D-Wave Defies World of Critics With 'First Quantum Cloud'

Geordie Rose has a Ph.D. in quantum physics, but he's also a world champion in Brazilian jiu-jitsu and a Canadian national champion wrestler. That may seem like an odd combination, but this dual background makes him the perfect fit for his chosen profession. Rose is the CTO and founder of D-Wave. He calls it the world’s only quantum computer company, but the world's quantum computer experts don't agree with him. The result is a nearly 10-year fight to prove each other wrong, and at least in some ways, Geordie Rose is winning. ”I’m not okay with losing at anything," he says, "at all.”
Image may contain Clothing Shirt Apparel Human Person Man Face and Head

Geordie Rose has a Ph.D. in quantum physics, but he's also a world champion in Brazilian jiu-jitsu and a Canadian national champion wrestler. That may seem like an odd combination, but this dual background makes him the perfect fit for his chosen profession.

Rose is the CTO and founder of D-Wave. He calls it the world’s only quantum computer company, but the world's quantum computer experts don't agree with him. The result is a nearly 10-year fight to prove each other wrong, and at least in some ways, Geordie Rose is winning.

”I’m not okay with losing at anything," he says, "at all.”

The quantum computer is the holy grail of tech research. The idea is to build a machine that uses the mind-bending properties of very small particles to perform calculations that are well beyond the capabilities of machines here in the world of classical physics. But it's still not completely clear that a true quantum computer can actually be built.

There's no shortage of quantum physicists, mathematicians, and computer scientists who say that D-Wave's machine is no quantum computer. "D-Wave's technology has been an enigma, in a negative sense," says Greg Kuperberg, a math professor at the University of California, Davis. Just this week, physicists at the University of New South Wales unveiled new research they hope will eventually lead to the first quantum computer.

But Rose keeps fighting. In May, D-Wave published a paper in the influential journal Nature that backed up at least some of its claims. And more importantly, it landed a customer. That same month, mega defense contractor Lockheed Martin bought a D-Wave quantum computer and a support contract for $10 million.

The critics have been so vociferous in large part because Rose isn't shy about promoting his company. But that's just the way he is. Rose likens D-Wave's quantum computers to the Large Hadron Collider, the world's biggest particle accelerator. "They're the largest programmable quantum systems that have ever been built by a long shot," he says. And his latest pitch is that D-Wave is on verge of unveiling the world's first quantum cloud. That's right, quantum-computing-as-a-service.

The Machine that Defies Common Sense

The quantum computer was first proposed in 1985 by British physicist David Deutsch. And it defies common sense.

The computer on your desk obeys the laws of classical physics, the physics of everyday life. But a quantum computer seeks to exploit the physics associated with very small particles, such as atoms. A classical computer stores "bits" of information in things like transistors, and each bit has a value of either 0 or 1. But a quantum computer stores information in "qubits," and these are represented by some sort of quantum system, such as the spin of an atom's nucleus. An "up" spin indicates a 1, for instance, and a "down" spin indicates a 0.

The trick is that a quantum system can exist in multiple states at the same time. It's called the superposition principle of quantum mechanics. At any given moment, the spin of a nucleus can be both up and down, holding both a 1 and a 0. And if you have two qubits, they can hold four values at the same time: 00, 01, 10, and 11. You can see where this is going: If you build a large enough quantum computer, it's exponentially faster than anything in the classical world. It would be fast enough to, say, instantly break the encryption algorithms that protect communication and electronic commerce on today's internet.

But first you have to build it. And that’s not easy. Qubits must be completely isolated from the classical world. If you interact with one, it collapses into a single state -- i.e., it changes into just an ordinary bit. The challenge is to find a way of connecting up lots of qubits without breaking them.

Rose says he's found one. Others say he hasn't.

D-Wave's 512-qubit chip, code-named Vesuvius. The white square on the right contains the quantum goodness.

Photo: D-Wave

Man With a Quantum Plan

Geordie Rose didn't invent his quantum computer before launching D-Wave. He decided to start a company before he knew what quantum computing was. "I'd never heard of it," he says.

While working on his doctorate, Rose took a class on entrepreneurship, and as the class brainstormed business ideas, another physics student suggested a quantum computing company. He read a book on the subject and told his professor that's what he wanted to do. The prof showed him how to write a business plan and wrote him a check for several thousand dollars. D-Wave was born.

The Multiverse

The D-Wave computer manipulates energy to solve optimization problems. A set of quantum bits, or qubits, has an energy landscape -- a series of energy peaks and valleys. Loading an optimization problem into the computer amounts to mapping the problem onto this landscape. "You're making a one-to-one correspondence between a mathematical optimization problem and the actual physical energy of the chip," says D-Wave CTO Geordie Rose.

