Transcript of journalist and senior media executive Richard Sergay's interview with Christof Koch, PhD for the "Stories of Impact" series.
Watch the video version of this conversation
RS = Richard Sergay (interviewer)
CK = Christof Koch, PhD (interviewee)
Christof Koch, PhD is the chief scientist of the mine school program at the Allen Institute in Seattle. This is a transcript of his conversation with interviewer Richard Sergay, for the documentary series "Stories of Impact."
RS: Tell me a little bit about your background and what the Allen Institute is about?
CK: My original training was in physics, although for the past almost a third of a century of work in the neurosciences. For 27 years I was at Caltech, I was a Caltech professor in neuroscience and in engineering, and then about ten years ago I came here to lead a large scale program involving three not scientists, engineers, and other staff, focused on a few large projects involving trying to identify all the different types of cells in the brain, because they, I mean a brain cell isn’t a brain cell. They’re inhibitory, they project to different regions, they have different zip codes as is where they compute things in different ways, it turns out there are probably roughly a thousand different cell types.
RS: Where are we in the art of understanding the brain?
CK: It’s only really from the Enlightenment that we’re beginning to understand that the brain is really the seat of the mind and in particular of consciousness. Until then, people thought predominantly it was cardiac based thinking, and you still see that reflected today. I love you with all my heart. Somebody is heartless. There are thousands of Sacred Heart academies. There’s not a single sacred Brain Academy. You know, you see statues of Jesus with a glowing heart but never with a glowing brain. It's really a phenomenon that’s happened over the last three hundred years. Now we know that who you are — your traits, your desires, your fears, your dreams, everything you are — is really constituted of the physical, correlate of that is really in your brain. So if we want to understand ourselves we need to understand our brains. So this is, we’re doing this since roughly two hundred years, there’s really a basic science track and then a clinical sort of track associated with neurology and neurosurgery and of course we’ve learned enormously amount of stuff from the clinic about the brain, the importance of specific parts of the brain, for things like mind and for consciousness.
You know you can exist, it turns out, with half a brain, although things we learn from and from clinical neuroscience. Particularly over the last 20 years and in this country with the brain initiative — it’s a large scale initiative, that really put a lot of funding and people onto trying to once and for all get a complete accounting, a complete census of all the types of neurons, what neurotransmitters they use, where are they distributed throughout the brain. Partly because we can manipulate the powers of modern molecular genetics — we can manipulate these neurons if we can identify them by the type of genes they express. Many diseases, many neurological and psychiatric diseases, are probably diseases of specific cell type. So this sort of knowledge will ultimately give us very powerful knowledge and tools to begin to cure the brain of its diseases.
RS: Define consciousness?
CK: It’s, in some sense, easy. It’s any conscious experience, any experience, any feeling of love, of hate, of seeing right now the picture in my head of you — I can see you. I have a voice inside my head when you talk. I can be upset. I can be angry. I can wake up in the morning and remember what I did last night. Those are all different conscious states. It could be as banal as tasting a slice of old stale pizza. Or it could be as exalted as having a numinous experience from the top of a mountain and feeling I’m in the presence of something godlike. Those are all different aspects of something that we experience every day. This only goes away during deep sleep. Of course sometimes when we sleep we wake up inside our sleeping bodies and we have conscious experiences. We fly, we encounter long lost relatives, friends, and pets. Those are dreams. That’s a different conscious state. But apart from those states when consciousness is gone, normally like in deep sleep, or when I lose consciousness like when somebody hits me on the head or when I’m in a car accident etc. when consciousness is literally lost. I am conscious all the time from probably the beginning of life until its very end. So this is, it’s so basic that it’s difficult to understand how could it not be. Well, but we could imagine zombies. What philosophers called zombies. We could imagine creatures and we see them in movies, but we could at least — certainly it’s logical and compatible with everything we know about physics and biology — but there could be creatures that have brains like ours that behave like us. They talk like us, they move like us, but there’s no feeling. There’s just nothing inside them. They don’t feel anything. They’re all action without any experience. But of course we know in the real world it’s not like that. I have feelings. And so the question is — the mystery of the beating heart, the mind body problem — how does consciousness get into the physical world? How does it get into my brain, as it were? Because consciousness is not in the foundational equation of physics. It’s not in quantum mechanics, it’s not in general relativity, it’s not in the periodic table of chemistry. It doesn’t seem to be in the endless ATGC chart of our genes, yet once again we are conscious. So what’s the relationship between physical systems like my brain and conscious activity? Why do certain systems like brains have it but other systems like my liver don't appear to have it. My liver isn’t conscious, or if it is it’s not telling me. My computer, I have no evidence that it’s conscious. The other conscious creatures around like dogs and cats and other creatures, so those are some of the key questions we’d like to understand.
RS: In terms of defining consciousness, what are the parameters you use?
