Alan Evans says available computer power is at a point where it can allow for a more comprehensive and realistic model of what’s actually going on in the brain.Christinne Muschi/The Globe and Mail
There are seven billion human brains on the planet, but even the best minds in neuroscience aren't really sure how they work. How does a network of a hundred trillion neural connections generate thoughts and feelings and motivate action? An international team of researchers is taking a crack at finding out by building a brain of its own.
Boosted by a massive funding commitment from the European Commission, announced on Monday, the Human Brain Project is a $1.6-billion effort that will draw on a growing body of research about brain structure and function and aim to translate it into a multilayered supercomputer simulation. While it is by no means the first effort to approximate the brain "in silico," the scope and scale of the project is beyond anything attempted to date. Skeptics question whether it can succeed.
Co-ordinated from Lausanne, Switzerland, the project is expected occupy about 200 researchers at dozens of institutions around the world for a decade or more. Among them is Alan Evans, a professor of neurology at McGill University and an expert in brain imaging at the Montreal Neurological Institute. He spoke with The Globe and Mail's science reporter, Ivan Semeniuk, on the art of brain-building.
Why do scientists want to simulate the entire human brain?
If you're interested in understanding how a particular circuit in the brain works, you have to explore how information flows through that circuit with all its many inputs and outputs. You often can't predict how it's going to behave unless you build a model and test it. You can make that same statement at multiple scales, right up to the level of the whole brain. The concept isn't new. But now the available computer power is at a point where it can allow for a more comprehensive and realistic model of what's actually going on in the brain.
What can you do with the simulation?
Once you have a working model of a normal brain then you can start to play around in any direction you like. You can remove a connection and see what happens without having to do invasive surgery on animals. You can study the underlying causes of neurodegenerative diseases and psychiatric disorders, which are perturbations of normal brain function.
There's no blueprint for the human brain, so where do you start?
We start by gathering information from different areas of brain science. At the gross scale we can look at neuroanatomy and we know a lot of the fibre pathways that link different regions of the brain. That's what will give us the broad strokes – the basic wiring diagram of the brain. What that won't do is give us the fine structure. So then you step down to a smaller scale to model the connections within a particular region, like the visual cortex, and see if you can get the simulation to reproduce the way the visual sub-system behaves. Different groups within the project will be focusing on different subcomponents of the entire brain. The objective is to build up in complexity to something like a rat brain and then on up to a human brain.
To what extent will the simulation be "thinking"?
The goal is to model the human brain, so that will require us to be able to interact with this entity in a way that is "human-like" in some ways. It's not going to be like science fiction – like having a conversation with a robot and thinking its human. But the project will go some way toward testing whether behaviour and cognition are properties that simply emerge from the way the brain is organized.
What do you expect to learn?
My sense is we will learn a lot of things are that are not immediately obvious. To me a simulation is the ultimate expression of everything you think you know about a particular system. Where it becomes interesting is when you test your knowledge by running through the simulation and it behaves differently than you predicted. Then you've learned something about how the brain works that wasn't apparent from traditional observation.
How will your lab contribute?
Two ways. We'll contribute high-performance computational infrastructure and database technology– ways to manage and process the immense quantities of data on the brain that will be collected. And we've also been working on a high-resolution three-dimensional map of the brain that can serve as an anatomical template onto which useful information about different brain regions can be overlaid.
What is the biggest challenge the project faces?
Organization. Like the brain, the project needs to be more than the sum of its parts so that all of this information is gathered and assembled in a coherent manner toward the higher purpose.
This interview has been edited and condensed.
Editor's note: The number of neurons in the human brain is estimated to be 100 billion. The number of connections among those neurons is somewhere between 1,000 and 10,000 greater, or approximately 100 trillion, not 100 billion as incorrectly stated in a previous version of this story.
Ivan Semeniuk