The path to understanding the brain: Henry Markram at TEDxCHUV

I'd like to tell you about the Human Brain Project, which is basically a CERN for brain research, to accelerate our understanding of the human brain. You have witnessed, I think, a very good example of the passion, Gregoire, the hope, and the progress of understanding the human brain, in lots of different forms. But the problem is, if you go to a neuroscientist today and you put them on the spot and say, "When are we going to understand the brain?" They'll take a step back and say, "No, no, no, no, no. It's going to take a very long time. It's going to take ages." The question we have to ask is whether we can afford to wait. Can we afford to take time to understand the human brain? Because it is already affecting 127 million people in Europe. It's already costing a trillion dollars, just in Europe. Pharmaceutical companies are giving up. They're saying that after spending billions of euros, decades, that digging into the brain is not solving the problem. They can't get better treatments. Novartis, Roche, others ‑‑ they're shutting down their neuroscience. So we should ask ourselves, "Why can't we understand the brain? Why is it so difficult to understand the brain?" So if you go to these neuroscientists that say, “No, it's going to take a very long time”, and you ask them, “Why can't we understand the brain?” What they will tell you is ‑‑ they'll tell you it's because the brain is too complicated. They'll tell you it's because we don't know enough. They'll say, "There's just too many holes in the research. We can't understand the brain. There's too many holes in the research. We don't know enough. It's too complicated. We're just going to take a very long time to understand the brain." The problem is that if you actually analyze how we do neuroscience today, there are about 100,000 neuroscientists in the world. They wake up in the morning, some of us have a great idea, and we compete for these ideas. The best idea gets a lot of money, and then they go off and they do their research, and they add a new piece of the knowledge, a new little piece of information about the brain. And in this way, we've actually generated about 5 million studies directly related to the brain. So, basically, they're saying that as these neuroscientists get up, we're generating all this knowledge. There's about five million papers. There's about 30 million papers that are indirectly related to understanding the brain. They come from the life sciences. So the question is, "How many more papers do we need to do? How many more studies? Is it 10,000 papers? Is it 100,000 studies? Is it another million studies? Are we going to solve that problem?" We have to think, "What is the real problem?" And actually, the real problem is that all this knowledge is fragmented. It is completely fragmented. Most of us do not even understand each other in neuroscience. So what we have to do is obvious. Even you know what the solution is, and that is, you have to bring this data together. But to bring this data together, we're going to need the help of most of the disciplines. We need the help of the engineers, which you see here. They have learned how to take all their data, physical data, of machines and the physics of it, and have been able to build technology. They've been able to build airplanes, cars, all kinds of machines. We're going to need the help of mathematicians and physicists and chemists and neuroinformaticians, so that we can actually look at this data, make sense of these data, analyze it and understand what the patterns of it and what the rules of this information all is. We may need the help of NASA, too. How do you coordinate all of this effort? We may need the help of CERN, because CERN has experience. How do you take thousands of people, scientists, together, and put them around the same table, and make them work together towards a single mission, right? So, what we started to do was to try to begin this process. And in the Blue Brain Project that was about 7 years ago, we started to build a facility. It's a prototype facility. It's a supercomputing facility, and it's a software initiative to try to synthesize all this data. We also have recently started to put together a very large consortium of about 256 scientists around the world, to begin putting this together and work our way systematically towards the human brain. Now, what we had to do before we could justify such an attempt is to build a piece of the brain, as a proof of concept. So the piece that we chose was the neocortical column. Now, we had to do it in a rat, or in a rodent, because that's where most of the data is. We chose, also, the cortical column because it's the most complicated part of the brain. It's the most sophisticated. So if you can build this little piece of the brain, you'll be able to build the other pieces of the brain. So with this mission in mind, we started. The first thing we had to solve was, if you want to build a piece of the brain, you have to know all the kinds of neurons. Now, this is a very big challenge. It's a difficult challenge. There's many different kinds of cells. We call it the "neurome." And in the beginning, when we started this, we had no idea how you would solve this problem. What we have subsequently found is that there are very simple rules that allow you to derive the composition of the cells. It's like finding out what's the good pasta recipe or reverse‑engineering the Coca‑Cola recipe. So this has been possible, and we can extend this further to looking at the whole brain, the composition of cells in the whole brain. We had to take all the data together that tells us about the molecules and the proteins that make neurons come alive. These are called ion channels, and we had to see how to put them together so you could automatically build different types of neurons and make them come alive. But we also had to solve one of the biggest challenges, which is called the connectome. The connectome is, "How do all the cells fit together?" Your brain has about a thousand trillion synapses, and every one of these synapses, which is these little red dots you see here, are positioned in a very precise location. We didn't know how to solve this problem when we started, but by taking all the pieces of information that were known from all different areas of research, we actually discovered very simple rules that today allow us, mathematically, to connect these neurons and account for all these synapses. We