Original URL: https://www.theregister.com/2013/02/04/human_brain_project/

Euro boffins plan supercomputer to SIMULATE HUMAN BRAIN

€1.19b for in-silico experiments to build robots driven by simulated people

By Simon Sharwood

Posted in Science, 4th February 2013 04:03 GMT

The European Commission has selected the Human Brain Project (HBP) as one of its Future and Emerging Technologies and will send it up to €1.19b over ten years so it can build a supercomputer capable of simulating the human brain.

The HBP wants to build a simulated brain because we don't know enough about our grey matter. The project's web site says we lack even a “casual understanding of the way [brain] events … produce cognition and behaviour,” while more than a century of research has yielded little understanding of “how changes in the synapses between neurons help us to remember important events in our past”.

Our crude understanding of our brains makes it tricky to fix one's grey matter when things go wrong. That lack of insight translates into economic loss, because the consequences of brain disease often lead to patients ceasing work and/or requiring costly care.

The project will require fundamental research, because as its FAQ points out “With today’s technology, an exascale computer capable of simulating a cellular-level model of the whole human brain would probably consume about a Gigawatt – billions of euros worth of electricity every year.“ Another problem is that boffins behind the project think they'll need hundreds of exabytes of memory, rather more than one can expect on hand at even the largest of tech distributors between now and 2020.

The team's alternative is described as follows:

“So what we plan to do is build fast storage random-access storage systems next to the supercomputer, store the complete detailed model there, and then allow our multi-scale simulation software to call in a mix of detailed or simplified models (models of neurons, synapses, circuits, and brain regions) that matches the needs of the research and the available computing power. This is a pragmatic strategy that allows us to keep build ever more detailed models, while keeping our simulations to the level of detail we can support with our current supercomputers.”

The computers concerned will be “neuromorphic” and will use designs the project says will not use common building blocks but instead will “be based on the actual cognitive architectures we find in the brain – which are finely optimised for specific tasks. Their individual processing elements – 'artificial neurons' – will be far simpler and faster than the processors we find in current computers. But like neurons in the brain they will also be far less accurate and reliable. So the HBP will develop new techniques of stochastic computing that turn this apparent weakness into a strength – making it possible to build very fast computers with very low power consumption even with components that are individually unreliable and only moderately precise.”

Such designs, the HBP says, will require massive parallelism and use “semiconductor substrates in new ways that allow them to perform individual operations orders of magnitude faster than their digital counterparts.”

It's hoped that work will not only enhance our understanding of the brain but also help to develop technologies that leapfrog current limitations in computing, especially the slow growth in computing speed of processor cores.

The project will run for a decade, centred at the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland, with collaborators across Europe and North America. One of the project's aims is to create a robot controlled by a neuromorphic brain. Nothing could go wrong with that, of course … ®