SourceWhy we're building a €1 billion model of a human brain
11 February 2013 by Jessica Griggs
The Human Brain Project has just won a €1 billion research prize. Its director Henry Markram says the initiative will unify our understanding of the brain
Your project aims to recreate the human brain in a supercomputer. Why?
We want to reach a unified understanding of the brain and the simulation on a supercomputer is the tool. Today you have neuroscientists working on a genetic, behavioural or cognitive level, and then you have informaticians, chemists and mathematicians. They all have their own understanding of how the brain functions and is structured. How do you get them all around the same table? We think of the project as like a CERN for the brain. The model is our way of bringing everyone, and our understanding, together.
What do you hope will come of the project?
There are three goals. As the model becomes more accurate and behaves more like a brain, we could couple it to a robot and see the robot learning. Then we'll be able to trace the chain of events from molecules to cognition.
Second, we plan to collect data from hospitals around the world to search for biological signatures of disease. We believe this will give us a new classification of brain disease, which can be used to diagnose people objectively - not solely based on their symptoms.
And finally, we want to build neuromorphic computers, which would have processors that can learn, mimicking the way the brain does.
"Once we've modelled a rat brain, then we update the rules and gradually move towards a human model"
How close are you to recreating a brain?
We've developed an algorithm to work out some of the rules for how to reconstruct the brain. We can now make thousands of predictions that when tested are accurate, and we can reconstruct small brain circuits of up to about 1 million rat neurons. Now we've got to connect those groups of neurons to make brain regions, then connect the regions to make a computer model of a whole brain. Once we've modelled a rat brain, then we update the rules and gradually move towards a human model.
Could the model be too detailed?
People are afraid of detail and complexity. But a drug is a molecule. It doesn't hit a cell, or the brain, it hits other molecules. There's no reason why we shouldn't try to explain how all the molecules are interacting with one another. Sheer numbers do not equal complexity. One hundred billion neurons is numbers, not complexity.
What about criticisms that it drains funds from more readily achievable neuroscience?
That is an incredibly short-sighted view of how a big science project impacts the rest of the field. Take the Human Genome Project - you can criticise them for all the promises they made at the time, but we would be in the dark ages now if we hadn't done that project. It has benefitted every biologist and geneticist on the planet today.
Once completed, could the simulated brain ever become conscious?
When we couple the model to a robot, the robot will behave, and we'll see this in the way its neurons are firing. Does that mean it's conscious? That's a philosophical question - and an unresolved one.