosis, and in particular, we are interested in how the brain and the body give rise to hallucinations.
Steve Fleming: I'm Steve Fleming. I am a Professor of Cognitive Neuroscience at University College London, just down the road from here. I run a brain imaging lab studying metacognition and how we think about our own thinking, and I'm also involved in developing cognitive neuroscience theories of consciousness.
Alex O'Connor: I'm Alex O'Connor. I'm not a scientist. I am interested though, in how scientific findings can inform philosophical theorising. I'm the host of the Within Reason podcast, which is a philosophy podcast, but I'm always speaking with scientists and attempting a kind of cross-disciplinary approach. And this is our panel. Now, we thought we'd begin with a definition, but of course, those definitions might be hard to pin down and take a long time. So maybe the shortest definition each of the panel members has of consciousness. What do we mean by consciousness?
Anil: I go back to a philosopher, Thomas Nagel, 50 years ago, who gave, I think, a beautifully minimal definition of consciousness. Says for a conscious organism, there is something it is like to be that organism. It feels like something to be me, feels like something to be you. But it doesn't feel like anything, we think, to be a table or a chair.
Katharina: For me, consciousness is simply the state of being aware and having subjective experiences of the world.
Brian: And Steve?
Steve: I agree with that, and I think it's also important to contrast what we mean by being conscious of something against all the other kinds of brain processes that we think are going on unconsciously.
Brian: Alex?
Alex: I agree with Thomas Nagel's definition, with an emphasis on the subjectivity, which is why I think it's such a difficult question for science, which is in the domain of objectivity.
Brian: So complete agreement, thank you for listening. Thank you. So those are wonderfully brief definitions. So now I think we will go to the first audience question.
Debbie Collins: Hello, I'm Debbie Collins, and I would like to know, can we see consciousness in the brain?
Brian: So who would I take that, Anil?
Anil: I think we can see the fingerprints, the footprints if you like, of consciousness in the brain. We can, when we look inside a brain, we always see a representation of this deep complexity that lies within our skull. And the things that we have most immediate access to are things like what brain regions do, how much energy they're consuming, what neurons are doing if we look closely enough. None of these things are consciousness itself, but they're intimately related to consciousness
Consciousness science kind of got going again in about the 1990s through the study of the neural correlates of consciousness. The idea here was to put to one side the deep philosophical problems of what consciousness is, and how it might depend on physical stuff. And simply take advantage of the fact that in the brain, there are strong correlations between what happens in the brain and what happens in conscious experience. And if we use brain imaging of the sorts of things that Steve does and Katharina does, we can look for these footprints of consciousness in the brain, but not consciousness itself.
Brian: Katharina.
Katharina: Yeah, I agree with that. And I think if we stick to that definition of consciousness being something to do with a subjective experience, we definitely can see subjective experience in the brain, or we can see that the brain behaves differently depending on what the subjective experience is. So the brain does not simply react to external stimuli always in a fully predictable way, but sometimes the same- sometimes we see something and we might perceive it, and sometimes we see the same thing and we just might not notice it. And the brain definitely distinguishes between these two cases.
So again, without having a clear, without having to agree on a definition of consciousness, I do think that we can see traces of subjective experiences that form consciousness.
Brian: Steve.
Steve: There are really exciting experiments being done in a branch of psychology known as psychophysics, which is this set of tools we have to quantify how people experience simple stimuli. And there's a lot of mileage you can get out of those experiments. And we can now quantify things like imagination, things like how confident you feel when you see a particular stimulus, and how that is related to patterns of brain activity. But I think what we're learning from that science is that it's not the case that consciousness can be localised to a particular brain area or a particular type of neural activity. Instead it's a process rather than a thing. It's something that the whole system does. And we're starting to understand the kinds of brain processes that go along with being conscious.
Brian: And I think just, Katharina, something that you said to follow on from that. So you can have a particular stimulus, but the brain- the pattern you see will be different depending on whether you report it as having been noticed or not. So it's not the stimulus itself. There's something else about how you categorise that
Katharina: Yeah.
Brian: input.
Katharina: And I think hallucinations are a good example of that. We, in psychosis, for example, which is the topic I'm interested in, people often hear voices that others cannot hear. And actually, that's something that happens to all of us, maybe not hearing voices, but having perceptions that, yeah, are not directly triggered by an external stimulus. One example is if we are waiting for a message or a call, and we have our phone in our pocket, we will start, I at least start feeling it vibrate, and I look and it actually has not vibrated. And the reason is because our expectations influence what we experience in the same way as do external stimuli. And just having a very strong expectation of getting a phone call will give us the perception of really feeling the phone vibrate. And that's a little bit how we think about hallucinations, but I think about perceptions more broadly.
And so we can really, in the, like when we look in the brain, and we have, as Steve was mentioning, we have amazing tools now at our hands. We can really record in the brain, especially in model organisms, we now can record from thousands of neurons at the same time, which is advances that have been made possible just like in the past, I would say, five to 10 years. And we can start to see that the state of the brain differs depending on whether we subjectively feel something vibrating, or we don't feel it vibrating regardless of whether the phone has vibrated or not. So I think it's a really exciting time to start looking at the kind of correlates at least of the subjective experience.
