以下是曼若提教授(哲學家)和瑟史教授(大腦神經科學家)關於「意識」的對談。我對「意識」的了解是從認知科學和大腦神經科學入手，自然接近瑟史教授的觀點。

兩位對談中提及「人工智慧」和DMT所產生的幻覺等；可參看本城市《人工智慧縱橫談》和《現實與對現實的認知》兩欄貼文。


The hallucination of consciousness

Is consciousness in the brain?

Anil Seth/Riccardo Manzotti, 06/29/23

The hard problem of consciousness has puzzled philosophers and neuroscientists alike for decades. Here a philosopher, Riccardo Manzotti, and a neuroscientist, Anil Seth, meet to discuss consciousness, the hallucination of reality, and whether consciousness is inside or outside the brain.

I met Anil Seth many years ago. I think it was 2003, give or take. We had the first of many discussions about the nature of conscious experience. Although we share many ideas, we are also very different in the conclusions we draw. He takes consciousness to be an emergent phenomenon taking place inside the brain and I take consciousness to be the external world impinging on the brain (外在世界強加到大腦上). He believes that perception is a reliable hallucination, I believe hallucinations are a form of perception. This shows the richness of scientific debates on controversial topics such as consciousness, where scientists with different approaches and insights can benefit from a candid discussion. It is common knowledge that consciousness is still a great mystery to science and that although there has been a tremendous growth in scientific work on consciousness in the last decade, there is still no accepted theory on how consciousness arises in the physical world.

Anil Seth has been involved in the study of consciousness for 30 years. As a young assistant, he worked with Nobel laureate Gerard Edelman and then moved to the United Kingdom, where he eventually became director of the Sussex Centre for Consciousness Science. He has published hundreds of articles and his latest book, Being You, is an international bestseller that has won umpteen awards. The book has helped make Seth a world-renowned celebrity in science. But with great power comes great responsibility, and so I am pleased to have a dialog with Anil Seth about his views all the more because our approaches are apparently so distant.

Given such a unique combination of friendship and differences, I very much look forward to our dialogue.

Hi Anil, how are you? I am just holding in my hands the new copy of your book, which has just been published everywhere and eventually in Italian. I see that it was incredibly successful: a Book of the Year for Bloomberg, the Guardian, the Financial Times and the Economist! That's very impressive for a scientist! What do you think is the secret of this success?

Hi Riccardo – I’m doing well, and I hope that you are too.  It has been a long time – but one of the great privileges of a life in academia is the pleasure of conversations that unfold over years or even, as in our case, decades. I am very happy to have the chance to continue our conversation – and productive disagreement – about the nature of consciousness.

Thank you for the kind words about Being You. I’ve been thrilled to see how many minds it has reached. During the writing process, I admit I lost sight of the prospect that people might actually read it, and I’ve been surprised and delighted that many now have. As to its relative success, it’s hard to say. I did my best to find the balance between detail and accessibility; to make it readable by anyone with a curious mind while still saying something new and distinctive about a well-worn topic. And consciousness, though well worn, is intrinsically fascinating to many people. We’ve all wondered about things like: do I have free will? Why am I me, and not someone else? What happens when I fall asleep, or go under anesthesia? The nature of consciousness is not just a deep scientific and philosophical mystery: it’s a mystery that matters, to each one of us, every moment of every day. This all helped. But really I must pay tribute to everyone who has helped me figure out my thoughts about consciousness over my years of research, and to all those who read and commented on the book. Although it's just my name on the cover, it was far from a solo effort.

I hope you do not mind if I steer this dialog to the heart of the matter. I am eager to ask you many questions and to understand why you draw your own conclusions.

Let’s do it! I’m eager to know your reactions.

Okay, thank you. First of all, most of your research is based on the idea that consciousness is generated in the brain. The Italian title, "Come il cervello crea la coscienza", is pretty clear on this, even clearer than the English version. Not even a question mark! On what basis are you so sure that our consciousness is created inside the brain?

As you probably know, authors rarely get to choose their titles – especially for translations. But I won’t wriggle out of the question so easily. I do take it as a working assumption that consciousness comes from the brain: more precisely, that the constitutive basis of consciousness lies in the brain. I do this because of the ample evidence linking the two. If the brain stops, consciousness stops. Any change in consciousness seems to be accompanied by a change in brain activity, whether this is global change, for example as happens in anesthesia, or a local change as happens when what we are conscious of changes from moment to moment. The challenge, of course, is to explain why this is the case. And this is the task I take on, from a particular perspective, in my work and in my book.

