Harnessing chaos to build faster computers

Chaos has its own internal order — and now a University of Florida researcher is harnessing those hidden patterns to create a first-of-its-kind, highly adaptable computer that could be as much as a hundred times faster than conventional computers. Despite their name, chaotic systems have underlying patterns, or “disorganized organization,” and are found everywhere in nature, from heart fibrillations to weather patterns, said William Ditto, a UF professor of biomedical engineering who is developing the computer prototype with funding from the U.S. Navy. NASA also has expressed interest in the project.

“Chaotic systems are common to biological systems. They’re common in every physical system we build,” he said. “We thought, ‘Is there some way we could take advantage of that behavior being everywhere?'”

Because these systems obey some simple rules, their behavior is predictable for a short time into the future, Ditto said.

“There are patterns that weave in and out, kind of like a symphony,” he said. “But it’s not like an orchestra tuning up — it’s not random. And it’s not periodic, like bang-bang-bang-bang. So the study of chaotic systems is to understand how those patterns interact, why those patterns exist and for us, how to actually exploit those patterns.”

But further into the future, they become unpredictable. That’s because chaotic systems also are highly sensitive to very small changes — known as the “butterfly effect” — and two very similar starting points can lead to two very different trajectories. That sensitivity to small changes is why researchers had not previously considered chaos as a useful computational principle, Ditto said. Conventional computers are designed to follow simple logical directions. For example, if condition A exists, the computer should follow command B.

However, Ditto and his team found that short-term predictability does have computational advantages. It enables a chaotic element, such as an electric circuit or a brain neuron, to mimic the “logic gates” used by conventional computers and then to surpass them. Unlike traditional circuits, chaotic elements can easily and rapidly flip from one function to another.

“Instead of having all these different complicated circuits, I can create every circuit the same, and every circuit can perform different computations, depending on what I want,” Ditto said. “And then it just finds an available connection — like a college student looking for a date.”

Ditto and colleague Sudeshna Sinha of The Institute of Mathematical Sciences in Chennai, India, have turned that seemingly far-fetched theory into fact. They already have constructed a chaotic circuit that can perform every logic operation that a normal computer circuit can — and more.

Researchers have been searching for the next generation of computers for decades, Ditto said. Scientists have experimented with a range of “soft” computing methods, from quantum computers to DNA computers to neural networks, each offering possibilities for faster, more efficient computing that sidesteps the traditional linear approach and is one step closer to how the human brain computes.

“We’ve now got 50 years of building (computers) and engineering them, but they have certain fundamental limitations,” he said. “They use too much energy, are way too slow and are way too fussy.”

“It’s very exciting,” said Michael Shlesinger, program director of the U.S. Office of Naval Research’s Nonlinear Dynamics division, which has contributed about $800,000 so far to Ditto’s chaotic-circuit work. “(Chaotic computing) allows you to use different types of elements that are more biological, because our brains have evolved to solve problems like pattern recognition, make cognitive decisions — things that are very hard for standard computers to do.”

Chaotic computing is just in its infancy, but it has great potential, Shlesinger said. “We’re interested in a computer that could make unmanned vehicles more autonomous, with better decision-making properties,” he said. “With its potential to approach the flexibility of biological systems, it may be better at enhancing the autonomy of unmanned vehicles. It would have to be programmed, but it may be able to self-organize to handle data in ways that you wouldn’t program it today.” The possibility of packing just a few highly flexible devices into a space-limited satellite also has NASA interested, Ditto said.

“If you send five different devices up on a satellite, you have to have five different circuits, five different power supplies, and then you have to have usually three versions of redundancy,” he said. Instead, one chaotic circuit can do the work of all those devices, from determining how much power is passing through the system to communicating with the ground.

“It’s like a chameleon architecture,” he said. “We just slightly jiggle it to change its behavior.”

Now that he has a working chaotic circuit, the next step in developing the chaos-based computer is to create an operating system and the programming language for it, Ditto said. Though it may be several years away, he envisions a time when a single operating system — named “chaOS” — may run both a regular computer and a chaotic computer.

“It’s something very new, and we’re going to have a lot of fun watching it grow,” Shlesinger said.

From University of Florida

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.