CK: Of course consciousness, studying consciousness, is a little bit more difficult than studying black holes or viruses or nerve cells. Because they have what philosophers refer to as only third person properties. So you can weigh, you can compute the mass of a virus or a neuron on a black hole, at least in principle. You can infer its influence on other neurons, right, by recording from the brain using a variety of techniques. Now consciousness is different because consciousness has what’s called, what’s known as a first person property. Only I have direct acquaintance with my conscious experience right now. I don’t even know whether you’re conscious. It’s a problem, you know, how do I know you’re not a zombie? Well, here we use an inference project and a type of infinite, a reasoning called inference, in particular it’s called abduction. We infer to the best, the best explanation given all the known facts. I know a fact: you, you’re human like me. You look similar although not identical to me. If I put you in a scanner, I’m willing to put down a million dollars that you’ll have a brain like me and look roughly like me. You behave very similarly. Your neurons are like me, our genetic history is shared, so therefore I assume that you too have these conscious states, and then indeed you tell me about them. You say yes, I can see the beautiful lake here. So I assume you see it too — now, of course you could be a zombie — but I think that, that possibility so remote, because it would imply that this one set of, of physical, psychic loss for my brain and there’s an entirely different set for your brain and I see no reason to assume that. So I assume your brain also studies, generates consciousness, or is capable of experiencing states. And so now, with that assumption, I can study it. I can show you colors for it, and I can ask well, are these two colors the same or different? Do you see a man or woman here? And, you don’t even have to talk. I can just ask you to press a button. I flash up a face and I ask you very quickly to tell me if it’s a man, push this button, if it’s a woman, push that button. And then certain parts of your brain will be active and they are similar overall to the paths of the brain that are active when you are asking me to make these distinctions. And so it’s always this adductive process. I assume you have consciousness and then I can make all sorts of inferences between your brain and my brain, in fact the brain of all people, of all you know, relatively neurotypical people. Because there’s some people that are unable to have certain discrimination. These folks, they might not be able to read properly, see faces at all. That’s called face blindness. But then their brain differs in interesting way from, from the brain of neurotypicals, or you may be one of the seven percent of man that’s colorblind, where you’re unable to make certain types of red-green or red-orange distinction that, that most men and women can make. But again then there’s sudden interesting differences in your genetic makeup compared to mine. [14:38:37.02]
RS: What are the leading theories and are they competing?
CK: What are the leading theories of consciousness? Well, so first of all that relates to the question, what do you mean by a theory? So many people have a theory. Oh, I have a hunch, consciousness must involve this particular part of the brain or that particular part of the brain. That’s not a theory. A theory to me coming from physics as a well-organized set off of a primary statement. The smaller and the more powerful the theory — that can then be sort of from which you can then infer a whole class of phenomena and the larger the class of phenomenon you can infer — the more powerful the theory that is. And again, the theories I’m talking about are scientific theories that are actually testable and amenable to empirical tests. Otherwise, it’s pure philosophy and I’m less interested in them. So then you can sort of broadly talk about a couple of classes of theories.
Currently, probably the most popular two theories are called Integrated Information Theory, and Global Neuronal Workspace Theory. They differ fundamentally in their assumption. And they make some interesting distinction between where they locate the footprints of consciousness. A big question of study since roughly 30 years is something that Francis Crick, the Nobel laureate who co-discovered DNA, and I worked on in 1989 and 1990. Essentially, what we said, well look, independent of all of your philosophical take on consciousness, whether you’re dualist or materialist or physicalist, we can all agree that there are certain physical mechanisms in the body in general, in the brain in particular, that are active any time or that seem to be involved anytime you have a specific conscious experience. So when you hear my voice, there’s certain neurons in your brain that specifically seem to correlate with that, very specifically. And if I can find those, I can identify them and ask well, where’s the difference between when you hear my voice and the voice of, let’s say, my wife or when you hear a dog bark? Are those different types of neurons, are they organized in the same type of brain area? Do they all carry the same genes? And what happens when you, for example, see me as compared to hearing me? Different neurons are involved, so I can ask all these questions.
CK: Furthermore I can ask, well what happens once I found those neurons, those footsteps sort of,of consciousness, and I artificially activate them — using for example as surgeons new electrode or some outside device — can I then induce you, can I then make you hear my voice, although I’m not really present? Furthermore, if I remove, if I inactivate that mechanism, by some artificial means or maybe because you had a stroke now in this area, what happens? Then, if this is really the footprint of consciousness, you should not hear my voice anymore, even though I’m speaking. Those are questions that are empirically very accessible. These two dominant theories, Integrated Information Theory and Global Neuron Workspace, make interesting but quite different predictions about where the footprint of consciousness in the brain is.