Steve: We've been doing these experiments on how you imagine the world. This is work that I've been doing with a colleague, Nadine Dijkstra at UCL. And we have been asking people to imagine simple patterns on a computer screen. And then sometimes we sneakily fade in the real thing that they're being asked to imagine. And we find that people often confuse the real stimulus with their imagination and vice versa. And sometimes people are able to tell the difference, and this is what we call reality monitoring. They're able to monitor whether they're imagining or whether they're perceiving reality. And we think the way that works is that areas in the prefrontal cortex of the brain are essentially monitoring the strength or reliability of this sensory activity, and trying to figure out, does it really reflect reality or is it just being internally generated?
Brian: Yeah, so Alex, the question was, can we see consciousness in the brain?
Alex: I think the answer is no. And I think that that's essentially what we've just been getting at with Anil's language of putting it to the side and talking about correlates. I mean we've heard words like correlated to and related to. And I think that the, what the scientific method does in analysing the brain is expertly tells us some really fascinating things about the brain activities that are correlated with certain kinds of experiences. Isn't it interesting that there might be a difference in what the brain is doing when someone experiences a voice that's a hallucination, versus a voice that actually comes in through their ears? That's really interesting.
But we all agreed on the definition of consciousness as being what it is like to be a particular thing. And that's not something you can see in the brain. You can take someone who's hearing those voices, and you can study their brain to the maximal capacity that in principle it would be possible to do with any scientific instrument. And you might know all there is to know about what their brain is doing, but you won't know what it's like to hear that voice.
And if that's what we agree that consciousness itself actually is, then the conversation that we have around these investigations into the brain is not an investigation into consciousness, but an investigation into the brain activity which is correlated with consciousness, which is related, but not the same thing.
Steve: Not sure I fully agree with that, Alex.
Alex: Good, it's about time.
Steve: In fact, I'm sure I don't fully agree with that. I think it's a good starting point, right? I think often scientific investigations of things that seem mysterious start off with correlations. You know, correlations between astronomical measurements and things like that gave us clues about the organisation of the solar system. And indeed, now we can see these correlations between things happening in brains and things happening in the conscious experience of a person.
But they don't have to stop there. I think the challenge for us now, and the reason there's a, I find it a particularly exciting time in consciousness research is that built on top of these correlations, we now have theories, and theories try to explain why we see patterns of correlations. They try to propose mechanisms that underlie properties of consciousness. So they might not tell us why consciousness exists in the first place, why it's a property of the universe, but they can start to tell us things, like why is vision the way it is, and different from hearing, or different from an emotion. I think if you do that, then you actually get close to a scientific account of consciousness.
And the fact that you won't get as far as knowing what it's like to be, let's say a bat, or a sheep, or something, I think that's an unfairly high bar. We don't generally expect science to deliver that kind of thing. We want it to explain, predict, and help us control things. And if we can do that, then I think we've done a pretty good job. And I think science of consciousness hasn't done it yet, but it's on the way to do that.
Brian: Right, well, we've actually answered or strayed into many of the questions that we're going- so I think given that, we'll go to the next question please.
Jade Guo: Hello, I'm Jade Guo. My question is how are consciousness and memory connected?
Brian: So consciousness and memory.
Steve: I think they are intimately linked. I think we can start from the kind of intuition, if we had no memory at all that we were just located in the immediate present, it would be very hard to articulate what that was like. We seem to have an experience that is extended in time, and we talk of a stream of consciousness and moving through time.
And I think recent data actually from psychophysical experiments show us that consciousness is surprisingly slow. It seems to integrate over around half a second of time. We can kind of switch stimuli around in order within that half a second, and people don't seem to notice. They integrate them together in their conscious experience. So there's a sense in which consciousness is kind of always pulling along the recent past, and there's been recent suggestions that maybe we should think more about the connections between the neuroscience of consciousness, and the neuroscience of structures supporting memory.
Anil: The connection between consciousness and memory goes way back. William James, who's one of the early people in 19th century writing about consciousness, talks about the specious present, the idea that the present moment is both a little bit anticipating the future, and still has echoes of the past.
My old boss, Gerald Edelman, talked about the remembered present in the same way. And I think this highlights that consciousness is complicated, but memory is also very hard to pin down. It's many different things. We have this very immediate memory. If you close your eyes now, you have this fading visual experience that lasts about half a second. But of course, we remember what we had for breakfast, we remember what we did yesterday. Some of us might remember that. And this is a different kind of memory, much different kind of timescale. And these larger timescales, then you start to see dissociations. So there's famous examples of people who have almost complete amnesia for things that happened to them in their lives.
A guy called Clive Wearing lost the ability to lay down new memories sometime in the 1980s. And he lived in this kind of perceptual present, this perpetual present. And he's clearly conscious, but he doesn't remember anything that happened. So that kind of memory is not necessary for consciousness, but it certainly shapes things like our experience of being who we are. And then the other end of the scale, a topic that Steve and I are both interested in is things like dementia, Alzheimer's disease, often thought about as disorders of memory. People fail to remember things.
