TRANSFORMING TECH A quantum leap forward
D-Wave Systems’ computers harness power of subatomic particles
D-Wave Systems bills itself as the world’s first quantum computing company, with a line of computers that promise to run hundreds or thousands of times faster than conventional computers.
Google and NASA think D-Wave is on to something. They bought a new one a few weeks ago for the Quantum Artificial Intelligence Lab they run with the Universities Space Research Association.
The Burnaby-based high-tech firm has also made a customer of defence contractor Lockheed Martin. Amazon founder Jeff Bezos and In-Q-Tel — the tech hothouse for the Central Intelligence Agency — are investors to the tune of $100 million combined.
The newest version of the company’s flagship model is the D-Wave 2X, which employs a superconducting processor cooled to a temperature far lower than the coldest corner of the universe, about -273 C. While D-Wave does not disclose what its customers pay, their machines have reportedly carried a price tag up to $15 million.
Quantum computers exploit the mysterious properties of subatomic particles to solve problems of unimaginable complexity in seconds, problems that would take years to calculate with today’s conventional machines.
The heart of this revolutionary new computer is the qubit, the quantum equivalent of a bit, the basic unit of information in conventional computers. Bits hold a single value, either 1 or 0. However, qubits can be simultaneously 1 and 0, a state called superposition. Plus they can exhibit mysterious and difficultto-explain states of wavelike “coherence,” exchange information with other qubits and “entangle” with one another, a special connection in which they can influence each other ’s values even when they are not in contact.
The D-Wave 2X processor has 10300 possible states — that’s a one with 300 zeros after it — a number that dwarfs the number of particles in the visible universe and potentially unlocks unprecedented computing power.
Problematically, you can’t measure or detect subatomic states without stripping the qubits of the information they hold in the process, which makes it very tricky to learn the result of a computation and pretty much impossible to know exactly what is happening on the chip during a computation.
D-Wave has used quantum simulations to determine that what is going on inside their “black box” is not conventional computing.
If this all sounds impossibly complex, that’s because it is. It’s also the reason much of the academic community harbours doubts that D-Wave’s chips are truly exploiting quantum effects.
Quantum machines rely on principles of particle physics that are utterly foreign to our experience of the physical world, leading Albert Einstein to dismiss entanglement as “spooky action at a distance.” The renowned physicist just couldn’t wrap his head around how differently things work at the quantum level, so don’t feel bad if you aren’t clueing in immediately.
“If anyone tells you they understand quantum physics, they don’t understand quantum physics,” laughed Jeremy Hilton, D-Wave’s vice-president of processor development.
D-Wave’s approach to quantum computing is a kind of quantum-lite, based on ideas introduced by researchers at the Massachusetts Institute of Technology, says Hilton.
Rather than controlling the quantum state of atoms, D-Wave uses tiny loops of niobium made so cold that current runs in both directions simultaneously, their version of superposition.
“We were at a cusp as a company in 2004,” said Hilton.
“We had been surveying all these approaches to quantum computing and developed some ideas, but at that point we needed to pick something and go big, start to build a real technology.”
D-Wave founder Geordie Rose chose to focus on a processor that would employ a limited roster of quantum tools, essentially leaving out the ones that scientists are not yet able to manage reliably.
“Instead of waiting 20 or 30 years, we wanted to create a technology that could solve meaningful problems as soon as possible,” said Hilton.
Rather than shoot for a Holy Grail universal quantum processor in which difficult effects such as coherence and entanglement are controlled, they have opted to create a simplified version using superposition, the easiest quantum effect to control.
“If you can start to access the more robust quantum mechanical effects like superposition, you can start to build something that is applying quantum mechanics to a computation and gaining some advantage over the classical approach,” said Hilton.
“This is a way to transition into this technology … while we learn how to push further along that spectrum so that ultimately we will have a full-scale quantum computer with all those delicate effects.”
The processor — housed in a big super-cooled fridge — employs superposition in a quantum state to run thousands of calculations simultaneously and then it reveals its results by “tuning” the qubits back to a classical state, so they can be read.
It’s a quantum breakthrough, even if it isn’t a fully realized quantum computer, according to Hilton.
Even D-Wave is struggling to figure out exactly what is happening on their chip, but after some computer modelling experiments they are sure it is not conventional and possibly much more.
“What has surprised us is that our own research shows that entanglement and quantum tunnelling are available and participating in the computation,” said Hilton.
Researchers in the field are justifiably suspicious when a private enterprise appears to leapfrog the best results achieved by university-based programs. However, D-Wave’s published results have sparked renewed interest and investment in their model, which is known in the field as adiabatic computing.
In addition to new funds flowing into D-Wave, the founders of BlackBerry established a $100-million investment fund for quantum computing research, and researchers at the University of Montreal recently won a substantial grant to pursue adiabatic computing.
Hilton reckons D-Wave’s quantum computer works more like a human brain than classical computers and sees a bright future for the technology in artificial intelligence and machine learning, in which computers must tackle pattern recognition problems that humans perform with relative ease.
The D-Wave 2X shines particularly in the field of optimization problems, again because like human thought processes, getting a mostly right answer quickly is far more useful than the exact right answer next year.
In the case of optimizing thousands of delivery routes for a global delivery service or financial portfolios with millions of combinations of assets, getting back the 10 “probably” best scenarios is just as useful as a single right answer.
“You are looking for a set of good answers with different cost thresholds to choose from,” he said.
“If you can find a solution that is close to the best solution 101000 times faster — in seconds rather than years — that’s good enough.”
We wanted to create a technology that could solve meaningful problems as soon as possible.
JEREMY HILTON D-WAVE’S VICE-PRESIDENT OF PROCESSOR DEVELOPMENT