skip to content | Accessibility Information

Computers which mimic the human brain

Discovery offers prospect of record computing speeds

image to illustrate 'Computers which mimic the human brain'

A SIMPLE invention by scientists in Manchester could hold the key to building the world's first neuron computer.

Two mathematicians, Jon Borresen and Stephen Lynch believe it is possible to design superfast micro-processors which mimic the configurations in the human brain.

If their models are correct, the circuitry would produce world-record computing speeds at a fraction of the power input.

The pair present their theoretical research today in the open access journal PLOS ONE.

Modelling neurones

Their binary half adder circuit, which is protected by three pending patent applications, is already attracting interest from leading figures in the supercomputing industry.

But Jon and Stephen, who work in the Novel Computation Group at Manchester Metropolitan University, admit the potential breakthrough in micro-processing didn't come from research into computing at all: "We were modelling the behaviour of neurones in the brain, just observing really, when it struck us that certain configurations could perform binary computations," said Dr Borresen.

"We thought, if these binary interactions could be replicated in electronic circuits, it had the potential to revolutionise silicon chips."

Brilliantly simple

The new circuit is brilliantly simple: two oscillators connected in such a way that the addition of each extra oscillator results in a doubling of processing power, such that (2+n) ~ 2^n (two to the power n) processing power, resulting in massive savings in power and components.

Following a series of theoretical tests, Jon and Stephen pitched their theories at Daresbury STFC and at the Rutherford-Appleton Laboratories in Oxford, where they were encouraged to obtain patent protection for their ideas.

But it was a recent trip to the New York offices of high-tech firm HYPRESinc which provided the real breakthrough.

"We had been researching how we could build one of these computers using biological neurones, CMOS, optical resonators or Josephson Junctions (JJs). HYPRES had already been using JJs since in commercial circuitry and we think they are the key to building the new computer," added Stephen.

Multiple applications

HYPRES recently announced they have built the world's fastest (20GHz) hybrid Arithmetic Logic Unit, using 8,000 JJ oscillators. The Manchester Metropolitan team calculate they could produce the same processing power with just 50.

The work may also have applications in neural biology, providing a much needed assay test for neural degradation -having particular implications for the design and testing of new drugs for conditions such as Alzheimer's and Parkinson's.

"We started this as an abstract exercise in mathematics, and we've ended up on the brink of theoretically building the world's first neuron computer."

Dr Peter Roberts of Marks & Clerk LLP said: "The vast majority of patent applications that I write are for incremental improvements of existing technology. However, Jon and Stephen's idea is an entirely new technology, and writing the patent applications has been absolutely fascinating," said Dr Roberts.

'Oscillatory Threshold Logic', invented by MMU mathematicians Drs. Jon Borresen and Stephen Lynch, presents an interesting opportunity to use oscillatory devices as basic components of future computers. It combines neuron-like dynamic behavior and digital logic principles to achieve high energy efficiency in information processing," said Dr Elie Track, Sr. Partner at HYPRES, Inc.

The research is published on November 16, 2012 in a paper entitled 'Oscillatory Threshold Logic' in the journal PLOS ONE. http://dx.plos.org/10.1371/journal.pone.0048498.

For further information, please contact Manchester Metropolitan University Press Officer Gareth Hollyman on (+44)161 247 3406 m. 07748 111322 g.hollyman@mmu.ac.uk

Go to Index