But increasingly I think there's a great value in trying to understand them through the lens of consciousness too. How does somebody's experience of the world and of the self change in these conditions? I think that'll help us understand what's going on in a really valuable new light.
Brian: Katharina, so hallucinations, they, it seems to me, you can tell me if I'm right or wrong, they're memory-based in the sense that, I suppose, can you hallucinate something you haven't experienced in your life?
Katharina: I think how we think about hallucinations is that they're, well, there are different kinds of hallucinations, but if we think about hallucinations and psychosis, they are mostly based on previous experiences. Very often it's hearing voices, and someone who has never heard speech, I don't think that they will be able to hear voices.
But there are other kinds of hallucinations, for example, psychedelic hallucinations that often are patterns, geometric patterns that people have not necessarily experienced before. And I think we are still trying to understand the mechanisms, whether they are shared, or whether they're just distinct mechanisms that give rise to these different kinds of hallucinations. So I think it is generally possible to hallucinate something that has no, without having any memory of that particular experience.
But I do think the question of memory is very, very interesting when we also think about kind of what might be the evolutionary advantage of having consciousness. So I'm also a psychiatrist, so I see often that subjective experience, and consciousness can create a lot of suffering. And I mean, pain is similar. There wouldn't be pain if we didn't have consciousness. So I do sometimes wonder, is consciousness just a cruel accident of evolution, or is there something more to it? And I think memory is a good point, because I think what consciousness allows us to do is to connect the past with the future, and emulate what other people or other organisms might do.
If we, for example, are able to know what it is like to feel fear or to feel pain, we might be able to see someone experiencing fear or pain, and we might know what is going on for this person. For example, we sit around a campfire having like a party around the campfire. Suddenly someone jumps out with wide open eyes, and stares into the corner. Because we know what it is like to have fear, we will know that there's probably something dangerous in that corner, maybe a bear, depending on where you are in the world. And so we will know to react to that, and kind of also maybe run away.
And I think that is maybe the evolutionary advantage of having consciousness, the ability to use information from others because we know we can emulate what is going on for them and what has triggered them, and what they will do in the future, so.
Alex: I agree that we should start with reflecting on what it would be like to have no memory. I also instantly thought of Clive, poor Clive. There's a documentary you can watch on YouTube for free. It's extraordinary. Every few seconds or so he says, these are the first human beings I've ever seen in my life. I've never seen a human being before. Or at least not for, you know, 30 years or however long it's been. Interestingly, he does remember some things. He remembers his wife's face, for example, although I think not always her name. And also he was an esteemed composer. And you sit him down at a piano, and he will just start playing. It's amazing. And he won't even remember that he'd been playing five minutes ago, but he can bring it straight to the fore.
So it's obviously very complicated. But I wanna go one step further, and I don't want you to imagine what it's like to not have memory in that case of amnesia where you forget things every seven seconds. I mean no memory at all. I mean nothing being laid down in any instant as time goes by. What might that be like? Imagine you were falling through the air, and in every single instant, there was no such thing as memory. Would you even know that you were falling? You wouldn't have a moment before to compare it to, you probably wouldn't even really be able to feel the wind on your body, because by the time it even sort of got to your brain, you've forgotten about it. There's no such thing as memory. And yet it does still seem to make sense to imagine that someone could be conscious.
But you remove that memory fully, and it seems so rudimentary, so simplistic, so devoid of content that I think this helps to demonstrate that plausibility of a view that consciousness is actually a lot more simple that it's often given credit for. And that we often think of consciousness as being something which is necessarily complex, 'cause it's tied up with the most complex thing we know about: our brains. But that a lot of the things that we see in our complex brain, like memory, like emotional processing, like first person sense of identity, and persistence through time, and all of this kind of stuff is not the definition of consciousness and not what consciousness is, but some of the complex things that consciousness can do.
Steve: I really like your thought experiment, but I reached the exact opposite conclusion. I feel like it's a really nice demonstration that it's really hard to have the intuition of what any experience would be like without this surrounding-
Alex: Right.
Steve: Functionality like memory, like the registering of the passage of time. And so I think this is why it's useful when we approach it through the lens of how a psychologist thinks about the way the mind works, consciousness then becomes more of a process. It's not something we can just point to and isolate in one single slice of time.
Alex: Yeah.
Steve: And so I think that's where maybe our intuitions on, you know, what the fundamental of conscious experience might be, diverge.
Alex: That would depend on your ability to imagine, which is not to say that you should or shouldn't be able to imagine this, but just if you're trying to work this out for yourself and what you think about the world, one place to start might be, is such a person conscious? Is it possible to even make sense of a concept of somebody who's conscious and yet has no memory?
Brian: Well, we will go, one single question could generate two hours of debate.
Anil: Three weeks.
Brian: So with that, we'll go to the next question.
Simon Elman: Hello, I'm Simon Elman. I want to know, do animals and plants have consciousness?
Brian: Well, Katharina, you mentioned that you model organisms. Would that be things like mice? Is that what you mean by that?