One important part of this story is that although I think that consciousness comes from the brain, the nature of what we perceive depends on the body and the world. Our brains are embodied, and our bodies are embedded, and the interactions between brain, body, and world are fundamental to the ways in which we consciously experience the world and the self. But you go further than me on this point, believing – if understand you right – that consciousness itself extends into the world around us. Quite what you mean by this has always puzzled me a little.

Let me ask you another general question that has always surprised me. If we did not know that there was such a thing as conscious experience, are there any experimental results that suggest that there is something else going on in the brain besides neuronal activity and metabolism? If you were, say, an alien AI without any consciousness studying the physiology of human beings with no access whatsoever to their subjective reports, would you have any cue that there is something more than neural processes?

What a wonderful question! It’s true that the starting point for consciousness research, whether in science or philosophy, has been rooted in the first person, in introspection. From this perspective, the idea that consciousness is a phenomenon in need of explanation seems undeniable. (Of course, some people - so-called ‘strong illusionists’ - do deny this, but I find this position highly implausible. I do find some appeal in a weaker form of illusionism (錯覺論) which recognizes that consciousness exists but suggests that it might not be what we think it is. I wonder where you stand on illusionism?)

Back to your question. Since your alien AIs lack consciousness, and are also unaware of the fact that humans talk about being conscious all the time, I suppose they would be (unconsciously) puzzled by all the neuronal and metabolic happenings inside human brains. They might try to be proper behaviourists (行為論者), explaining human behavior without recourse to mental states – and especially not conscious mental states (since they’d not have the concept of consciousness). But, like the actual behaviourists that dominated Western psychology throughout much of the 20^th Century, I think they’d come up short.

I think conscious experiences play important functional roles for humans and other animals too. Quite what these roles might be is still much discussed, but a good way to think about it is that consciousness brings together a large amount of organism-relevant information in a format that helps guide adaptive and flexible behaviour, to best help the organism stay alive. We experience the world, and the self, not as it is, but as it is useful for us to do so. So, understanding human (and animal) behaviour likely requires recognizing that consciousness exists, and that it is useful. This is one reason why the neuroscience of consciousness is probably best not left to unconscious alien AIs.

In you book and your talks, you often use the expressions “inside”, “inner world”, “from inside”. These expressions seem a physical version of the mentalistic notion of interiority. What do you exactly mean with “inner world”?

I mean something rather simple. For me, there is an outer world – the world of objective reality, as described by physicists. The inner world is the world of conscious experience. Of course, conscious experiences include experiences of the world, as well as experiences of the self, but they are ‘inner’ in the sense that they depend on the brain.  The world that we experience is an active brain-based construction, it is not the world as it really is. I’m with Kant here, when he claims that it doesn’t even make sense for us to directly experience objective reality. The world as it really is, is always, for all conscious creatures, hidden behind a sensory veil. Just to be especially clear: for me the inner world is all of experience, not only the experiences that we feel as being inside us, like thoughts and feelings.

I suppose that this is a point where you and I may disagree, and that for you, consciousness involves some kind of direct access to the world. I am wondering how this disagreement, if there is one, might be resolved. In one sense, objects in the world, as they appear to us, are part of consciousness – which may be close to what you have in mind – but, for me, the objects as they appear in our experience still remain distinct from actual objective reality.

One important consequence of this way of thinking is that because we all have different brains, we’ll each experience the world in a unique way.  Just as we all differ on the outside, we will all differ on the inside too – even if we do not realise this. One of my current projects is to map out this hidden landscape of ‘perceptual diversity’ by inviting people to take part in The Perception Census (https://perceptioncensus.dreamachine.world).  This is a large-scale citizen science project consisting of interactive illusions, simple experiments and surveys, all designed to measure individual differences in various aspects of perception, including color, time, sound, music, emotion – and much more. So far we’ve had more than 25,000 people from over 100 countries take part, but we’re hoping for many more.  So I’d be very grateful to any reader who is willing to give it a go – you’ll be helping our project, and you’ll also learn more about your own powers of perception, and how they relate to others.

Your main argument is that the reality we perceive (or rather that we believe we perceive) is a hallucination, a reliable one. But if that was the case, how would we know we have a brain? Should that not also be a hallucination? Friedrich Nietzsche raised a similar objection to sensualism (感覺論) in his Genealogy of Morals: If everything we perceive was created by our senses, how would we know there is a body? The body too would be an illusion.