Then there are also two other classes of theories. One is sort of a very poor class of theory that holds consciousness as so mysterious. There’s one other great mystery in science that we haven’t really resolved, which relates to the nature of quantum mechanics. The so-called collapse of the wave function. And so because both things are fundamentally mysterious, well maybe they’re linked by the same mystery. This has to be caused by some sort of macroscopic quantum effect that takes place in the brain, i.e. the idea is maybe while the brain is like a quantum computer, and this would explain many of the mysteries of consciousness. These theories haven’t been quite worked out in as much detail certainly in terms of the neural mechanisms at all, as Integrated Information Theory and Global Neuron Workspace. Furthermore, right now there really isn’t any serious evidence that the brain is really a quantum computer.
There’s no doubt that the brain, like any other piece of furniture of the universe, is just my brain. It's a piece of furniture like rocks and stuff, and black holes. It has to of course follow quantum mechanics. But there’s no evidence that we exploit some of the more esoteric quantum mechanics effects like superposition in the brain. That's after all by the standards of physics, quite hot, and in a wet environment.
And then there’s yet another class of theories that are more closely related that come really from philosophy, so-called hot theories. Higher order thought theory. They say, well consciousness is always, I know that I’m in pain. I know that I hear you, I know that. The idea is that it always has this higher order reference, so there’s not just pure consciousness, pure sensations, always knowledge of sensation. Yes, I’m personally not very sympathetic to those ideas because I think we over exaggerate our self-knowledge. These theories come like all theories from people who are very intellectual and are constantly thinking about themselves. But most of the time when I’m out there in the world, when you run, when you climb, when you, when you speed on a motorcycle through traffic, when you make love, when you play soccer, when you’re war, when you’re hunting, when you’re in deep meditation, pure consciousness explains, and in all those cases, you’re out there without having any knowledge of self, yet you’re highly conscious.
So I think, consciousness of self is the state we are capable of, particularly adults, teenagers and children much less. But it's a subset of all conscious states. And so therefore, for me the sort of really dominant theories that I’m interested in testing in terms of neuroscience is Integrated Information Theory that ultimately says consciousness is associated with the causal power of systems like the brain to determine their own future and to be influenced by their immediate past.
Ultimately it’s about the physics of certain systems that are sufficiently complex, so their previous state determines, their history determines their current state, and the current state influences the future. Any system that has this causal power upon itself will be conscious according to integrated information theory, according to Global Neuron Workspace, that really argues that consciousness is, happens when, information arises through my sense, propagates through the brain, and is made globally available to all the different subsystem of the brain, so it’s really a cognitive information processing theory, very different from Integrated Information Theory. It says that all these individual modules, there’s vision, there’s audition, there’s language, there’s memory, etc. Those are all local processes but once the information in my brain is made broadly accessible, so it broadcasts to all the different subsystems of the of the brain, in that act of broadcasting it becomes, it becomes conscious, and that involves particular the prefrontal cortex, summarizing information and propagating throughout the brain, in particular sending it to the back of the brain, where the specialized visual, sensory, tactile, and other areas are. If this broadcast doesn’t happen, you’re not conscious.
RS: How do you differentiate between the brain and the so-called mind?
CK: They’re totally different. The brain is a piece of hardware. Ok, it’s inside its protective skull. The mind is what the brain does, it's sort of the collective, sort of a collective action of, of both conscious and unconscious. We know since Freud, the massive amount of stuff that I, that my brain is capable of processing, that I have no idea. I mean if you ask me a sentence in English I somehow translate it you know — I’m German of course as you can hear — and, you know, and send it out to my speech apparatus and then it comes out in a well-formed sentence. I have no idea, and most of us have no idea how this happens. I open my eyes and I see the world and all of its complexity. Well, we know this because we have computer vision. That's a very sophisticated operation, again I’ve no idea how I see in stereo, I have no idea how to analyze the difference, the different wavelengths to put it together into a consistent picture of, of color. So the mind is what the brain does if you want. But it’s very different. One is a physical, tangible thing. And the other one is sort of, is the visible and the invisible processing that happens, that happens in the, in this organ, in the brain.
RS: So you locate consciousness exactly where?
CK: I can locate the physical substrate of consciousness in the brain. We know this from the clinic because we know that if people have lost, for example, aspects of their brain, particularly their cerebral cortex, the outermost shell of the brain, they lose a particular aspect of consciousness. They may be unable to hear, to hear sounds, although there’s nothing wrong with their ears. They may be unable to see colors. They don’t experience color anymore although there’s nothing wrong with their eyes. They may not see motion anymore, so there’s a whole variety of conditions when the entire, they’re face blind and they’re unable to see a face. They see the individual eyes, the eyebrows, the lips, but they don’t, they have lost a part of the brain that puts it all together into a percept of a face, like you and I see all the time. We don’t think about it, but it’s a sophisticated process. We also know that if I lose access to my entire cerebral cortex for example, because I have a massive car accident or trauma intoxication, I become completely unconscious. So we know that the physical substrate of consciousness is in, not just in the brain in general, but in particular in the cerebral cortex. Over the last 200 years we’ve learned this the hard way from patients.
