Transcript for Sebastian Seung uncut

 

Jim Fleming: Well, I think maybe the best place to start is with the word connectome. Can you tell me; what is a connectome?

Sebastian Seung: You could imagine a connectome as a map of a network, and I'd like to have people think about the back pages of airline magazines. You know, those airline route maps? If you replace each city by a neuron and every flight between cities by a connection between neurons, then you would have something that looked like a connectome, although not that many cities are in those airline maps so it might be the connectome of a simple animal like a worm. Your connectome would be the equivalent of a hundred billion cities and thousands of flights leaving every city.

Fleming: I get confused, I confess, by those small scale projects in the back of the airline magazine. What you're talking about is massively more complex than that. This is a truly remarkable project that you're engaged in then, isn't it? To try to establish a human connectome.

Seung: Well, I don't think about it as a project because really it will be the work of generations I think to get there. It's like defining a grand challenge for humanity. To map the entire brain of a person at the neuronal level. Right now, we're focused really on small level chunks of brains. Those of human brains would be great, but we tend to do our work mostly on small chunks of animal brains.

Fleming: And when you do the work, I watched you on a recent internet talk that you gave showing how difficult it is to even begin this project. Even when you talk about something as small as a mouse brain, you still can't look at it as a whole. You have to go smaller and smaller and smaller even to find the beginnings of this map I suspect.

Seung: Yea, so all those connections between neurons, however they made, well you've probably seen neurons and they are these funny looking cells because they look like trees. They have very long branches, and those branches are like the wires of the brain. You could compare a brain to an electronic device in which the connections have to be made by wiring, and the amount of wiring in your brain is really staggering so in one cubic millimeter there might be several miles of those branches of neurons. Which means that…

Fleming: One cubic millimeter. Several miles of branches.

Seung: Yes. Several miles. And that would mean that inside your entire brain, there are millions of miles of branches. So it's an incredible mess of tangled up wires in there. Think about the back of your stereo but multiply that by a big number to get to millions of miles.

Fleming: When you begin to think about this, when any researcher begins to think about mapping even a small section of this, is it disturbing to you to think about the back of your stereo, to think about how complex the wiring is? I know that when I've hooked up stereos, it is not all that hard to put one wire in the wrong place and come up with a disaster.

Seung: Well, you're bringing up two points. One of them is the appreciation for the brains' complexity. We're often told that the brain is the most complex object in the universe or some other sort of hyperbole, but I really recommend that people look at the images available online to see how complex it really is inside there. You bring up another issue which is that of miswiring. It's easy to miswire an electronic device, and people have long hypothesized that brains might be wired differently. So let's consider certain psychiatric disorders like schizophrenia or autism. People have tried very hard to find evidence of pathological neurons inside the brain. Could there be dying neurons the way there are in Alzheimer's disease? And that's not typically found, so the theory has emerged that perhaps the neurons are fine, they're just wired up in an abnormal way.

Fleming: I think that that is, by itself, worthy of investigation but in order to even get to discussing that, we have to have a better understanding of, or I have to have a better understanding of the basic project. Forgive me, you said project was the wrong word. The basic investigation. And realistically, how difficult is it to begin this kind of wiring mapping?

Seung: Well, we have to have images of the brain that are extremely high resolution. High enough to reveal the brains wires, even though those wires are extremely thin. And that's provided by a means known as electron microscopy. That's the keenest microscope that we have. We also have to slice brains, these are dead brains. We have to slice them into extremely thin slices, a thousand times thinner than a hair. So if we image each of those slices and then stack them up together, we'll form a three dimensional image. A kind of virtual brain in an image. Once we have those images, then we can trace the paths of the wires and we can also identify the synapses. Those are the point in which neurons touch and also communicate with each other. If we could do both of those things, we could map connectomes.

Fleming: Bless the creation of the computer for allowing you to even think about the number of different images you have to coordinate in order to come up with the smallest possible picture.

Seung: And just to make sure, so the images from one cubic millimeter would amount to one petabyte of data. That's like a billion photos in your digital album, and that contains let's say one hundred thousand neurons and a billion connections, but the garden variety MRI brain scan you see in the newspapers, in those images, one pixel is one cubic millimeter. So the MRI blurs over all of that complexity into one pixel. So that's a huge difference in resolution between electron microscopy and the MRI machine.

