Project simulates human brain to enhance computer power

A new project to simulate the human brain could make computers more powerful and less energy-hungry, researchers claim.

Teams from nine European countries, including the UK, are contributing to proposals for €1bn of funding to support the Human Brain Project, which could also aid medical research into neurological conditions.

The researchers plan to create a virtual human brain on a large supercomputer, using the vast amount of existing research to build up mathematical models of how it works.

The research team would be more surprised than anyone if the virtual brain became self-aware

‘It’s about modelling what we know about the brain,’ the project’s spokesman and chief proposal editor, Richard Walker, told The Engineer.

‘Neuroscientists have done a huge amount of work on the brain and there’s a huge number of papers but they tend to be extremely fragmented.

‘We want to use these to fix parameters for different parts of a single unified model. And that means we can use information from one piece of research to illuminate what’s happening in others.’

As well as creating a research tool, the project participants, led by a team at the Swiss federal technology institute EPFL, hope their work will inspire so-called neuromorphic technology — computer chips that run on analogue circuitry inspired by the brain.

This could drastically reduce the power consumption of computers, which currently use far more energy than human synapses — the supercomputer planned to run the project will be likely to use one million times the 20 to 30W of the brain.

Cutting computers’ energy use is particularly important to maintaining the trend of doubling processing speed roughly every two years.

Another aim is to make computers better at things the brain is very good at but current technology struggles with, such as recognising faces, making decisions based on incomplete information or using different parts of its circuitry when one part breaks down.

‘When we use computers they’re almost always extremely rigid today,’ said Walker. ‘A robot on an assembly is great until a piece arrives the wrong way up and the computer doesn’t know what to do but any human worker would turn it round.’

Mimicking the brain’s ability to deal with large and varied amounts of information from multiple sources could help computers to run complex networks of sensors, such as a city’s traffic-control system or within a nuclear reactor.

The technology could also be used to simulate brains for medical research, both to investigate mental and neurological illnesses and to reduce the need for testing on animals.

To create the simulation, the researchers will use mathematical models to gradually build up a map of neural pathways rather than individually measure and recreate each one.

‘We have models of the different types of neurons, we know their densities, we know how they connect to each other and we look for general statistical principles governing these things,’ said Walker.

‘We know for instance that if we take some neurons of specific shapes and place them randomly in a space they will connect in a certain way.’

Through an earlier EPFL project that has succeeded in simulating a rat’s neocortical column — the Blue Brain project — researchers were able to recreate the electrical wave patterns of the brain just by mimicking its architecture.

The team also hopes to create a visual representation of the model that shows it operating in real time, rather than having to export part of the programme from the supercomputer to another machine that can convert it to images.

Thirteen organisations, including the Wellcome Trust Sanger Institute in the UK and other groups in Austria, Belgium, France, Germany, Israel, Spain, Sweden and Switzerland, will take part in the project.

The EU has awarded the scheme €1.5m under its Future and Emerging Technologies programme to put together a proposal that will enable it to compete against five other projects for up to €1bn of funding.

Although the simulation could help scientists understand more about human consciousness, Walker said the research team would be more surprised than anyone if the virtual brain became self-aware.

He added that the computers resulting from the project would be limited to finite tasks and wouldn’t be capable of creativity.

‘We don’t see them as replacing current computers — it’s more complementary. Humans are very bad at adding up so if we imitate that on our chips they will be bad at adding up too. But we hope it will do other things our current computers can’t do.’