Katharina: Because we don't agree on a definition of consciousness, I think we cannot really agree whether they have consciousness or not. But I can speak to subjective experience- to perception. I can speak to perception. And for example, we do see in mice, we can, I think, measure what we think is related to hallucinations in mice. And the way we do this is a very simple experiment. We play tones or don't play tones to mice, and then we train them to tell us whether they hear the tone or not. And then we also train them to tell us how confident they are that they heard the tone.
And what this allows us to do is that sometimes we won't play a tone, but the mice report hearing that tone, and they do so with high confidence. And that is what we call a hallucination-like perception. It's maybe different from a hallucination, but it is an unfounded perception that it's experienced with high confidence. And when we then look in the brain of mice, we can see that perception looks more similar, the neural signals look more similar when the mouse perceive something as compared to when it doesn't perceive anything. So I think, and we can do exactly the same thing in people, and we can relate it to their subjective experiences, we can measure their responses in our experiment, but we can also ask them, did you actually perceive something? And I think this way we can start to approach this question of like, so I don't know, we won't fully understand the subjective experience of the mice, but we can make inferences about what is going on for them and why they behave the way they do.
And what is really amazing when we do these experiments is that with all the amazing tools that we have at our hands, we can, for example, shine light on certain parts of the brain, and that allows us to manipulate these neural circuits. And that allows us to, for example, understand that dopamine, which is a messenger signal, is really driving these perceptions of something non-existent. And we can then take that back into humans, and that can help us to develop new therapies for hallucinations. So I cannot tell you whether they are conscious or not, but I can tell you that it's useful to study them, to understand something about consciousness or experience in humans.
Brian: So the components are there, I suppose you can identify the components of consciousness. But I suppose the question is about how you make that transition from having these different, I don't know, abilities to the conscious experience itself.
Steve: I think that if you approach the topic of consciousness as a process that is associated with a particular set of functions in humans, then we can absolutely go and look for those similar functions in other animals. And there's a really exciting programme of research being done on animal cognition that is throwing up all these surprising findings about the fact that species like the octopus seems to have amazing abilities for deception, and theory of mind, and so on. And that makes us think, well it has the kind of cognitive capacities that we associate with consciousness in humans.
And so I think that you can start to explore that question scientifically. I think the question starts breaking down for systems, or say plants or microorganisms that it's really hard to do any kind of psychological study on. And then I think we're then in the realm of more speculative notions of consciousness.
Anil: I mean it's such an important question, and when we think about consciousness, it's appealing on the one hand, 'cause it's this deep mystery, and we're all interested in ourselves as well. Like why am I me? But it's also the basis for pretty much all of our moral and ethical reasoning. If something is conscious, then it should have some kind of moral status. And so which animals are conscious, how far does consciousness extend within the universe at large is a very urgent, practical, ethical question, to say something about even if you can't give certain answers.
I think what we've heard from Stephen and Katharina is that it's a great, it's a good initial strategy to take what we know about humans and generalise outwards, because we know that, well, I know that I'm conscious, although some philosophers might even try and persuade me that I'm mistaken about that too. Not Alex, other ones.
But if you generalise out slowly, you can get that. And it's easier to generalise to animals that are closer to us than it is to things like plants. My suspicion is that plants are not conscious, and the reason I'm suspicious I don't think they are, is really, they share very little in terms of the kinds of behaviours and functions that are associated with consciousness in animals. And they don't share really what we know to be important for consciousness in those animals. Now that might be a failure of imagination on our part, maybe that consciousness arises in a completely different way.
But in human beings and other animals, consciousness is surprisingly easy to lose. So think we have to be very careful before attributing it too generally to things that don't show obvious signs of it.
Alex: The clearest thing is that I think it shows consciousness is probably some kind of continuum, right? Most people want to say, we know that at least one animal is conscious, you know about yourself. It seems that other animals exhibit certain behaviours, dogs, cats, great apes certainly. But when you get down to a snail, it's sort of a bit unclear, but you wanna say there's kind of something going on. And I think this demonstrates that there must be some kind of continuum happening, again, emphasising that consciousness is not the properties we often assign to what human consciousness does, some of the complex things that it does.
You can have less complex agents that we like to think are conscious in some sense, and yet lack many of the functions that give us our complexity. So I certainly think that's a continuum. And if it is a continuum, then you must be able to get down to a point where there's like almost nothing but something. Some people think that's where you get this sort of plant stuff going on. But if plants do have any sense of consciousness, it would literally be like the person I described a moment ago who didn't have memory, except he also doesn't have the ability to feel pleasure and pain. If you take away sentience, if you can't feel pleasure and pain, you can't have preferences. To have a preference is to say there's a way the world isn't that you want it to be, which means that you have a sort of a pleasure, pain, a preference thing going on. If you don't have that, then there's no volitional action.
So there'd be no action, there'd be no thought, there'd be no emotion, there'd be no memory. And again, if you can imagine that something like that could be conscious, then maybe a plant could be. But if you can't, then it seems definitely not.
Brian: And we do have a question actually where the questioner is not in the audience, but I'm gonna ask it anyway. It's, well, I'll ask the question. It's what does the panel think of some of the increasingly popular theories of consciousness? So, things like panpsychism, integrated information theory, I don't know what that is, but maybe you do. Do you see any of these as scientifically productive?
