Q&A with the man who seeks to unlock the secrets of humanity
Mapping the brain could unlock the secrets of human individuality, but with billions of neural networks per cubic millimeter, the task ahead for scientists is immense. One of these scientists is Sebastian Seung, tipped as a rising star of neuroscience. He trained as a physicist at Harvard University, but later joined MIT as professor of computational neuroscience. In his first book, Connectome: How the Brain’s Wiring Makes Us Who We Are, he argues that our individuality lies in our connectome, the complex map of our neurons.
By Ian Sample / The Tapei Times Observer Monday, June 18, 2012
Ian Sample: What does a connectome look like?
Sebastian Seung: A connectome is a map of the brain as a network of neurons. If you want a mental picture, think of the flight maps you see in the back of airline magazines, except replace each city with a neuron and each flight by a connection between neurons. Now scale up the map to include 100 billion cities and 10,000 flights per city. That would be your connectome.
Sample: How do you map a connectome?
Seung: One way is to cut really thin slices of brain and image every slice. You then have to identify every neuron and trace its path through the chunk of brain. You need high-resolution images and automated ways to analyze them. I am proposing a long-term goal, to get to a point where we can do an entire human connectome, but that requires decades of exponential innovation. Just by doing small pieces of brain, we will learn a tremendous amount about how the brain works.
Sample: What might it tell us?
Seung: The brain is behind the really big questions we have: Who am I, what is my identity? What is that based on? If memories are encoded in connectomes, your personality might be in your connectome. If that is the case, that is then the basis of your uniqueness as a person. That is why I have this slogan: “You are your connectome.”
When some people think of a wiring diagram for the brain, they imagine an electronic device and a diagram that never changes, but the connectome changes when you have an experience and it is been thought for a while that is how the trace of a past experience is stored in your brain. The hypothesis is that minds differ as connectomes differ.
Sample: How does the brain change?
Seung: I define the “four Rs” of connectome change as: reweighting, reconnection, rewiring and regeneration. Reweighting is a change in the strength of an existing connection. Reconnection means the creation of an entirely new connection or the elimination of an old one. Rewiring is growth and retraction of branches of neurons. Regeneration is the creation or elimination of entire neurons.
We know these four kinds of change happen. What we do not know is how exactly they are related to learning and memory. We do not know the extent to which they continue in adulthood and old age. Reweighting occurs all your life, that is why you are a lifelong learner. Reconnection was thought to stop at adulthood, but now there is evidence it keeps going. All the debates are over rewiring and regeneration.
Sample: You talk about “connectopathies.”
Seung: This is the theory that certain psychiatric disorders are caused by miswiring of the brain or connectopathies. They are pathological patterns of connections. One of the most puzzling aspects of certain psychiatric disorders is the lack of a clear neuropathology.
If you take the brain of a person who died from Alzheimer’s or Parkinson’s disease, it is clear that the neurons are degenerating and dying. We do not have cures, but at least we can see something wrong with the brain. But if you look at someone who had schizophrenia, you cannot see a distinguishing neuropathology. That is why for many years people denied that schizophrenia was a brain disorder. The neurons might be healthy, but maybe they are connected in an abnormal way and we just cannot see it.
Sample: Where do you start?
Seung: The bottleneck becomes analyzing the images. Suppose we can image a cubic millimeter of brain in two weeks. To trace the neurons through those images manually we estimate would take 100,000 years. For 1mm3.
The challenge is to speed up that process. One of the ways is to automate the process, but artificial intelligence (AI) is not perfect. It can speed people up, but it cannot replace them. Even if AI does 99 percent of the work, that would only reduce 100,000 years to 1,000 years.
Sample: So you crowdsource the job.
Seung: My lab is launching a citizen science project, which is trying to recruit volunteers on the Internet to interact with the AI to map out connections. Angry Birds consumes 600 years of human attention a day.
I joke that if we could make mapping the connectome a game that was 1 percent as fun as Angry Birds, we would be done with a cubic millimeter in two days. We need a lot of people to do that cubic millimeter, but our AI is based on machine learning, so the people who interact with it are also teaching it.