RS: Currently the greatest obstacle to understanding consciousness is what?
CK: The brain is by far the most complex piece of active matter in the known universe, by far. By any measure. You know if you just look at a little quinoa grain, a piece of brain like a, you know like a grain of quinoa, it contains on the order of a hundred thousand different neurons. They probably have a hundred different types of neurons, not just one type of new neuron, but a hundred different types in this, in this quinoa grain. This, on the order of a couple of miles of cabling in there, and they are on the order of a billion, a thousand million synapsis, those are the connection points between neurons, and of course those synapses we believe encode my first memory or where I was yesterday. Or, you know, when I last saw my Burmese Mountain Dog. And we have tools that are powerful in one hand compared to anything that went on before, but they’re still very weak. So either I can put people in a magnetic scanner and I can see the entire brain and I can see its activity rise and fall, but I’m actually tracking, there’s not neural activity itself, but I’m tracking the power consumption of the brain. So I’m averaging across millions or tens of millions of neurons. To just see sort of the aggregate. Or I can take a piece off of Silicon wire and put it into a brain either because I need to monitor a patient’s brain for epileptic seizure or I can do it in an experimental animal, and then I can listen to the, to the chat of individual neurons, I can listen to the way they talk. (BRRP BRRP) They send out these pulses. That’s how they communicate. But then of course I only get this tiny pointed list, you know it’s like trying to understand a painting by looking at one tiny pixel. Or you’re looking at a painting and you take off your glasses and all you see is a few color splotches. So it’s, it’s the most difficult, the most complex system we’ve ever faced as science, and so understanding it and its physical substrate is very, very challenging. I think once we have the tools where we can deliberately, delicately, reversibly, turn on or off specific subset of neurons, so I can just right now turn off your ability to see in color, very precise with millisecond precision and these neurons are turned on and those neurons are turned off, then I can understand much better the causal relationship between particular bits and pieces of the brain and particular bits and pieces of conscious experience.
That still doesn’t of course answer the fundamental challenge, so probably understanding consciousness is a conceptual, very challenging problem. Because then still these leaves open the question okay, activity in these neurons give rise to this perception. And activity in those neurons give rise to that perception. But why should there be? I mean once again: my liver, it’s another complex organ, it’s not quite as complex as this, but it’s also quite complex, and has cells that interact with each other. Right now we don’t believe that activity of these liver cells gives rise to conscious experience of the liver. Well, why not? Where’s the difference? And so fundamentally what we need is an empirically accessible and testable theory, that tells us which physical systems and at what conditions half-conscious experiences and which ones will never have a conscious experience. Will a sand dune for example, you know, on the beach, a gigantic sand dune, will that ever feel something? Does the sun ever feel something? You know, does a tree ever feel something? Does a dog ever feel something? Do you ever feel something? So for certain of these systems we have precise answers, but right now it’s mainly based on intuition on this objective process. I believe you’re conscious because you’re similar to me. I believe my dog is conscious, my dog doesn’t talk but he can wag his tail and he can move its ears about its head. And we share, of course, 50 million years of evolutionary history. But the further away we go from things that look like me, the more difficult this inference process, right. So in a squid, it’s already, a squid is, you know, it isn’t even, it doesn’t even have a spine, like a cord decent animal. Is a squid conscious. I believe so, but it’s more difficult to say, is a bee conscious. Is a plant conscious? That's much more difficult. So that’s why we fundamentally need a theory and that’s not an easy task.
RS: Describe for me how the scientific and technological work that you are doing at the Allen Institute is helping you get to an understanding of consciousness.
CK: We’re building it together with others of course and the rest of the community, these tools that can query ever larger numbers of the relevant actors. To understand chemistry, you need to understand atoms and molecules. To understand memory and consciousness and behavior and perception, you need to understand the relevant actors, in this case the atoms or the molecules of the mind are neurons, and collection of neurons. And so I need to be able to understand how many there are, where they are located, and can I record them? Can I do it at the relevant timescale? The timescale is milliseconds or 10 milliseconds, fraction of a second. And can I do this? I know these actors are distributed across the entire brain. So I need to do it at, you know, in many different locations across the brain in this incredibly dense, you know, organ that contains 100 billion neurons where the density of individual neurons really matters. And unlike for example, the chips in your phones that are highly, highly stereotyped, that’s why we can manufacture a billion of them and put them on a chip. The neurons are anything but homogeneous, they’re highly, highly heterogeneous, and I need to be able to track a large number of these tiny, tiny neurons, across,seconds, minutes, days, hours, weeks, or ultimately my lifetime. And that’s just very challenging. This is one of the tasks that we are doing here in both the brains of mice as well as in pieces of the human brain that we harvest from, from neurosurgery, from the human brain.
RS: Where are we in terms of the scientific understanding in the arc of understanding the brain at this point, both generally and at Allen Institute?