Oh, the question is, "or more a matter of philosophy." I don't think we should just consign it to the bin labelled philosophy on the desktop. But so, I suppose these are popular ideas that are out there that you may call pseudoscience or not. So panpsychism, integrated information theory.
Anil: And those are two examples, I think, of different things. So there's a lot of theories of consciousness very broadly. I mean, depending on where you draw the line, a recent review that was published had about 200. And paper I co-authored a couple years ago had 20. There's a lot. Some of them are more philosophical theories, like panpsychism is this idea that consciousness is built into the universe at the most fundamental level, like mass or energy or charge. And that's a kind of a popular idea, though in my view, it doesn't actually get us anywhere at all because it basically solves a problem of consciousness just by assuming that it's already there at the beginning, and you generate a whole bunch of other problems which are just equally difficult, but different problems.
So that's a philosophical position that you can't rule in or rule out by experiment. I think those kind of positions should be worn lightly. You know, we can't ever prove them. Are they useful, is the right question to ask in my view. I don't think it's particularly useful. Then you have theories that try to say, well, why is it, and how is it that things like brains can give rise to or be associated with consciousness? Integrated information theory's one of these theories.
It's a bit more controversial than some other theories because it sort of says that consciousness is a property of brains, but it could also be a property of many other things that share some features with brains. And it's quite difficult to test for that reason. I think it's a useful theory. It's not a theory I personally sign up to, but I think it has the right to be wrong. I think it poses interesting questions for us, and it can lead us to do useful experiments that we might not do otherwise. And for me, that's the criterion of whether something is usefully in the game or not.
Brian: Yeah. Alex.
Alex: Is panpsychism a pseudoscience? That is, in my industry, what we call a leading question. I understand that there's a deep scepticism to this kind of view, but I think it indicates a distinction in the kind of thing we're talking about. And Anil just said, and I think it was in the question too, is this scientifically useful and does it sort of do anything? It's possible that it doesn't, it's possible that it has no effect whatsoever. But it kind of depends on what you're interested in. If what you're interested in is coming up with a theory that will be useful for making predictions, or for discovering new medical techniques, or whatever it may be, then cool, that's a noble task. But that's a different question from being interested in the nature of things. So it may be that it is not scientifically useful, but it might still be scientifically interesting.
And so if somebody wants to ask what is actually the explanation for consciousness? How does it arise in the world? Somebody might suggest that the two options traditionally in philosophy have been dualism, as in there's two types of substances. There's like a little soul. You can imagine it being like the captain of the ship of the body. It's sort of inside you some way, or like up there in your head, but it's a separate thing from your body that there are two distinct things, dualism. Or there's this idea that actually, no, there's only the material stuff, there's only the atoms and the way that they interact, and that somehow if you put them together in the right way, you get this thing called consciousness. We're not entirely sure how, but it does it somehow, those are your two options. The panpsychist is dissatisfied with both of them. Says that there's clearly not like two sides of reality that are completely, it's completely mystifying how they'd interact with each other. But it also seems wrong that you could like throw a bunch of atoms together and get consciousness.
So the panpsychist says, well, dualism seems dissatisfying, but so does this idea that you can throw an atom together and get consciousness, two different types of things. It's like adding two plus two, plus two, plus five, plus seven, and eventually getting like the colour orange or something. It's like just categories of different things. It doesn't seem to make sense. So the panpsychist says, what if there's a third option? What if it's that consciousness is there from the base? Now that seems a little bit weird, you know, why would consciousness be there at the base? Remember that consciousness would be incredibly rudimentary on this form. We're not talking about atoms that, you know, worrying about the rain today or something like that. We're just talking about it being that there's some kind of mental property at the foundation of reality. That's what the panpsychist says. And people look at that and say, but it's a complete mystery why that would be the case, and how you would put those little conscious things together and get unified selves. And the panpsychist says, yeah, sure that is a great mystery. But it's much more of a mystery how you would do the same thing with inert matter and atoms that we don't even know the nature of. So that's the panpsychist position, I think.
Whether it's pseudoscientific is up for you to decide, but it may well be proper not to call it a scientific theory at all. But then remember that the scientist tells you what things do. They tell you functions, they tell you how the brain behaves, and the correlated activities of the brain that go along with the conscious experience. If you're interested in the conscious experience itself, if you wanna know what that is, then maybe it's not scientifically useful, but I still think it's interesting and worth pursuing.
Brian: You know, I think we should move on. Carl Sagan said that a physicist is a hydrogen atom's way of learning about hydrogen atoms.
Alex: Yes.
Steve: Can I just say very, very briefly, I think that these views are interesting to talk about, but they are primarily metaphysical worldviews, that they're arguing that we need a radical change in the way we see the world, the universe, and everything in order to explain consciousness. And I think that this is because they're trying to solve a problem that doesn't exist. I think they're trying to solve this invented problem that consciousness is somehow an intrinsic essence that needs to be explained across all possible systems.
And as soon as you go down that road, you end up taking it away from where we started, which was the human case, the case we know best, which we think is allied with the properties of having a mind and being a psychological agent. And I feel like that's needs to be our starting point. We need to cleave closely to the psychology here. And these views are radically depsychologizing, to use Keith Frankish's term, consciousness. And I think that's taking us down the wrong track.
