System simulates fault tolerance of the brain

University of Manchester to build brain style computer.

Scientists at the University of Manchester are to build a computer that mimics parts of the human brain to improve the reliability of computers.

The project, which has been awarded £1m of government funding, will simulate neurons firing in a brain.

Project leader professor Steve Furber says the computer will be built to simulate one million neurons, a fraction of the 100 billion neurons in a human brain.

‘Simple insects such as the bumblebee have sophisticated control algorithms with about that many neurons,’ said Furber. ‘This is applicable to real-time control problems – mobile robotics being an obvious example.’

The research will examine the potential of a neuron-style system – more fault-tolerant than present computer architectures.

‘In the brain neurons die at a rate of about one per second, yet things keep working as normal,’ said Furber. ‘This fault-tolerant characteristic is of interest to engineers who want to make computers more reliable.’

Furber says the project is not aimed at creating artificial intelligence (AI) in the short term.

Ian Pearson, BT futurologist, said: ‘Eventually machines could be developed with neural networks that have the potential to be intuitive enough to fit in with a human workforce.’

Copyright © 2010 Computing

brainfaultoddwareofsystemthe