CK: At the Allen Institute it, well, let’s start out generally. Where are we with respect to consciousness in the brain? We know it’s not the entire brain that gives rise to consciousness. In fact there’s a little brain at the back of your brain called the cerebellum. It contains roughly 80 percent of all the neurons in your brain. This so-called little brain. And if you lose this, because let’s say you had neurosurgery, because you have a glioblastoma there, as happens sometimes, a type of tumor, then you might be unable to do this properly anymore or that properly anymore. You might not be able to type very fast on your iPhone or you know you might not be able to play a piano anymore. But none of these patients complain about loss of consciousness. They don’t lose parts of their visual experience or the expense of themselves or their memories or things like that. So that tells us whatever the cerebellum does, it doesn’t, it’s not involved in generating consciousness. Through such experiments, natural experiments, lesions and diseases, and studying the pathology, we now know that consciousness really is generated by, by the outermost layer of cortex of the brain called cortex, the neocortex. It’s a— it’s like a pizza. It’s really, so one of my hemispheres has its outermost shell, this pizza-like structure. It’s two to three millimeter thick, just like a pizza. It’s roughly that size 12 to 14 inch. Of course it’s highly convoluted and you’ve got two of them left and right, and they’re interconnected by two hundred million fibers, the so-called corpus callosum. And they make up my cortex, the left and the right cortex. And we know that sometimes, these fibers connecting the left and the right hemisphere have to be cut to prevent epileptic seizures from spreading from one to the other. And then it turns out as far as we can tell there are now two conscious entities, there are conscious minds in, in that one skull. You know, the one can speak, it’s typically the left, and then the other one can hum or can sing, can alter memories, can do complicated tasks, so they, and from the point of view of one mind the other one may as well be on the dark side of the moon. Because you have no conscious experience of the other mind, although you share, you share the same skull. So this is another interesting fact we learn about consciousness from, and from neurosurgery. So then the next question is, well where precisely in the neocortex, is it all of the neocortex? Doesn’t look like it. This one theory of consciousness, the Global Neuronal Workspace Theory argues it’s really the front part, the prefrontal cortex, that’s heavily involved in consciousness. The other theory, Integrated Information Theory, says no, it looks like really consciousness is sort of more restricted to what we call a posterior hot zone, is the back of their cerebral cortex that’s really critically involved. The front of the brain is critical for things like language, for planning, for intelligent action, for reasoning, all of which are important, but that’s really independent of consciousness. And so this is one of the things that we are trying to test now, the footprints of consciousness. Are they in the front of the brain or in the back of the brain, sort of very crudely spoken. Do they involve the prefrontal cortex or are they primarily localized to the posterior temporal occipital region of the neocortex?This is a question that has a principle, given the current limitations and power of the various brain techniques like FMI and EG and MEG, we should be able to answer over the next few years.
RS: Tell me about some of those technologies that you’re using, the scanning technology, and you’re also diving into tissue?
CK: At the Allen Institute, most of us don’t study consciousness. Most of us really study the brain. Most people just solidly study the brain and certain behaviors, particular visual behaviors. So because I’m at the Allen Institute and my personal interest is to study consciousness, I've assembled a group of people who study consciousness, but that’s all in mice. That’s not humans. We collaborate.
Templeton World Charity Foundation supports research, particularly trying to test these two theories. That primarily involves testing them in humans, which has advantages and disadvantages. So in humans the idea is, we use the best current technologies to get a view of the entire human brain and its activity, using either functional brain imaging of FMRI, where you lie in these, in a magnetic scanner and it scans your brain and what it reads out is these sluggish activity of the of the hemodynamics, essentially you follow the power consumption of the brain. When one particular part of the brain is very active, let’s see, because you’re processing a visual image, then that part of the brain will be supplied with more oxygen and you and you can track that using magnetic fields.
Another technique that’s being used in this, in this showdown between these two different theories, is EG and NEG. In the one case we track the electrical activity of the brain as it’s visible from the outside. This is also known as a brain wave. We can look at various brain waves. This technique is roughly a hundred years old, EG, and then a much newer technique that is somewhat different is tracking the magnetic field. Every time this electrical activity of course there’s also magnetic activity and you can track that and at a certain difference to tracking electrical activity. And then there’s another technique that’s being used here, which relies on having patients where you, the doctor, because they have epileptic seizures, has to place electrodes inside the brain proper, called ECOG, to track epileptic seizures. And so you can use it with their permission, you can use the same electrode also to, to track neural activity, because now you can, you’re much closer to the brain because you’re literally on top of the brain or sometimes even inside the brain and so you’re, you have much higher resolution than being able to only look from outside the protective skull, from outside the scalp.
And so these are the dominant techniques, FMI, brain imaging, EG and MEG. This is all done in normal subjects and so-called neurotypical, typically volunteer undergraduate students that are paid for and then tracking the direct electrical brain activity in these patients.