Fleming: I'm finding a huge grin spreading on my face because it's so difficult to imagine the complexity of the image that you're describing. It's extraordinary to me that science is able at this point to talk realistically about that kind of imagery.

Seung: Yea, so we can acquire those images but that's something that's not unique to science. It's really across our entire society. In business now, the buzzword is big data because companies have enormous store houses of data about their consumers and their products and so on and so forth. So in all areas of life, I think we're confronted with huge amounts of data and the problem of having to make sense of it.

Fleming: And again, when you talk about this, you're talking about very small investigations at this point. You described the one complete connectome as that of a worm. I believe it's approximately one millimeter long, and that took seven years?

Seung: Longer I think, so yes. In the 1970's and 80's, the 300 neurons and 7000 connections of a worm named C. elegans, this connectome was found in its entirety but it took over a dozen years of labor.

Fleming: Has the science improved in the years since the 70's and 80's sufficient that you can imagine being able in some point in the future to do that kind of mapping of the human brain?

Seung: Well, the technologies have improved so the imaging now is much easier. It used to be much more laborious to capture the images of this worm but due to the advances of the technologies for cutting and for gathering the electron microscopic images, we could now look forward to possibly imaging a cubic millimeter of brain tissue within just a few weeks. That should be possible within two or three years let's say. So the problem, the bottleneck really, is shifting to that of the analysis of images. It's like a deluged or a flood of image data and we're drowning in it. How can we find a way to breathe and eventually to conquer the waters?

Fleming: Well I guess that brings us to the real question of all of this. Why are you doing this?

Seung: Yes, the connectome. The wiring diagram of the brain, sometimes called. There are many approaches to studying the brain, and I would divide them really into three kinds of observations. Some neuroscientists study the activity of neurons. The signals that circulate between neurons as you think and you feel. Those are the signals that are most intimately related to consciousness. Then there are scientists like me who study the connections between neurons. And then finally there are scientists who study the genes that control the processes inside neurons. So here's the three levels; activity, neurosignals, number two is connections, and number three is genes. We have to have all three of these to understand the brain. Now why do I find connections so interesting? Well, experiences are able to change your connectome. We know enough about connections to know that they're not fixed in your adult life, so the metaphor of the wiring diagram is a little misleading because an electronic device is not subject to changes in its wiring typically. But your wiring, your connections can actually change, and I believe that.

Fleming: Just to be clear, I mean the example would be again the stereo system. You're always going to connect left and right speakers to left and right outlets. The brain may say one day; I don't like that anymore. I want up and down or maybe it would sound interesting if both of them were on the same side. We're perfectly capable of changing things constantly.

Seung: That's a great metaphor, and indeed it's believed that your memories are actually stored by changes in your connectome. Every time you have an experience, it leaves some trace of the past on your brain. Possibly by modifying the pattern of connections in your head. So the effects of experience on the connectome are really important. At the same time, genes also have a powerful influence on your connectome because they guide the processes by which your brain wired itself up when you were in the womb and during childhood and so on. So the amazing thing about the brain is it's a computer or a device that wires itself up spontaneously. So the connectome is shaped by both genes and experiences, so effectively it's the place where nature meets nurture, and we've had a lot of arguments about that interaction and the twentieth century. Some people want it to say that the mind is so malleable that it's not affected at all by genes, but by now everyone acknowledges, everyone sensible acknowledges that genes have some effect. The question is; exactly what effect, and how exactly do experiences effect the connectome. These are the questions that we need to address rigorously through data and through science and hopefully no longer through polemic.

Fleming: So you have to talk about the big idea. The big idea, the connectome, is to have some understanding of the way in which the mind works, but you have to approach it in these tiny ways without letting the big idea get in your way. If you stop to think for a moment of the daily perhaps, every second changes that occur in connectome, I would imagine it would scare you away from the very idea of trying to map it. You have to assume that the active mapping a tiny portion of it, even though that will change at some point, is going to give you a bigger scope idea about how the thing works.

