3 years

Neural Networks

Recreating the Brain

Hartosh Singh Bal turned from the difficulty of doing mathematics to the ease of writing on politics. Unlike mathematics all this requires is being less wrong than most others who dwell on the subject.
Page 1 of 1
Scientists in Switzerland have embarked on an ambitious plan to reverse engineer the human brain, molecule by molecule
Unlike the molecule by molecule simulation being attempted by the Blue Brain team, neural networks used in artificial intelligence are simplified models of the actual processing that takes place in the brain. The work assumes that biological systems have features that may not be relevant to the functioning of the network. Such networks have been extensively used in artificial intelligence and are principally used for speech recognition and image analysis

Is it possible to reverse engineer the human brain molecule by molecule? The answer, surprisingly enough, is that we might already be on the way. On 21 April, Henry Markram, keynote speaker at the Science Beyond Fiction conference at Prague, told the audience that his team had been able to recreate the functioning of a real brain through a detailed molecule-by-molecule computer simulation of a slice of the brain. The simulation, he said, “starts to learn things and starts to remember things. We can actually see when it retrieves a memory, and where they retrieved it from, because we can trace back every activity of every molecule, every cell, every connection and see how the memory was formed”.

Markram is director of the Blue Brain Project at the Ecole Polytechnique Federale de Lausanne (EPFL) in Switzerland that, in 2005, embarked on an ambitious plan to reverse engineer the brain with the help of IBM.

The first step of the Blue Brain project involved recreating the neocortical columns that make up the grey matter in our brain. Each neocortical column is 0.5 mm wide by 2 mm high and contains about 10,000 neurons connected in an intricate, but consistent, way. This column is repeated across the cortex millions of times, and it is only the density of these columns that differentiates the human brain from that of a rat.

The project has used one of the world’s fastest supercomputers to simulate a single neocortical column of a rat. This was done by virtually building the column neuron by neuron, and testing its output against the output of a real column in a rat’s brain.

According to Markram, the adding up of millions of such columns could lead us to understand the emergent properties of the brain, perhaps even of consciousness. All it would take, he says, is greater computing power. “It’s a matter of whether society wants this. If they want it in ten years, they’ll have it in ten years,” he added.