The metaphor for perception I use in the book is indeed one of ‘controlled hallucination’ – where the control is just as important as the hallucination. The ‘hallucination’ part emphasizes that our perceptual experiences are active constructions, brain-based predictions, that come largely from the inside-out, rather than direct reflections of reality as it is. The ‘control’ part emphasizes that the brain’s perceptual predictions are constantly calibrated by sensory signals coming from the world, and the body. Our conscious experiences are tied to the world by criteria of utility, not accuracy.

So how do we know we have a brain? Only indirectly, but that’s OK. We have no direct acquaintance with the fact that we have a brain, whatever that might even mean. That we have a brain is an excellent hypothesis to explain all sorts of data – both sensory data, for those of us that have been lucky enough to actually see, or touch, a brain – but also all sorts of scientific and medical evidence too. What a human brain is really like is still up for grabs, just as what anything is really like is up for grabs. Something similar goes for the body. Our experiences of the body – both from the outside and from the inside – are perceptual constructions, but, like all (non-hallucinatory) perception, they are constrained in useful ways by sensory signals. One major difference between the brain and body, from this perspective, is that for the body we have many sensory signals to go on – reporting on both the body from the outside (vision, touch, proprioception, and so on) but also the body from the inside (so-called ‘interoception’ – the sense of the body from within). By contrast, the brain doesn’t sense its own activity at all – and why should it? This is why there is vivid body-related perceptual content, where there isn’t for the brain.

I should add that I’m not keen on the term ‘illusion’ here, because it implies the existence of a more accurate, correct, way of perceiving. Our experience of the body is not an illusion: it is a perceptual construction that plays a useful role for the organism. When I hear things like “the self is an illusion” I only partly agree.  The idea that there is an immaterial ‘essence of me’ – well that is false, and so any experience that such an essence exists is illusory. But perceptions of ‘being a self’ exist and play a useful role in guiding behaviour and keeping the body alive – there’s nothing illusory about that at all.

OK. Here comes something we agree on! Intelligence is something different from consciousness. So we could have very intelligent machines with little or no consciousness ... a scary possibility!

Artificial intelligence is moving fast, especially in the realm of language models. One point to make here is that we humans are often vulnerable to anthropomorphism (擬人論或人類中心主義). We project properties into other things (machines, animals) on the basis of what it would feel like for us in similar situations. So, people project ‘understanding’, as well a whole raft of more specific cognitive competences, and even consciousness, into large language models like GPT4, in ways that are not supported by evidence. Certainly there is no reason to suspect that current or near-future AI systems have any kind of consciousness at all.

Returning to your question, and as I mentioned before, I think that consciousness has functional utility in integrating large amounts of survival-relevant information from an organism’s perspective. From the perspective of evolution, consciousness provides a very effective solution to the problem of guiding behaviour, and regulating the physiological condition of the body, when things get sufficiently complex. As I put it in the book, we experience the world, and the self, with, through, and because of, our living bodies.

Note that for consciousness to be functional, we do not have to assume that conscious experiences have causal power over-and-above the physical stuff inside our brains and bodies. Making this assumption means smuggling in an unhelpful dualism, where conscious experience and the physical world inhabit separate domains of existence, interacting somehow or other. Consciousness being functional is perfectly compatible with a non-dualistic worldview in which conscious experiences are properties of embodied brains. Is it also compatible with your view in which consciousness extends out into the world itself? I imagine it would be, but I’m curious to know.

If there are no colors in the physical world, how can the brain, which is part of the physical world, have colors? You might answer that the brain has no colors, but that the brain sees colors. But that seems even more confusing to me. How can the brain see properties that are not part of the physical world? How can the brain actually see anything at all? Is the brain the kind of thing that sees? I do not see how.

It was Cezanne who said that “color is the place where the brain and the universe meet” and he was on the right track. When we experience color, we experience both less than what’s really there, and more than what’s really there.  Less than, because our eyes are only sensitive to three wavelengths of a spectrum of (colorless) energy that goes all the way from radio waves to x-rays and beyond. More than, because out of these three wavelengths, the brain conjures a palette of millions of distinguishable colors.  Why does it do this? It turns out creating perceptions of color is a useful way for the brain to track the reflectance properties of different surfaces, especially under changing lighting conditions. Again, the point is that we don’t see things as they are, but in ways that evolution has determined it is useful for us to do so. 