That is the only way we will ever do a cubic millimeter of brain and scale up to do the entire brain. We need to get on this curve of exponential innovation.
http://www.humanconnectomeproject.org/
 |
| Uber brain nerd Sebatian Seung. |
Ian Sample: What does a connectome look like?
Sebastian Seung: A connectome is a map of the brain as a network of neurons. If you want a mental picture, think of the flight maps you see in the back of airline magazines, except replace each city with a neuron and each flight by a connection between neurons. Now scale up the map to include 100 billion cities and 10,000 flights per city. That would be your connectome.
Sample: How do you map a connectome?
Seung: One way is to cut really thin slices of brain and image every slice. You then have to identify every neuron and trace its path through the chunk of brain. You need high-resolution images and automated ways to analyze them. I am proposing a long-term goal, to get to a point where we can do an entire human connectome, but that requires decades of exponential innovation. Just by doing small pieces of brain, we will learn a tremendous amount about how the brain works.
Sample: What might it tell us?
Seung: The brain is behind the really big questions we have: Who am I, what is my identity? What is that based on? If memories are encoded in connectomes, your personality might be in your connectome. If that is the case, that is then the basis of your uniqueness as a person. That is why I have this slogan: “You are your connectome.”
When some people think of a wiring diagram for the brain, they imagine an electronic device and a diagram that never changes, but the connectome changes when you have an experience and it is been thought for a while that is how the trace of a past experience is stored in your brain. The hypothesis is that minds differ as connectomes differ.
Sample: How does the brain change?
Seung: I define the “four Rs” of connectome change as: reweighting, reconnection, rewiring and regeneration. Reweighting is a change in the strength of an existing connection. Reconnection means the creation of an entirely new connection or the elimination of an old one. Rewiring is growth and retraction of branches of neurons. Regeneration is the creation or elimination of entire neurons.
We know these four kinds of change happen. What we do not know is how exactly they are related to learning and memory. We do not know the extent to which they continue in adulthood and old age. Reweighting occurs all your life, that is why you are a lifelong learner. Reconnection was thought to stop at adulthood, but now there is evidence it keeps going. All the debates are over rewiring and regeneration.
Sample: You talk about “connectopathies.”
Seung: This is the theory that certain psychiatric disorders are caused by miswiring of the brain or connectopathies. They are pathological patterns of connections. One of the most puzzling aspects of certain psychiatric disorders is the lack of a clear neuropathology.
If you take the brain of a person who died from Alzheimer’s or Parkinson’s disease, it is clear that the neurons are degenerating and dying. We do not have cures, but at least we can see something wrong with the brain. But if you look at someone who had schizophrenia, you cannot see a distinguishing neuropathology. That is why for many years people denied that schizophrenia was a brain disorder. The neurons might be healthy, but maybe they are connected in an abnormal way and we just cannot see it.
Sample: Where do you start?
Seung: The bottleneck becomes analyzing the images. Suppose we can image a cubic millimeter of brain in two weeks. To trace the neurons through those images manually we estimate would take 100,000 years. For 1mm3.
The challenge is to speed up that process. One of the ways is to automate the process, but artificial intelligence (AI) is not perfect. It can speed people up, but it cannot replace them. Even if AI does 99 percent of the work, that would only reduce 100,000 years to 1,000 years.
Sample: So you crowdsource the job.
Seung: My lab is launching a citizen science project, which is trying to recruit volunteers on the Internet to interact with the AI to map out connections. Angry Birds consumes 600 years of human attention a day.
I joke that if we could make mapping the connectome a game that was 1 percent as fun as Angry Birds, we would be done with a cubic millimeter in two days. We need a lot of people to do that cubic millimeter, but our AI is based on machine learning, so the people who interact with it are also teaching it.
 |
| The super-popular Angry Birds game. |
That is the only way we will ever do a cubic millimeter of brain and scale up to do the entire brain. We need to get on this curve of exponential innovation.
 |
| Image from the Human Connectome Project. See link below. |
http://www.humanconnectomeproject.org/
No comments:
Post a Comment