by Henry Markram
Today, brain disease affects some 164 million European citizens at an annual cost of many hundreds of billions of euros. Yet despite huge investment in research, there are still very few brain disorders whose causes are fully understood, and very few drugs that do more than relieve symptoms. In the past, 95% of brain research was funded by pharmaceutical and biotech companies. Today they are pulling out. The number of new patents for brain-related drugs and brain implants is dropping every year. “Brain-inspired” computers face similar problems. The human brain is a highly flexible, very fast, massively parallel system that consumes just 20-30 Watts. Computers and robots with a fraction of these capabilities would transform 21st century society. Yet we still cannot build computers that match the human brain on even the simplest cognitive task.
Figure: Simulated neural network. The brain, with its billions of interconnected neurons, is without any doubt the most complex organ in the body and it will be a long time before we understand all its mysteries. © Blue Brain/HBP.
What is missing is a true understanding of how the brain works. A century of neuroscience has accumulated vast amounts of data and knowledge but it is badly fragmented. Yet we still lack an “integration strategy” – a way of putting them together.
This is the goal of the Human Brain Project. The project proposes to build a new research platform for modelling and simulating the brain, a new kind of scientific instrument– the equivalent of a new telescope. With the new instrument, the project will pool experimental and clinical data from all over the world and integrate it in detailed computer models, flexible enough to incorporate the rising flood of results from new research. It will look for patterns in the data and use them to understand basic rules of brain design. And it will use these rules to predict hard-to-measure features of the brain. Given that many rules are valid across different areas of the brain and different species, it will not need to measure every single neuron, synapse, molecule, and gene. In fact, it will model the human brain without using invasive methods at all.
Every time we use the instrument, we will learn more about the brain, and everything we learn will help us improve the instrument, until it gives us a full and comprehensive picture of the way the brain works. The new instrument will make it easier to design brain prosthetics for disabled people. Medical researchers will use it to investigate the causes of brain disease, to produce better diagnostic tools, to find and test new therapeutic leads, and to develop better treatments. The end result: lower health costs and better lives for patients.
Engineers, for their part, will take models of brain circuitry developed with the platform, simplify them, and use them to build new kinds of computing systems and robots. Even if the first machines only have limited capabilities, everything we learn about the brain will make the task easier.
Ten years ago none of this would have been thinkable. What makes it feasible now is enormous progress in supercomputing. Already, the world’s most powerful machines can perform a million billion calculations per second; by 2018 that will reach a billion billion - more than enough to build realistic models of the brain. The instrument proposed by the Human Brain Project will exploit these capabilities. Of course it will not resolve every problem in neuroscience, medicine or technology. What the project will do is integrate what we know in a completely new way. If we are to understand the brain and develop effective treatments for brain disease, if we are to build completely new Information Technologies, based on knowledge of the brain, this is the essential first step.
Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland