Brain-Like Computer Closer to Realization

Neurons
Image: PLoS Biology via Wikipedia

(PhysOrg.com) -- Almost since computing began, scientists and technologists have been fascinated with the idea of a computer that works similarly to the human brain. In 2008, the first "memristor" was built, a device that is designed to behave in a manner that mimics the junctions betweens the neurons in the brain. However, until recently, the memristor was just a device. Now a group at the University of Michigan, led by Wei Lu, has demonstrated that the memristor can actually be used in computing. Their findings were published in Nano Letters: "Nanoscale Memristor Device as Synapse in Neuromorphic Systems."

The Michigan team used the same sorts of materials that we have readily available when building : silicon and silver. The team joined two metal electrodes at their crossing using the and silver mixture in order to reflect behavior of the synapses in the brain. This set-up is believed to provide a way to store memories as the memristor learns new firing patterns. New Scientist reports on the importance of this computing memristor:

In the brain the timing of electrical signals in two neurons affects the ease with which later messages can jump across the synapse between them. If the pair fire in close succession, the synapse becomes more likely to pass subsequent messages between the two. "Cells that fire together, wire together," says Lu.

The Michigan device exhibits the same behaviour. When the gap between signals on the two electrodes was 20 milliseconds, the resistance to current flowing between the two was roughly half that after signals separated by 40 milliseconds. "The memristor mimics synaptic action," says Lu, adding that the next step will be to build circuits with tens of thousands of memristor synapses.

However, it has yet to be proved that memories are being stored by this set-up, and stored information hasn't been retrieved. But the fact that a team of scientists has managed to create a situation in which the is mimicked by human-developed technology means that a brain-like computer could be closer to realization.

Brain-Like Computer Closer to Realization
A memristor is a two-terminal electronic device whose conductance can be precisely modulated by charge or flux through it. Image credit: Nano Letters, DOI:10.1021/nl904092h

Explore further

Memristor chip could lead to faster, cheaper computers

More information: *Paul Marks, "Electronics 'missing link' brings neural computing closer," New Scientist (2010). Available online: www.newscientist.com/article/m … omputing-closer.html .
*Sung Hyun Jo, et. al., "Nanoscale Memristor Device as Synapse in Neuromorphic Systems," Nano Letters (2010). Available online: pubs.acs.org/doi/abs/10.1021/nl904092h .

© 2010 PhysOrg.com

Citation: Brain-Like Computer Closer to Realization (2010, March 17) retrieved 21 August 2019 from https://phys.org/news/2010-03-brain-like-closer.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
0 shares

Feedback to editors

User comments

Mar 17, 2010
The learning process described here is called: http://en.wikiped...n_theory . It's not very effective. If they got back-prop learning working, that would be impressive http://en.wikiped...rceptron but regardless of the learning algorithm, we still don't know how the global brain network operates, which, IMHO, is what's holding up the whole show

Mar 17, 2010
There is a whole boat load of stuff holding up the show, but you have to admit that this is really cool.

Mar 18, 2010
Really interesting, thou I bet my money on computing a brain in a computer is more effective - with more plasticity - rather than making it physically with silicon. But I might be wrong.

Mar 18, 2010
Perhaps one could better mimic brain neuron/synapse development by using a method that produces an "infant" network with many thousands of connections where some how (here is where the magic is) the synapse strength could be increased or decreased by some controlled chemical action. But what I was thinking when i saw this article and this one http://www.physor...84.html, that perhaps the two ideas (contained in this article and the one i reference) could be made to work together nicely. In our brains from what I understand it starts as a raw cell with many dendrites connecting all over the place and the actions of the brain either causes these connections to strengthen or break. Perhaps doping the ends strands (from the pasted article) with molecular compounds, the transmission of electric signals along the strands could cause this hardening/breaking to occur to transform it into a developed "brain cell".

Not mind you that i know much about this type of science! ;-)


Mar 21, 2010
This sounds a bit like the NOMFET (http://en.wikiped...NOMFET), that some French guys recently have invented. Take a look.

May 03, 2010
I'm sorry, but by no means is backprop a better option than Hebbian learning for a biologically-based system. Backprop can be a highly effective learning algorithm, but is purely ARTIFICIAL in its neural representation of how real neurons work. Something more efficient (hardware-wise) is surely going to be needed than either Hebbian or backprop can currently provide.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more