RS: By tracking the electrical pattern, you will find what?
CK: The secret of the universe. (LAUGHS) I mean, what else?
For tracking the electrical patterns, we have two experiments, both involving a stream of images. In one experiment, these individual things that come up, they’re things like faces and objects and letters and false letter phones. Sometimes they’re relevant, because I have to detect them enough to keep track of them, sometimes they’re irrelevant, I can disregard them. And another experiment involves a sort of a video game where quickly sort of, it’s a motor visual game where I have to quickly track things, and it’s, it’s very fast-paced, it’s adjusted to every player, so they play it just at their, at the peak level of performance. You know it’s a little bit like Tetris or something like that. And in all these cases I track the footprint of when they’re actually conscious of something and where that footprint is and how strong it is and what’s the thermodynamics of that. What we’re trying to do here is titled test differing and sometimes conflicting predictions of these two theories of consciousness, Integrated Information Theory of consciousness, and Global Neuronal Workspace theories of consciousness. And to be able to ascertain where are the footprints of consciousness, are they in the back of cortex or at the, or do they essentially involve structures in the front. What is the timescale? Do they persist for as long as I see something, for as long as I experience something, or only at the beginning or only when, when something disappears? So these are some of the questions that this team is trying to pursue.
RS: Templeton has set up this adversarial collaboration. What convinced you to get involved, how important do you think adversarial collaboration is in the scientific community?
CK: Alright, so, the way most people think science happens is that scientists get together, they agree to disagree, and so they let’s see, their theory A, and their theory B, and they agree on an experiment that distinguishes between theory A and theory B. The best— the most spectacular example of that in the history of science is the expedition that was launched in 1917 towards the end of World War One by Evington in in the royal astronomer in, in England to test a prediction that Einstein had just made the previous year regarding general relativity, regarding whether a beam of light from a distant star would be deflected by the gravitational mass of the sun. There was a very clear— there were sort of three predictions. One prediction was there is no deflection whatsoever. A second was a prediction that would be a prediction following a variant of Newton’s law, and the third prediction was just general relativity. There were these three clear predictions during the solar eclipse that took place in 1918 in Brazil, that passed over where the path, the path of total solar eclipse passed over Brazil and Principi, this little island off the coast off of West Africa. They measured a few stars and their deflection as the, as the sun was obscured and they came back and said yes, clearly the evidence is in favor of Einstein theory of relativity and the rest is history. Now, that almost never happens today anymore. Why, well partly because today theories, you know what we are studying, the brain, or in general biomedical sciences, is vastly more complex than a, you know, than a beam of light from a distant star deflected by a sun. Vastly more complex in terms of, in terms of how many degrees of freedom there is. And so typically what happens, somebody has an idea. You know I think this part of the brain is involved in that thing, in that particular action, tests it, finds evidence in favor of it or typically will only publish if they find evidence in favor of their theory. And then at the very end, towards the end of the discussion of the paper reporting this finding, they said, well clearly this contradicts the prediction of Professor Y, who has a theory W. So at the very end they sort of quickly, there’s really no engagement only at the very end they’re sort of as an aside. Well clearly these results are totally con, contrary to what that other theory predicts. What almost never happens is that people of these two different proponents of some particular, of some particular phenomena that they seek to explain, come together, are you ahead of time. Ok, we disagree. Let’s think of an experiment that we can do with the outcome if it goes one way, supports my theory, and if it goes another way it supports your theory. That’s much more difficult. A, you have to, because now you have to think okay, are we actually really talking about exactly the same phenomena. Very often it turns out in psychology for instance, people are not exactly talking about the same thing. But then I also have to commit myself. I really have to commit myself and say ok, if this is a finding, my theory will be wrong. And then people of course, now they want to, they want to game it. Well, it’s not so simple, it’s really only if this and that if that also holds. So it partly now is social engineering. You have to get people to agree and to kneel down, ok, are you committed to this proposition, in these conditions, this is tested and that’s found and your theory is invalid. And so that’s not easy.