Once the problem is mapped, the computer operator turns the equivalent of a volume knob all the way up, which puts the chip in a deep freeze. This shifts the computer from the world of ordinary physics to the weird realm of quantum mechanics. In this state, each qubit goes into superposition, meaning it represents 1 and 0 and everything in between at the same time.

British physicist David Deutsch, the father of quantum computing, describes this state as creating parallel universes. According to the parallel universes theory, using a 512-qubit computer splits our single universe into 2 to 512 parallel universes where everything in all the universes is exactly identical except the values of the bits on this chip. “You can think of it as opening up a little rift in the multiverse that sits right on your chip,” says Rose.

To get the answer to the optimization problem, the operator turns the volume knob back down, the chip warms up and a particular energy landscape emerges. Translating the landscape into a string of ordinary bits gives the result. In practice, the operator runs the computer many times to get a sample or the best approximation of an answer.

During the company's early years, it sponsored quantum computing research and harvested the intellectual property. At the height of this phase, it was supporting 15 different research organizations. "We were probably the leading repository for all knowledge about quantum computing in the world at that time," Rose says. "All of that intellectual property was coming to us exclusively."

Then, in 2003, D-Wave settled one particular method, something called the adiabatic model of quantum computing. It was a bold choice. Virtually everyone else in the quantum computing community favors a different method -- what's known as the gate model. "Over the years," says Rose, "I've come to strongly believe that [the gate model] is just simply a really rotten idea."

The first version of the D-Wave machine was released in 2007, and the company billed it as a 16-qubit quantum computer. The current version has expanded to 128 qubits. The machine includes a chip that contains 128 superconducting circuits, and each circuit is a microscopic loop of continuously flowing current. The chip is encased in a high-tech refrigeration unit that cools it to almost absolute zero, and at that temperature, the circuits enter the fuzzy realm of quantum mechanics where the current flows both clockwise and counterclockwise at the same time.

Basically, you feed the machine a problem, and it uses a set of algorithms to determine the probability that a set of qubits will emerge from the quantum realm in a particular pattern when the computer’s temperature is raised. That pattern -- a string of 1's and 0's -- is the answer to a problem.

Wall Street, Monte Carlo, and Beyond

D-Wave's computer is designed to solve what are called combinatorial optimization problems. The classic example is figuring out the most efficient route for a traveling salesman going to multiple destinations. There’s no mathematical shortcut that computers can take to solve combinatorial optimization problems. They have to use brute force: Simply check all possible combinations. The trouble is, the number of possibilities explodes exponentially with the problem size. For example, if you have six destinations, there are 64 possible combinations. If you have 20 destinations, there are 1,048,576 possible combinations.

D-Wave’s next-generation computer is designed to handle problems with as many as 512 variables. In theory, that lets you solve problems involving two to the 512 possible combinations, and a problem of that size is beyond the reach of any classical computer that could ever be built. "It's bigger than the number of atoms in the universe," Rose says. "It doesn't matter how big a supercomputer you make."

Combinatorial optimization problems are everywhere, and solving them is big business. In addition to route planning, they include image recognition, genome sequence analysis, protein folding, scheduling, and risk analysis. Airlines, pharmaceutical companies, and financial firms deal with combinatorial optimization problems every day.

D-Wave's CEO Vern Brownell has a history with such problems. He was the chief technology officer at Goldman Sachs from 1989 to 2000. The core of financial computing is calculating risk and analyzing portfolios, tasks that involve sampling from huge data sets to estimate the nature of the data and simulate how it will evolve. The Big Data software that carries this out is called Monte Carlo simulation.

The world's financial firms have hundreds of thousands of computer cores dedicated to running Monte Carlo simulations, and according to Bronwell, D-Wave's computer is well-suited to the task. "Think about a world where all the Monte Carlo simulations that are done by all the investment banks in the world could be replaced by a handful of chips running in a 10- or 15-kW refrigerator," he says.

The F-35 fighter is a mighty complex system.

Photo: Lockheed Martin

Software Complex

The investment banks haven't bought one yet. But Lockheed Martin has. And its aim is a little different.

Lockheed Martin makes some of the most complex systems in the world -- things like the F-35 Joint Strike Fighter. On average, half the cost of developing a new complex system at Lockheed Martin is system verification and validation, and the major component of this is software verification and validation. The concern is that as it builds ever more complex systems, this cost will rise. “I have quipped in various board meetings that maybe we ought to give the airplanes away and sell the software maintenance contracts,” says Ned Allen, Lockheed’s chief scientist.

Instead, Allen decided that the best way of debugging the company’s software was to throw out the computers. The great mathematician and founding figure of computer science, Alan Turing, showed that it is impossible to eliminate all errors from software. But the laws of physics are another matter.