Seung: Well, let me emphasize that we believe that the connectome changes, but slowly. It changes more slowly than neural activity. Right? You can be angry one moment and sad the next, and that's due to the fact that the activity of neurons is ephemeral. It changes all the time. But think about the difficulty in learning something new. Think about the difficulty in quitting smoking. There's a lot of aspects of our minds that don't change quickly, and these may be the ones that are most associated with the connectome. Because really the connectome is based on a material structure. You can't really arrange material structures as rapidly as you change signals.

Fleming: At one point in your book, you talk about the oath that every member of Alcoholics Anonymous takes. Talking about the courage to face change. That is one of the things that this may help us to understand. Isn't it? That it may not just be courage, it may be having to become someone else. Having to find a way to work through how the brain works. To find new ways of being?

Seung: Yes, lately you'll find many books, self help books, written by neuroscientists who claim the authority to tell you how to fix your marriage or quit smoking and so on and so forth because of their knowledge about the brain, but I generally find that those books don't go much beyond common sense and the facts they bring in about the brain really are irrelevant to the current discussions. And the reason is that we still don't have a clear idea of what happens in the brain when you change yourself. It's as simple as that. We don't know what happens in your brain when you learn a new skill, what happens when you store new memory and so on and so forth. We have to answer those really basic questions, otherwise neuroscience can never inform the quest to change ourselves.

Fleming: So do you believe that the search for an establishment of the connectome will help you define how the brain works? Help us at least see how the brain works?

Seung: Yes. I completely believe that that kind of information has been missing and it's held back. I'm not saying that it's all that we need, but it's an important piece of missing information that has held us back from understanding how the brain works.

Fleming: Let's take a look again at one of the little pieces that you're looking at. I've read that one of your goals is to actually see a memory.

Seung: Yes.

Fleming: I think even saying that makes people think immediately of science fiction, but that's not fiction at all to you. Is it? So, what would a memory look like?

Seung: Well a memory would look like some kind of pattern of connections between neurons. The example I give in the book would be the memory of your first kiss, let's say. Let's say that's an indelible memory. You never forgot it. Maybe the appearance of your beloved, your fellow kisser. Maybe the particular room in which that happened, let's say. The music that was playing and so on. Let's assume that all of those stimuli are represented in your brain by the activity of different sets of neurons. Now one of the basic principals in neuroscience that has come to be accepted through some experiment is that when neurons are activated together, the connections between them strengthen. So the neurons that are signaling the presence of your beloved, the neurons that are signaling the music that's playing and so on, they become linked together by connections, which is the material basis then of the associations in your mind that make up your memory. All right? So I hope that's clear, the idea then that co-active neurons, neurons that are simultaneously active become bound together by stronger connections. There's empirical evidence for that. And even after the experience is over, those strong connections can remain inside your brain and store the trace of the memory. Now how does it get recalled? Well, suppose you hear the music playing one day. The same music that happened during your first kiss. That activates a particular set of neurons, but those neurons are connected to all the other neurons that were involved in that experience, and they can awaken the other neurons to activity too. So the activity spreads throughout this assembly of neurons the same way that your memory just comes flooding back. That's the theory.

Fleming: You are both articulate and eloquent about this, but I have to confess that when I start thinking about this, it's easy for me to think about all of those different things you described and realize how many of them are. It's not ten, it's not a hundred, it's not a thousand. It's some number well beyond that. Tracing those is an extraordinary task.

Seung: Yes, so we may not start out with trying to read the memory of somebody's first kiss, but we could find similar examples in animals where we could try to find the memory that's embedded there, and what I propose in my book is the memory of a bird's song. Could we somehow, by looking at the connections in the bird's brain, be able to read out the sequence in which the neurons are activated in its brain as it sings its song?

Fleming: Have you listened to a bird's song recently? They are incredibly complex.

Seung: Well, they are complex, but there are some birds that are quite simple in the sense that they only sing one song and they sing it over and over again, more like a figure skater doing compulsory figures. So those would be the kinds that we would study first. Not mockingbirds, for example, which can just imitate any song.