think we can extend this to the whole brain, which means that you will be able to connect these neurons and account for the positions of these trillions of synapses. We had to develop the software that would allow you now to take these models and put them onto super computers and simulate the individual cells, and simulate how they communicate together. This is what it looks like. You saw it in Richard Frackowiak's lecture. This is what we call the "Unified Model of a Cortical Column." Now, the magic of this is that it actually holds ‑- it's like a library. It holds 100 years of knowledge. At a certain level of detail it holds 100 years of knowledge about this piece of the brain. It's an integrated view of all this fragmented information that was available. Right here it was available, we could put it together. But the power of it is, today we can use this model to make literally thousands of predictions that would take decades of experimental research to do. That's what we mean by accelerating our understanding of the brain. By putting these models together, we can actually leapfrog decades of research that would allow us to see how the brain is put together as an integrated system. We have to go from mouse to man because we absolutely have to learn how to carry with us all the information that we've learned. All the knowledge we've learned from animals into understanding the human brain. Because, as you heard previously, the human brain is not a place where you can just access anything that you want. You have to learn the principles of biology, and use them to reconstruct the human brain. Now, you could ask once you've built this, “How do you understand what it is doing?” Well, one of the things that we're doing is to connect this to a virtual avatar that is behaving in a virtual environment. As it is behaving, and as it begins to learn and make decisions and to perceive, we will be able to X‑ray across all the levels of organization. If we've built models with molecular level detail or genetic level detail as well as all the other areas, you'll be able to see across the entire spectrum of what's happening in the brain, and delineate, decompose all the steps that are involved in cognition. This is probably the only method that I can think of today that would give you this multilevel view of how a behavior originates. Now, once we've done that, we'll be able to ask some of the really big questions that we all would love to answer. For example, "How does the brain capture all the laws of physics so that it can build your little universe?" If your little universe is not correct, then you're going to be visiting professor Frackowiak to find out what is wrong. There are lots of people that have diseases because they build a wrong world. By being able to do this, we can not only see how this world, your universe comes crumbling down, but we'll be able to identify where it's crumbling down. So you would be able to use these models like an X‑ray system, or zoom system in diagnostics, to be able to pinpoint where is the most likely place where these diseases could go wrong. Then you could run simulations to try to understand what's the best treatment. Now, to make this all happen, we also have to do a lot of other computer science innovations. We have to change the way supercomputers work. We have to make them actually interactive instruments, so you can literally fly through the brain at different levels and explore how it is functioning. Now, if the excitement of understanding the brain and the seriousness of diseases is not enough to make you feel that we have to understand it now, today, as soon as possible, well, then maybe the money is. There is a lot of money in this, right? Because the brain is the ultimate technology. It doesn't transmit information the way that technology does today. It sends a hint, and the receiver imagines the answer. Use that or discover that principle ‑‑ mobile phones, Internet, will never be the same again. It doesn't store information the way that common technology today does. It probably stores the equivalent of petabytes or exabytes of data. Discover that and the way that data is managed in the world will be completely changed. It doesn't compute the way that a computer computes. It actually creatively guesses at the answer. OK? It is highly energy‑efficient. You heard from Pier's lecture, 20 watts. The equivalent supercomputer would require a gigawatt. That's a billion dollars of electricity a year. So the equivalent computer would take a gigawatt. It is highly reliable. As you also heard, you could lose 80 % of your cells in a certain part of the brain and you won't even notice it. You could lose half your brain before somebody notices you have dementia. Highly robust, highly reliant. If we discover those secrets, you will be able to build technology that is incredibly reliable. Understanding the brain will change the technology of the 21st Century. So if the diseases are not enough of a motivation, there is another motivation. And in fact, this is right around the corner. The technology to put brain circuits onto chip are there. There's been a race, in the U.S., in the East, and in Europe, to build silicon chips that can receive brain circuits. And they're ready. They've been developed already for the past 15 years. They're just waiting for the blueprints. They need the blueprint circuits that you can print on there so that these will become the next generation of what we call neuromorphic processors. These will radically change the way computers work today. And we see that they're going to hybridize, so you will have combined conventional computers with neuromorphic computers. Now, this is the consortium that is being put together, and we believe that this is going to be one of the most unprecedented experiments of putting a group of multidisciplinary scientists together. If it all goes well, we'll start in 2013 and the agenda is 10 years. So we are actually not waiting for this, in Switzerland. We are beginning. You may have heard that there is an announcement. There will be a new building, the first in the world, dedicated to actually bringing all the fragments of information together. But this is going to be a unique building because it is not just about research. It is about engaging the public. This building is aimed at ensuring that the public, you, can go on this adventure with us because understanding the brain will empower you and it will completely change society. Thank you. (Applause)