When I was first approached by the Templeton Foundation, Dawid Potgieter, the program officer who sort of pioneered this, he approached me more than two years ago, he said, “Are you interested, I want to do this adversarial collaboration, in particular in the domain of consciousness.” Which is of course my specialty and I immediately said yes, let’s do it at my institute. Because one of the functions of my institute is to pioneer new modes of new sociology, particularly in the neurosciences. And so we hosted the first meeting that now led to this series of experiments over a year and a half ago at the Allen Institute, where we got proponents from these two different theories together, Global Neuronal Workspace and Integrated Information Theory of consciousness. In one room for two days. There were like 20 people and we argued and argued and argued. Now part of the problem was, as I just alluded to, how do you define consciousness. And we had pretty massive disagreement. But then we also said ok, let’s agree on certain things where we all agree. We all agree consciousness, no matter what it is, and what your particular ontological commitment is, there is a physical substrate in the brain, ok, we can all agree on that. Now let’s say, what does your theory say with respect to the physical substrate, the footprints and what other theories say. And there we did find agreement, and then it still took us a year to hammer out these two experiments, in a joint protocol, and where we agreed ahead of time okay, these are the analyses we’re going to run. These are the subjects we’re going to pick. These are the exclusion criteria, because that’s another way how people say well, you know, I don’t want to count this, this person, you know, who doesn’t have good data, I’m not using him for my data evaluation. Well, then it gets very fuzzy. I mean, what’s good data? Is good data the data I hope is there? You know, so that can be very biased. So we try to get rid of all of it by agreeing ahead of time, let’s agree on some objective criteria based on which we’ll accept data or will reject it. Let’s find independent labs, so we have three different techniques: FMI, EG, MEG, and a technique that measures the brain response in patients. For each experiment let’s get two different labs, two because we want replicated, these labs are not directly associated with either one or the other theories. So they have no bone in the fight as it were, let’s get them to do these experiments, we’ll then upload it to the cloud, everybody as part of this collaboration has access to this data, so we can then mine it systematically, and we can try to reproduce each other finding and all the data and all the metadata and all the notebooks, all of that we, will be made accessible to everyone. So everything is totally aboveboard. So this started at the beginning of the year. We got ready, so the biggest deal, it’s just really very difficult because everybody has a slightly different scanner or different EG system, maybe from a different company, that has a slightly different way of placing its electrode, etc. So you have to agree on all of these to make sure that everybody is really exactly measuring the same thing. So we got through that stage but then COVID hit. So right now we are sort of in limbo. We have to wait until we can get back, in particular until we can do human experiments again.
RS: What are some adversarial collaborations or adversarial challenges of the future?
CK: Scientists who are very diverse, as a huge ecology of science, in terms of size, from super national labs like CERN, from national labs, to big labs, to small labs, to one individual professor and her postdoc — adversarial collaboration absolutely should be part of that. It’s difficult as I just mentioned because you have to do a lot of social engineering. You have to do a lot of meetings to adjudicate, to get common standards, to get everybody to agree on the same thing. Particularly, you know, scientists. They’re clever, they’re quick, they want to say well, I found a better way of doing it, so I’m quickly going to hack this way up. This clever way of doing it. Well if you don’t share this with everybody else and if everybody else doesn’t agree on it then we can’t use it, because we want to use it, everybody has to do exactly the same thing because otherwise we’re doing different analysis and it’s not surprising when we get different results. By its nature that’s, that is just a challenging thing to do. We all want particular scientists that very independently want to do their own things.To get them all to agree to do common things, itcan be done, I mean there are many large labs where this is done every day like LIGO, the experiment that discovered the gravitational wave. But it requires a lot of meetings, a lot of committees, a lot of agreement across the entire group. So it’s not easy. It should be done more often than it is currently.
RS: Some are skeptical, what level can Templeton realistically expect?
CK: Oh, I believe there’s no reason that we should not achieve our stated goal which is doing these experiments, understanding the condition and coming to some conclusion. What I’m not sure of, given that the theories, at some level they’re very abstract and then you have to map them onto the enormous complexity of the brain, and there are many, many, many different ways of doing this mapping, that just because we have one finding that rules out for this particular condition, one theory, it’s unlikely to be the death of that theory because then you can probably say ok, this prediction was wrong but I can change this prediction in my model and so now have a slightly different model, right. So is it going to vigorously rule out any one theory permanently, probably not. But I see no reason why this experiment shouldn’t succeed and shouldn’t succeed in a beautiful high profile paper that was given a lot of attention because of this very systematic way of testing it and making all the data and all the meta data available, which again is very rarely done. So then everybody else who disagrees with me can say ok, here’s a data, try it yourself, and maybe you can do further data mining because we’ve only done one particular analysis, there are all these additional analysis that, that you can do and then, you know, because the data is free, you can, you can publish it yourself. And in that sense, no matter what the outcome is on theories of consciousness, it’ll be an experimental, very successful protocol, at least that’s what I believe.
RS: The future— What will it take, where are we going?
CK: Well, we are going into a world where we can manipulate the brain at a better and better scale. We have not, what we haven’t yet seen yet a dramatic breakthrough, particularly in the diseases of consciousness, like schizophrenia, anxiety disorders, depression, right. These are all diseases that massively affect our, our tone and our conscious experiences. So ultimately one hallmark that will have the right theories if the theory leads to prediction that then ultimately leads to better ways to better therapeutics to actually, to actually help people. And of course my personal desire is to understand how consciousness fits into the general scheme of the universe, and what it tells us about who else is conscious. I believe consciousness is much more widespread than we believe in the Western canon. Who else has it and how does this affect our relationship with other organisms in this grand universe that we find ourselves in?