Allen’s idea is to write what amounts to a compiler to translate digital code into analog code, and then run the analog code on an analog computer of some sort. The code is directly connected to the physics of the computer, unlike digital computers that involve a logical abstraction. “The most fundamental analog computer that I know of is called quantum thermodynamics,” he says.

For help finding an appropriate quantum system, Allen turned to an authority on the subject: Daniel Lidar, a professor of electrical engineering, chemistry, and physics at the University of Southern California. And Lidar pointed him to D-Wave. “I was aware of the controversy of whether the D-Wave machine was a quantum machine or not,” Allen says. “So I was sort of reluctant to go deal with them. Lidar convinced me that I should take another look.”

Allen sent D-Wave a sample problem to run on the D-Wave system: a 30-year-old chunk of code from the F-16 aircraft. The software has an error that took a crack team of Lockheed Martin engineers several months to find. Six weeks after sending the code, Allen visited D-Wave and was given a demonstration that included identifying the software error. “I was just bowled over,” says Allen.

He then convinced Lockheed Martin’s management to buy a D-Wave computer and install it in a lab at USC’s Information Sciences Institute. Lockheed Martin and USC split time on the machine, and Lockheed Martin’s access is via a secure network. The machine came online at noon on December 23, and the company now has 50 people working on it.

The Quantum Computer That Isn't

But even Daniel Lidar -- the man who set up the sale to Lockheed -- says the D-Wave machine is not a "universal" quantum computer. He calls it a "special purpose optimization engine." Lidar is the man who actually operates the D-Wave machine as the scientific director of the USC Lockheed Martin Quantum Computing Center, and one of his first priorities is to build tests that will show -- conclusively -- whether the chip is quantum or classical.

In the field of quantum computing, the closest thing to unanimity is the oft-repeated assertion that it is likely take decades to develop a practical quantum computer -- if it can be built at all. Having someone claim to be build one -- someone whose work is cloaked in a veil of intellectual property and trade secrets -- has been more than some researchers can tolerate.

D-Wave burned its bridges with the scientific community in 2007 when it unveiled its first machine but didn't offer any peer-reviewed research to explain the thing. According to UC Davis' Kuperberg, the press conference misrepresented not only quantum computing but theoretical computer science in general. "Since then, discussions with D-Wave or about D-Wave have tended to veer toward the question of D-Wave's credibility," Kuperberg says.

On some level, Rose understands the criticism. "It's the same basic human reaction that everybody has to something that someone says that sounds outrageous," he says. "If you're a naturally skeptical person, which most scientists are, you're going to think it's bullshit." But he also says there are mismatched expectations. D-Wave's technology is a niche within quantum computing, and it differs substantially from the reigning quantum computing paradigm. From Rose's perspective, critics are using the wrong measuring stick to judge D-Wave's technology.

But the war of words is settling into something resembling a truce. "Things have gotten better," says Wim van Dam, a professor of computer science and physics at the University of California, Santa Barbara and longtime D-Wave critic. "People got tired of arguing in the aggressive way it was done a few years ago. People are talking to each other. Experimentally they seem to be doing solid work."

But van Dam still doubts that what D-Wave possesses is a quantum computer. "If I were to predict the future, their systems would [turn out to] be classical systems. But it could be that they're using a kind of technology to implement what I think is classical computation that is very different from the silicon technology we use nowadays." In other words, D-Wave could turn out to be wrong about what it thinks it has but still end up with a useful product: one that is faster, more energy-efficient or in some other way better than today's silicon computer chips.

Quantum Computing in the Heavens

Lockheed Martin connects to its D-Wave machine over a network, and this presages the next step in D-Wave's strategy: the quantum cloud. The company still plans to sell individual computers, but it will also offer quantum-computing-as-a-service. "That capability is already there. We just haven't productized it," says D-Wave CEO Brownell. "Eventually there will be a quantum cloud and there'll be a classical cloud, and all of your significant computing problems will be addressable in those two clouds."

D-Wave is looking to work with a cloud partner, but Brownell declined to identify candidates. One possibility is Google. In 2009, Google researcher Hartmut Neven cowrote a paper with D-Wave researchers that showed that the D-Wave system might be useful for searching across large image databases.

Further out, D-Wave is looking to develop machine learning technology and get into the data analytics field. The company has found that a range of machine learning problems are at the core really optimization problems. "A lot of the world is faced with the problem of how you draw inferences from huge amounts of data and can you do that in an automated way without human involvement," says Brownell.

Ten years on, Geordie Rose is still fighting. And he'll fight for another 10. "We’ve still got a long ways to go," he says. "It may take another 10 years before the history textbooks will know what to say about this."

Update: This story has been updated to clarify Daniel Lidar's postion on the D-Wave machine.