Fleming: That suddenly led me to wonder about the neurons involved in a double axle, which seems almost beyond belief. So you think that you, through the study of neuro-connections, the mapping of the neuroconnections, establishing a connectome might reveal to you how memory works.

Seung: That's right. How memories are stored and encoded in the brain, and I feel like that's the most direct route. If neuroscientists hypothesized that memories are stored in the connectome, then clearly the most obvious means of studying that hypothesis is to map out a connectome.

Fleming: One of the things that is difficult, I think, to describe using words alone, unable to show the listener what it is, is that you're talking about a thousand different connections. Maybe more. Maybe 998. I don't know.

Seung: Yes.

Fleming: But even when you get to that, you're not talking about straight line connections, are you? You're talking about connections that go, forgive me, every which way. The complexity of it just increases every time you add one more element.

Seung: Well, the wiring might be very chaotic looking. The wiring behind your stereo system, for example, could look extremely disorderly, but in fact there is a very logical rule that's behind what's connected to what, and so that's the hidden order that we're trying to discover in this tangled up brain wiring.

Fleming: But then of course you, and I'm sorry to stick with the stereo, but I work in radio after all so I guess it's reasonable.

Seung: That's right.

Fleming: It's one thing to be able to map the wiring of the stereo.

Seung: Yes.

Fleming: Especially a very sophisticated stereo.

Seung: Yes.

Fleming: It is quite something else though to listen to what the stereo produces.

Seung: Yes.

Fleming: And understand how that came from the wiring.

Seung: Yes.

Fleming: I mean, being able to map the brain. I wonder how much it tells you, if you could map a memory in your brain, how much does that really tell you about the effect of that memory on the person. How much does it tell you about the self?

Seung: Well in this sense though, the stereo is not a good analogy because the wiring itself of the stereo is separate from the remembered song which might be on a CD. It's in a separate medium. Right? But the hypothesis in the brain, the hypothesis, the theories that we've had for decades are that the wiring itself contains the information.

Fleming: So in the case of the CD, the brain is irrelevant. It is not something that you insert. It is all there together at the same time.

Seung: It is not a CD. That's right. So for example, in the case of the bird's song, we know that there's a particular part of the brain called HVC in which the neurons are activated in a stereotype sequence. So the first neurons are active, then the second neurons are active, then the third neurons are active in a very precise sequence every time the bird sings its song. And it's hypothesized that the first neurons make connections onto the second neurons. The second neurons make connections onto the third neurons. And it's through those connections that each group of neurons drives the next group into action. So if that's the case, then there's a very direct relationship between the way the neurons are wired together and the sequence in which they're activated during the song.

Fleming: OK. It's becoming clearer and I see.

Seung: You're asking really good questions. It's actually hard to explain many of these concepts without some diagrams, but imagine a chain of neurons. You could imagine that the neurons are wired together to be something like a chain of falling dominoes so that every neural activation sets off another neural activation inside this sequence.

Fleming: It's always difficult dealing with metaphors, but in this case perhaps we can continue that you are looking at falling dominoes, but at the same time those dominoes are stacked. They seem to relate all the time to each other, and I suppose that dominoes are not a bad metaphor there either. You can put the pips together and find some kind of creation. Anyway, this all as basic science, sounds extraordinarily exciting. You are looking long term at more difficult and perhaps more practical applications. I know you have considered the effect that this may or may not have, in helping us to understand what you've called miswiring. Brain disorders. Is that an approachable goal, do you think?