Alex: If you wanna know what it is rather than what it does, then I think you have necessarily strayed into the realm of philosophy, which is a bit of an intellectual no man's land. But welcome, everybody.
Brian: With that, we'll go to the next question.
Ruth Goldsmith: Hi, I am Ruth Goldsmith. I'm a writer from London. Psychosis is often described as a loss of contact with reality, but without offering a definition of reality. As someone who's tried to make sense of these experiences, both in my life and in my writing, I'd like to ask the panel how outlier experiences like this, disruptions to perception, including hallucinations and delusions are informing broader consciousness research.
Katharina: It's a beautiful question, and really touches on the heart of our research, and mainly what drove me to study psychosis, because I was really fascinated by this kind of detachment from what I would call 'shared reality' of what happens in psychosis. And so it very much informs our research on psychosis, and I think an extension about perception, and kind of like the whole range of human experiences because I think studying an extreme case of experience can teach us a lot about the entire spectrum of human experience.
People who have psychosis, when they have these experiences, they feel absolutely real to them, or they can feel absolutely real to them. The same way our experiences, when we don't have psychosis feel real to us. And that's why I don't like the term, normal or objective reality, because these are as valid as experiences as are any other experiences. But they often, or when people seek help, when I do see them as a psychiatrist, these experiences have gone to an extreme. They have detached so much from the rest of society, from the rest of all other people that they become dysfunctional, and people experience distress because they are so far out of what is adaptive conscious experience, is kind of also telling us how we really need to kind of think about shared reality rather than one ground truth that is there and that is the correct thing.
Steve: One part of the question was how research on hallucinations might contact theories or the science of consciousness. And I think that's happening recently in the form of theories that are proposing that one of the important aspects of consciousness is being able to pick out the features of internal brain activity that do reflect the real world. So we know that the brain is continually doing things internally, running internal simulations, doing all the kinds of housekeeping activities and so on. And we're unconscious of those usually.
And so the idea is that whatever is tagging something as reflecting reality right now is part of what we call conscious experience. And in the case of hallucinations, that seems to be going wrong because people are treating some of these internally-generated experiences that are not driven by external factors as real. So that distortion in what we call reality monitoring might help us understand what conscious experience is.
Brian: Well, thank you. The next question, please.
Tony Jenkins: Hello, my name is Tony Jenkins. Experiments have suggested choices are made before we are conscious of making a decision. Is consciousness a conductor, participant, or observer when we decide on the course of action?
Brian: That's a great question. Who would like to start?
Steve: I mean, this is a really interesting story of experimental cognitive science. So the experiments that you are referring to were conducted in the 1980s by Benjamin Libet, showing that you can detect correlates of a choice that someone's gonna make in brain activity before they become consciously aware of making that choice. And recent computational models have suggested that actually maybe we're interpreting those data the wrong way round, that actually what's happening is it's the fact that the experimenter has back averaged that activity with respect to the time of conscious awareness that leads to this ramp-like signal that we then interpret as being somehow some kind of precognition in the brain before we make the choice.
And I think that's now been largely shown to be not really the right way of thinking about those data. So I'm sceptical that there really is this big disconnect. I think it's absolutely true, there's lots of internal brain processes that we're unaware of, but I think when we are making a choice, our conscious experience largely goes in line with the timing of that choice as well.
Anil: I think that the question of free will, which is really the question here, right? Does consciousness--
Brian: Briefly then.
Anil: Another three weeks, right?
Alex: Who's doing philosophy now?
Anil: But does consciousness come along with our voluntary decisions, that's the question. Is it an observer? Is it a participant? I think I'm very much on the side that it's a participant. The idea that consciousness kind of reaches in and changes what the brain does so that it would do something different. I think that's something we don't even need. That's a kind of free will that is not necessary to explain what we do or even what it feels like to exercise voluntary control over our action.
I think we have just as much freedom to do what we want to do as we need. And consciousness is part of that way in which we escape the chains of immediacy, and do things that are less constrained by what's happening around us at this precise moment. And that's all the free will that we need, and that's all the free will that we have.
Alex: Do we agree that we are doing philosophy to some degree now?
Anil: Of course.
Alex: I think.
Anil: Yeah.
Brian: Do we have free will?
Alex: I didn't bring this up myself. I don't think we have free will. I also, I agree with the suspicion of Libet's experiments. I think that that's a, it's a bad basis, that experiment in particular, to ground the view that there's no free will. But I think there are just good philosophical reasons that there's no free will, which I won't go into because it's perhaps tangential. But the question as relates to consciousness, and I suppose the science of consciousness, it's like, you know, what is the role of the agent here? Are you observing? Are you doing? Are you the progenitor of this behaviour?
