Monday, June 27, 2011

The big plan to build a brain

By Roger Highfield

A scientist is trying to use the world’s largest computer to mimic our mind , writes Roger Highfield.

Your brain has capabilities unmatched by any machine. It consumes only a little energy, equivalent to that used by a 30-watt light bulb. Yet it can do amazing things – it can learn and think, as well as predicting the consequences of its own actions, expressing itself in language and understanding the language of others.

Now the owner of one such brain, a South African-Israeli professor called Henry Markram, is preparing to build an instrument that can simulate this magical machine down to the molecular level. The Human Brain Project, at the École Polytechnique Fédérale de Lausanne on the shores of Lake Geneva, intends to build a carbon copy of the brain by around 2023. Rendered in software rather than flesh and blood, it will draw on the help of 13 institutions across the EU – as well as the largest computer man can build.

You might think the project is a monumental example of hubris. But Markram’s motivation is straightforward. A quarter of the population is affected by one form of brain disease or another, and drug companies are giving up much of their research into the brain because they have found it too complex and costly. Also, the 200,000 scientists currently studying separate tiny facets of the brain are generating a dizzying array of fragmented information. "In all," he says, "there are around five million papers relevant to understanding the brain and its diseases, and around 100,000 more are being published every year."

The aim of the Human Brain Project is to create a tool that can turn all this work into a unifying model of the brain, using real data to hone simulations, which can then guide experiments. Markram has spent the past six years building his proof of concept, "Blue Brain". By 2008, his team had simulated the first microcircuit, which consists of a unit of 10,000 brain cells in the cerebral cortex, the thin layer where the most interesting functions reside. Today, the Blue Brain can simulate about 360,000 neurons, which is about the size of a single brain region in a small rodent. To simulate a human brain, he will need to simulate around 100 billion neurons.

This sounds daunting, but Markram has figured out both the structural rules that determine how the brain is built, which rely on a diverse range of nerve cells, and the functional rules that govern how the cells are able to interact in such a way that brain circuits are unchanged by growth, or minor damage. These explain how we can still get wiser, even though we lose 10,000 neurons each day.

One fascinating finding is that each microcircuit can spontaneously generate a rhythm of activity, called a gamma oscillation. This was once thought to be a property of interactions in the whole brain that could generate a unified perception, even consciousness. "Our finding does not mean that a microcircuit is conscious, but that it it can spontaneously generate this emergent property. Using our current Blue Brain model, we can systematically search for the mechanisms behind these oscillations." Using the simulation, he can also do virtual experiments to figure out which ones will be most worthwhile to carry out using the real thing.

Markram is leading leads one of six consortia which are competing for grants from the European Union that are worth a billion euros over 10 years. In the summer of 2012, he will find out if he has been successful. The funding would largely be invested in building an exascale supercomputer (one capable of billion billion operations). And the fascinating thing? It still wouldn’t be as smart as any one of us.

• Roger Highfield is the Editor of ‘New Scientist’

© The Telegraph Group London 2011