Seung: I believe it is, as long as other scientists can tell us where in the brain to look. Remember that, at this point in time we can only map very small pieces of brain at high resolution, so we have to rely on other kinds of scientists with other observation methods who can tell us; Hey look at this particular location in the brain. This is the part that's probably gone awry in this brain disorder. And that kind of study is being done a lot these days. Not only that we would like to have animal models of brain disorders, so I'm not sure you're aware, but when researchers try to develop drugs or study diseases, they often try to study that in not just in humans, but also in mice and other animals. Basically our cousins because the disease processes can be similar. Clearly, you could do that with heart disease. You could imagine a mouse with heart disease and a person with heart disease. Right? It starts to strain your imagination with psychiatric disorders. Can you really have a mouse model of schizophrenia? What's a psychotic mouse? Can you have a mouse model of autism? And yet, scientists are trying very hard. Some of my colleagues are trying very hard to create such mice. And one of the way they do this is through genetics. So it's known that certain genetic mutations that can lead to autism. There's been a lot of excitement about this in the past few years, and you can take those gene variants and put them in mice and see whether the mice display any symptoms reminiscent of autism. It's an amazing idea. Right? And indeed one of my colleagues at MIT has developed a mouse by altering one gene and he's found that this mouse tends to avoid other mice so it's kind of asocial and thereby exhibits some conditions that are reminiscent of autism. They're not really autism, but there's some relation, and the question is: can we find something different about these genetically altered mice's brains?

Fleming: And if we can, then can we make any correlation between the mice and us and it is the same way the connection is made with more mundane diseases?

Seung: Yes.

Fleming: There is a kind of over arching discussion that we haven't talked about, and that is how much of this is who we are, not just what we are. I guess what I'm asking is if you think that the research into establishing a connectome is going to tell us anything about the nature of consciousness. Our subjective experience of who we are and how we think and how we feel.

Seung: You're asking about the self and I've argued that memories, your memories may have been coded in your unique connectome and maybe other aspects of your personality and your intellect and so on and so forth, which is why I've coined the slogan: You are your connectome. It's really just a hypothesis because we don't know whether that's true, but it could be that your individuality, the key to your uniqueness really lies in that connectome. Now what's the relationship with consciousness? So here I argue that there's really two notions of self. Right? The conscious self is the one that changes all the time. The stream of consciousness. Right? We have these thoughts that are ceaselessly occupying our minds all the while that we're awake. But there's another notion of the self which is the stable self. The one that doesn't change or changes with difficulty. And actually that's really important, right. If I have a really close friend, and one day I meet them and they're behaving totally strangely, totally differently than I've ever seen them before, that's disturbing. We've come to count on the stability of people's selves. So both notions of the self are important, right? I mean, how do you think about that? Do you ever think about the stable self as well as the conscious self?

Fleming: I don’t think I had ever used those terms before, but yes, you’re exactly right. We all know what we might term the basic nature of a person that we know, of ourselves for that matter, with the variations that can change from Monday to Friday.

Seung: Yes. Yes. So the relationship between those two selves is, of course, fascinating. I would argue that…I guess I use the metaphor in the book of the “stream of consciousness”…so you think about consciousness as the water. But the connectome may be more like the stream bed because it guides what you think about the way that you act and so on and so forth. So it provides that material substrate behind the dynamic processes of our conscious selves.

Fleming: I kind of like that because if you deal with it as the stream bed, then you have a difference between what’s in the bed, and what’s running along it—could be water, could be wine.

Seung [laughs]: Ok..do you prefer wine…I don’t think it’s the weekend yet.

Fleming: When we’re talking about consciousness, perhaps..

Seung: So I’m not sure if I really answered your question. I’ve just multiplied the notion of the self to an extra self, right, that perhaps you didn’t think about before.

Fleming: I hadn’t. I do think in some ways, though, you’ve answered it, which is that the connectome, difficult and complex as it is, is a marvelous beginning, as so many things are, to an exploration of who and what we are.

Seung: Well, I’m looking forward to that voyage, and I would also add that we are really trying to recruit everyone to help us, because we have built a website—eyewire.org—which is based on electron microscopic images of the retina, the sheet of neural tissue at the back of the eye. These images were acquired by our collaborator, Winfred Denck, in Germany and we built a website around them where any person can come and learn to be an amateur neuroscientist and trace the wires of the eye, the branches of the neurons at the back of the eye to find the retinal connectome together.

Fleming: Sounds like a marvelous project. Thank you, and thank you very much for being so clear about it

Seung [laughs]: Ok…I don’t feel like I was clear, but I really try my best, at least.

Fleming: I understand, and I’m very grateful. It is not something you can explain in seven words.

Seung: Yes, well maybe when I become older and much wiser I’ll be able to do it.