A Darwinian approach has struck gold (Image: Rex/Shutterstock)
Move over, microchip. A random assembly ofcan perform calculations normally reserved for neatly arranged patterns of silicon.
Traditional computers rely on ordered circuits that follow preprogrammed rules, but this strategy limits how efficient they can be.
“The best microprocessors you can buy in a store now can do 1011 operations per second and use a few hundred watts,” saysof Twente University in the Netherlands. “The human brain can do orders of magnitude more and uses only 10 to 20 watts. That’s a huge gap.”
To close that gap, researchers have tried buildingthat do calculations without their innards having been , but so far no one had found a material that could reliably perform .
Now, van der Wiel and his colleagues have enabled a clump of gold grains to handle bits of information in the same way that conventional microprocessors do.
The team used gold particles about 20 nanometres across. They laid a few tens of these in a rough circle, each about 1 nanometre from its nearest neighbours, and surrounded them with eight electrodes. The computations happened when they applied just the right voltages to the cluster at six specific locations. Then the gold effectively forms a network of transistors that lacks the strict order of connections in a regular microchip, allowing them to run calculations using less energy.
Nothing about the particles told the researchers what the voltages should be, however. They started with random combinations of voltages and learned which were the most useful using a, a procedure that borrows ideas from Darwinian evolution to home in on the “fittest” ones.
It compared many sets of voltages, discarding those for which the unit’s behaviour made no sense, creating slightly different versions of those that seemed promising, and trying again. In effect, the clump of gold particles was.
New path to logic
As a proof of concept, the algorithm found the voltages that transformed the system into any one of the six “logic gates” that are the building blocks of conventional computer chips. It even found the combination for a higher-order logic unit, which can add two bits of information. “This shows that you can get to calculating ability by a completely different route,” van der Wiel says.
The clump of particles has to be cooled to just 0.3 degrees above absolute zero, but making the particles smaller would allow the working temperature to rise. Van der Wiel says there is no reason the approach couldn’t work at room temperature.
Van der Wiel hopes the research will lead to specialised processors that can solve problems such as, which are difficult for computers that do calculations one after the other. If a whole clump of grains is doing the calculation, then they are operating in parallel – much like neurons in the human brain, which is especially good at such tasks.
He could well be right, saysof the University of Alberta in Edmonton, Canada. “The physics is there, but of course you still have to demonstrate it.”
Another hurdle is creating devices that can handle more complex calculations, which could be done either using more nanoparticles or more electrodes. “The electrodes are probably more important, so that you can have more inputs and outputs.”
This entry passed through the Full-Text RSS service – if this is your content and you’re reading it on someone else’s site, please read the FAQ at fivefilters.org/content-only/faq.php#publishers.