In my book I give an example of looking at a red chair. In this example, the redness I experience depends both on properties of the chair and on properties of my brain. It corresponds to the content of a set of perceptual predictions about the ways in which a specific kind of surface reflects light. There is no redness-as-such in the world or in the brain. As Cezanne said, redness happens where the two meet.

An important point in your approach is the notion of prediction. This is another point on which we agree. The brain is probably a prediction machine, similar to recent advances in AI, from RNN to LLM and Transformers. This idea has been around for some time, possibly since Helmholtz himself. You’ve just said that our perception of, say, the color red, is the brain's best guess about the reflectance properties of surfaces. But why should the brain's guesses have a phenomenal character? Should not they just be guesses?

This question dangles the hard problem of consciousness temptingly in front of us: What is the ‘special sauce’ that conjures experience out of mere mechanism? I do not think this is a productive question to ask. In my preferred ‘real problem’ approach, the goal is to explain, predict, and control properties of consciousness in terms of things happening in brains (and bodies and worlds). The short account of redness I gave above is one example, though there’s much more to be said about even this (as I do say in the book). My overall strategy is to explain every kind of conscious experience as a form of perceptual prediction. At the end of this road, my hope is that the ‘hard problem’ question of why there is phenomenal character at all will fade away, disappearing in a puff of metaphysical smoke.

In a physical world, what are hallucinations made of? I do not see what hallucinations can be made of, unless you already assume that you are in a hallucinatory world. Is not that a kind of dualism in disguise?

Definitely not! My working principle is pragmatic materialism. To rehearse what I just said: I explore how far we can get by assuming that conscious experiences are properties of matter, organized in particular ways. The questions I am interested in accept that conscious experiences exist, and ask how their properties can be explained in terms of processes unfolding across brains, bodies, and worlds. I do not ask, or answer, the metaphysical question of what an experience is ‘made of’. But if you press me: well, they’re made of the same stuff that brains and bodies are made out of.

On the one hand, there are hallucinations. On the other hand, hallucinations and dreams are very commonplace. For example, why do not we hallucinate completely alien colours? Or completely otherworldly geometries? Or completely new and strange phenomenal properties. Even the reports of hallucinators using psychedelic drugs are remarkably similar to our experiences in normal perception. Why is this so?

I think you’re only partly right here. Sure, many dreams and hallucinations have the character of seeming rather normal, or perhaps variations on normal, everyday, perceptual experience. And this is perfectly understandable since perceptions, dreams, and hallucinations all depend on the same underlying neural machinery of perceptual predictions. But there are some cases of hallucination which rejig this neural machinery sufficiently that very strange experiences can occur. Under some psychedelics, such as DMT, experiences of time and space can collapse completely, in ways totally different from everyday experience. In a less dramatic example, we’ve found that hallucinations induced by flickering bright light on closed eyes – in our “Dreamachine” – people can experience colors of a depth and vividness they’ve never previously encountered.  Alien colors, one might say.  And this works for colorblind and visually-impaired people too – they find the Dreamachine extremely moving (https://dreamachine.world/). Over last summer we had about 40,000 people experience the Dreamachine, and the vast majority came away with a new fascination about philosophy and neuroscience. Because the mystery of consciousness really is a mystery that matters, to each one of us, throughout the whole of our lives.

The dialog will go on forever – nothing can stop a philosopher and a neuroscientist from debating the problem that lies at the natural crossroads between science and philosophy: our very existence, namely consciousness – of course, but we must stop here. Nevertheless, I would like to add a final remark. First, I must stress that the intellectual honesty and scientific soundness of Anil Seth's responses. They show how scholars can indeed be unite in their diversity. Second, and surprisingly, our dialog caused me to reconsider the apparent distance between our views. After reading Seth’s responses twice (or “trice”), it has started to dawn on me the impression that, surprisingly, there may be common ground. As in the beautiful quote by Cezanne, "redness happens where the brain and the world meet.”. But where is it? For Anil Seth, who is a neuroscientist, it is natural that redness happens inside the brain. For me, who am a philosopher and therefore able to afford to be bolder, redness happens in the external objects – it is one with the external object. But is there anything empirical that prevents our views from merging and complementing each other? I wonder ... maybe not. But there will be another time, in Seth's words, to continue our “conversations that have evolved over years or even, as in our case, decades.” That's the joy of intellectual pursuit.

Anil Seth is Professor of Cognitive and Computational Neuroscience at the University of Sussex, and co-director of the CIFAR Azrieli program in Brain, Mind & Consciousness
Riccardo Manzotti is Professor of Philosophy at IULM University (Milan).