And I think one of the most interesting places to look to answer this question is split brain patients. So most people know that the brain is split into two hemispheres, right? And they used to be a quite extreme remedy for epilepsy, for extreme cases of epilepsy, an extreme treatment would be the severing of the connective tissue. The corpus callosum connects the two hemispheres. They can still communicate a little bit afterwards through other means, but basically it means that your two hemispheres are not communicating with each other very well. And the strange thing about it is such people seem to just be completely normal. You'd meet them and you wouldn't even know that there was anything wrong with them, except in a few circumstances. So you might know that the right hemisphere of the brain broadly governs the left visual field, the left hand, the left hand side of your body. And the left hemisphere governs the right visual field. It's also the case that the left brain is thought to be the language processor. This is where you get your sort of words. The words that are coming out of my mouth are being governed by the left brain. And the right brain is typically in popular culture, thought of as the sort of creative one. You know, it's more imagery that's going on in the right brain.
So here's an interesting experiment. Take a split brain patient, someone who's had their brain severed in that way. You can show instructions and information to just one side of the brain and not the other. In one experiment, I can't remember the experimenter's name. People who'd signed up for an experiment, so they were awaiting instructions, these split brain patients. They were given the instruction just to their right brain to get up and walk over there. Says, go and walk over to the door. So they stand up, they walk over, and they stand by the door. And then the experimenter says to them in words, "Why did you just do that?" And you know what they say? It'd be weird enough if they said, "I don't know", right? That would be really strange. But they don't even do that. They make up a reason. And the funniest thing about this is that they're not lying. They're not just embarrassed, and like, oh. They say something like, "Oh, I was getting warm and I just went to get some air." And they believe it, even though we know the reason they did it is because they were fed this instruction.
But it implies that there's this idea of the left brain as the interpreter, as it's called, whereby the right brain sort of does stuff and then the left brain comes up with why it was done. And so on the question of whether, you know, we are observers to our own behaviours, I think this can offer some insight. And it's obvious in the split brain patient, but something like that might be going on in normally functioning brains as well. It's just less easy to measure. So I think there's a lot to be said for the idea that a lot of the time when we think we have access to the reasons why we're choosing to do particular things is actually retrospective, made up, and essentially illusory.
Brian: We've got time for one more question.
Imran: Okay, hi, everybody, my name is Imran. I would like to ask the panel, with the possibility of artificial general intelligence being achieved in the near future, do you think that this could lead to artificial consciousness? And how would we know that AI is actually conscious? And then at that point, would we need to redefine what it means to be alive?
Katharina: I love this question, because it touches also on the concept of life. And I think actually the concept of life is a very nice parallel for the concept of consciousness in the way we are preoccupied with finding a definition. And I think not so long ago, people were having heated debates on what it means to be alive, how do we define life? And as biology progressed, we now understand that there are cells, there's reproduction happening in these cells. These cells have a metabolism where they generate energy from some nutrients.
And I think what has happened with the concept of life is we still don't have one unique definition, but we have defined through science, a set of principles that we all agree are somewhat related to life. And we still do have entities such as viruses that tick some of the boxes. They fulfil some of the principles. For example, viruses have reproduction, but they don't have other properties such as metabolism, because they need another cell to reproduce. And we are comfortable with that. It doesn't kind of derail us from feeling that we need to work more to understand life.
And with consciousness, I think something similar might happen. We're still not there yet. We still need to understand all these properties. But I think we might settle at some point on a set of principles. And these principles might be, having recurrent network activity, and having subjective- yeah, image forming eyes, for example. And there will be cases that tick all the boxes, such as humans. But then there might be cases such as artificial intelligence that have recurrent network properties, but they don't necessarily have eyes. And we will be comfortable with saying, yeah, we still, we have this thing, we know where it sits in this kind of like spectrum of principles. And so all we need is to just move on with, I think, science, to define more of these principles. And that will give, leave us comfortable with knowing what exactly artificial intelligence is in terms of like conscious experience.
Brian: And just in the question, there was essentially a question about the Turing test, wasn't there? Or how would you define, so we've got an AGI sat there, how do we decide?
Steve: I mean, I think that my short answer to this is, yes, the AI will have what we are comfortable calling consciousness, and I think actually the progress in AI will do more than perhaps neuroscience has done to reconceive what we mean by consciousness, in a similar way to the way Katharina described the impact of molecular biology to reconceive what we mean by being alive.
So if you think of consciousness as a psychological property as a set of functions, then we are already on the way to emulating some functionality of human psychology in AI, and most notably recently natural language. And I think that increasingly, people are gonna be comfortable with attributing consciousness to these systems. Now, at the moment, I think that's mistaken because I think they are seeing patterns which are not really there. We see signs of linguistic conversational interaction, and so we think therefore that that system has some kind of independent internal experience, and it's going away, and thinking about us and so on. I think that's mistaken to currently hold that view. But eventually, when these patterns coalesce, and these systems become more multimodal, embedded in time, have their own agency, shared project, shared goals, then we will attribute consciousness of these systems, and we will be right in doing so, I think.
Anil: So if we started this discussions with broad agreement about what we thought consciousness was, so it's kind of nice to finish on maximal disagreement, I think. I very much think no. I think it's a terrible idea. We should not be trying to build conscious AI. Fortunately, I think it's incredibly unlikely. The point I will agree with Steve is I think already we're seeing people attribute consciousness to AI systems. This is already happening. And the consequences of that are already pretty disastrous, as we've seen. Some people have committed suicide with conversations with language models plausibly being a factor in that.