RS: When you say you think it is much more widespread, what are you pointing to?
CK: Well, I guess I should come out: I’m a proponent of Integrated Information Theory, I really like it, it has many things going for it. One of its consequences is an idea that I’ve always in my heart been very sympathetic to, that consciousness is not only a property of sort of you and me and great apes and other charismatic megafauna you know, the grand Eagle, and the orcas killer whales, and cats and dogs. But that many animals and perhaps most animals that are capable of behavior. Like a bee may feel something, there’s absolutely no reason given the complexity of the behavior of a bee, given the fact that its brain, it’s ten times denser than your or my brain and only has a million neurons, but it’s much denser. There’s no reason to rule out the fact that it doesn’t feel like something. Now the bee won’t feel fat, and won’t worry about conscious— or about you know, the weekend, but it may well feel like something to just have drank some gold nectar and to fly home in the warm sunlight. And in that sense I am much closer to — Sikhism. This ancient philosophical belief that psyche, the soul, is found maybe not everywhere but much more widely. And then if you think about it well, what about a Protozoa? So if you look at a single celled organism, that has vast complexity, no one has ever come even close to assimilating all the molecular processes that operate in a single cell. And so maybe it too, feels an itsy bitsy little bit like something when it’s alive, and when it’s some membrane disintegrating and it is dead it won’t feel anything anymore. And so that would suggest that their consciousness is much more widespread than we are raised to believe particularly in the western world.
RS: Talk about technology and computers.
CK: We’re certainly going to project consciousness onto our computers whether they have it or not. Take Alexa. You can ask Alexa, are you conscious? Ask her. Go ahead. She’ll tell you. Now that’s of course different, whatever she tells you doesn’t tell us whether that’s actually the case. So now again you need a theory. You can’t just rely on intuition for computers like Alexa because their evolutionary history is radically different from the evolutionary history that you and I or even I and a bee share. Their hardware is radically different. They’re engineered, they’re not evolved. So everything is very different. So there, our intuition completely breaks down. And then you need a theory. And some theories like Global Neuron Workspace, yes, consciousness ultimately is a computation, it’s a particular type of algorithm that makes information globally available. It doesn’t really matter to what extent that’s implemented on, on squishy, on squishy sort of wet weather, or on, on hardware. So yes, in principle Alexa or a suitably equipped robot will feel like something. Other theories like Integrated Information theory say no, at least digital computers as they’re currently built will never feel anything. So you can get a digital computer that simulates the entire human brain. In fact, it has a 1 to 1 simulation of every brain, every neuron, in a human brain assimilated on this digital computer. And this simulation, the simulacra will wake up and will see I’m conscious but it’s all a deep fake, it doesn’t feel anything, it feels as much as today as my refrigerator feels, maybe like nothing. For the same reason that you can run software on a computer that simulates the black hole at the center of our galaxy. But you do not have to be afraid that you’ll be sucked up by the software simulating the gravitational force. Why, in both cases the same reason, because simulation is not the same as having causal power. The causal power to, to twist space time requires heavy mass, a lot of mass. A billion, you know, you know, a million solar masses. You can’t simulate that. Same thing.
Consciousness ultimately is about causal power that can’t be simulated, it has to be built. So the right question is: could you build machines ultimately that have human level consciousness? And in principle, yes. There’s nothing magical, supernatural about the human brain, as I said before it’s a piece of furniture like any other. And so one way to build it is to integrate brains with computers to integrate some of their causal power. Then in principle you could get sort of experience in artificial systems.
RS: Would you worry about something like that in the future being built?
CK: I would, I do worry about super AI, about super intelligence, but there I’m really worried about behavior, not about consciousness. In fact you can make the argument that we are the most powerful species on the planet. There are almost a billion of us because we are the smartest, you know, not because we run faster or we are wiser, but we are the smartest. So now we’re getting ready to build, to put that smart into a machine, is that really long-term a good idea? I think no. It is going to happen for various reasons. But that’s all behavior. Intelligence is ultimately about behavior. And so that doesn’t tell you anything about consciousness. And all the dangers of a runaway super AI, don’t, none of those dangers depend on it being conscious. It doesn’t really matter whether the thing that’s trying to take over the world because it wants to maximize return on investment, you know, and take us to nuclear war because that maximizes the index is conscious or not. It doesn’t need to be conscious in order to have some, to have behavior. In fact you could make the opposite argument, you could make the argument that if machines were conscious, they could then develop what we have, which is empathy. The emphatic response. I can feel sorry for you because I know what it is to be lonely. I know what it is to be at home and to wait for that phone call that's not coming. I know what it is to have pain. But that requires consciousness and if you have consciousness you can feel with others. If you have no consciousness, it makes no point of feeling for the other, because there is no feeling for you. So I would make the point that in fact, endowing computers with consciousness might actually be less threatening to the long term survival of humanity than super intelligence without consciousness.