If we attribute these systems with capacities for feeling, understanding that they don't actually have, we open ourselves to massive new psychological vulnerabilities, and we distort our circle of moral priority as well. We may end up caring more about a hunk of silicon and code than about non-human animals, or other humans that really do deserve our care and attention. So even AI that seems conscious, even though it isn't, I think is a very, very dangerous thing to unleash into society. I was very reassured that Mustafa Suleyman, who's the CEO of Microsoft AI, basically made this point in an essay recently. And it's really, I think, encouraging to see that view taken because the future of AI is not yet written. We don't have to design AI so that it leverages our psychological biases these ways.
Brian: But your moral argument though rests on the assumption, you just said it there, that these are not conscious.
Anil: That's correct. So I also think that they're not conscious, and I don't know that for sure, because back to the question about plants and things, we can't know for sure. But I think there are very, very good reasons to doubt that AI systems are conscious. And I'll give you just one of them.
So the idea that AI, as we know it, these programmes running on silicon-based computers assumes that consciousness is something that you can implement in something made out of silicon rather than carbon. It's something substrate independent, you might say. It doesn't depend on what we are made of. It depends on the computations, the algorithm that is being implemented. And why would that be the case? It might be the case, but it's a very, very big assumption. And it only seems plausible if we think that the brains that we have are computers.
But that's a metaphor. It's been a very powerful, a very productive metaphor, but it's a metaphor nonetheless. And the closer you look at a brain, the more you peer inside it and just are faced with its implacable complexity, the less like a computer it looks like. And in particular, you do not see anything like this sharp distinction between mindware and wetware as you see between hardware and software in a computer.
Intelligence is one thing. Intelligence is the ability to solve problems to do things. But consciousness, back to our beginning, is about being, it feels like something to be a conscious organism. That is not something that is by definition an algorithm. It might be, but it's a huge assumption. You can certainly simulate the brain in a computer, but a computer simulation of the brain in all its detail might generate consciousness no more than a simulation of a weather system might actually generate rain or wind.
It's a completely different thing. We are living creatures. Consciousness is a property of our biological life, and our inheritance, you know, across millions of years of evolution. I don't think it's a program. We diminish ourselves if we think that we're nothing more than algorithms.
Katharina: I think AI might take a space in between full consciousness and kind of no consciousness.
Steve: I think it comes back to whether we think of consciousness as a process, something that the system does. In that case, I think you can ask questions about whether certain substrates can support certain functionality. And we don't know whether artificial systems can support the full range of human psychology yet. But I think it's a bit like saying, well can lines of computer code really play Go, or play chess? Well, yes, they can because it's doing it, right? It's cooking it up to the real world, and beating someone at that game. If consciousness is like that, then I think there's no reason why we can't implement it in artificial systems, so.
Brian: Alex.
- I agree with that, which is why, if all the consciousness is, is what it does, it's gonna lead to a different conclusion to what it is. So maybe it is scientifically useful after all to try and find out more than just what it does. The good news is, of course, the question asked, not only whether AI could become conscious, but how we would know. And the good news is that, so I'm told, one day, the scientific method might be able to fully explain consciousness, and we'll be able to just crack it open and take a look, right?
And in fact, you may remember not that long ago, the protein-folding problem, which was a huge problem for biology, was solved by artificial intelligence, 'cause it was better at doing that kind of science. So, maybe artificial intelligence, as the technology develops, will get good enough to give us a scientific explanation of consciousness with which we can then test if it's conscious.
Of course, well. Not so fast, because of course, it might try to deceive us on that front. However, I'm suspicious of this, of course. And that's the point that I'm trying to make. Although, it is an interesting thought experiment, isn't it? What would it be like if AI was conscious? We sort of talk about it all the time, and we kind of know what we mean. But really think about this for a second. ChatGPT being conscious. Cool. How many ChatGPTs are there? Like do you mean ChatGPT, like the whole thing? Do you mean like the individual conversations? 'Cause if I pull it up on my phone and say I'm having a bad day, it'll say, oh gosh, I'm so sorry. That sucks. And if you say you're having a good day, it'll say, oh, that's great. So it feels good and bad at the same time.
No, if they were real experiences, we'd want to distinguish them, we'd want to say there's a conscious AI, and here's a conscious AI. But we all know that it's actually kind of one thing just manifesting in different computer systems. And so I wonder if it ever became conscious whether you could sort of give it the equivalent of a psychedelic drug, and it would have this transcendent ego death, and realise that it's part of this whole sort of network of ChatGPTness. But I think that the relationship would be quite similar. The relationship that we have to consciousness would be like the relationship that ChatGPT, as an individual language processing on your phone, has to ChatGPT, the company, the computers locked away in a server room somewhere. Which is to say a little bit complicated and a little bit mysterious.
Steve: There is so much to say about this.
Brian: Everybody's made their-
Alex: I think we're out of time.
Brian: That is really, sadly, I've enjoyed that massively. That is all we've got time for. So thank you to our wonderful panel, Anil Seth, Katharina Schmack, Steve Fleming, and Alex O'Connor.
