I’m standing in front of a $70-million (and counting) machine named Dragon, which one day could be able to solve problems that would stymie even a universe-sized conventional computer slogging away for millennia. Or it could turn out to be less useful than the eight-meg laptop collecting dust in my storage space. Such is the bizarro world of quantum computing.
Dragon lives on the main floor of a nondescript office park in Burnaby, in the shadow of the gigantic western headquarters of McDonald’s Restaurants of Canada Ltd. It’s actually one of 30 generations of quantum computing machines owned by D-Wave Systems Inc., the makers of the world’s first commercial quantum computers. Special fridges can cool the hardware to temperatures 250 times colder than interstellar space (so that the computer chips can act as superconductors), but figuring out exactly what these machines are actually doing is a tall order, and a hotly contested one at that – at least among quantum physicists and computer scientists.
In theory, a quantum computer could solve problems exponentially faster than a classic computer. In an eye’s blink, it could break cryptographic codes used to protect top-secret banking or military databases. Or build a teleportation machine. Or unlock the secrets of biology by mimicking the behaviour of even the smallest molecules, which would be a breakthrough for biosciences and the pharmaceutical industry. It could also apply to robotics or miniaturizing technologies or be used to study the quirky, mind-bending laws of quantum mechanics, which could turn our understanding of life as we know it on its head. As the world demands ever-more powerful computers and ever-more tiny devices, hitting the upper limits of conventional computing power has become a real threat. Keeping increasingly muscular chips cool while also hogging less power is a challenge, which makes the lure of quantum computing even more appealing.
No wonder the Defense Advanced Research Projects Agency (DARPA, an agency of the U.S. Department of Defense), the National Security Agency, the U.S. Army, academic groups and corporations such as IBM Corp. and Hewlett-Packard Development Co. LP have been trying to build a quantum computer these past 15 years. But the so-called race perhaps fits a tortoise-hare analogy. These groups have invested about a billion dollars into developing their own quantum machines, but so far they’ve only made baby steps forward, with many scientists believing we’re at least a decade or two away from designing a useful quantum computer rivalling today’s fastest silicon-based computers. D-Wave, which seemingly came out of nowhere in 2007 with a machine that allegedly far outpaced anything else developed in test labs, aims to do it within years. The question is, in today’s uncertain financial climate, will D-Wave run out of money before its quantum machines are ready for prime time?
The scientific community has remained skeptical since D-Wave debuted its first machine in 2007 at the Computer History Museum in Silicon Valley, but that hasn’t stopped several high-tech investors and Canadian government groups from sinking more than $70 million into D-Wave since it launched in 1999, including over $22 million in 2008 alone. “I’ve always been drawn to crazy, wacky technologies,” says Steve Jurvetson of U.S. venture capitalist firm Draper Fisher Jurvetson (DFJ), which manages a $6-billion portfolio of emerging technologies. DFJ backed such projects as Skype and Hotmail, and started investing in D-Wave in 2003. Jurvetson, who now sits on D-Wave’s board, believes its machines will leave conventional supercomputers in the dust within five years. “Almost all the big winners in the high-tech field seem crazy at first, so the fact that this is an unusual technology right now is a big draw for us. Especially a commercial one like this that has the capability of being more powerful, more flexible and have much more longevity than any computer we’ve seen before.”
[pagebreak]
Quantum computers differ from conventional ones by exploiting the mysterious laws of quantum mechanics, using subatomic particles called quantum binary digits – “qubits” – to manipulate and store data. These qubits not only represent the binary units – the zeros and ones of classical computers – but each qubit can represent zero and one at the same time, in parallel. The more qubits, the weirder things get: two qubits can equal four values, four can equal 16 and eight can equal 256. Scientists theorize that a 30-qubit machine could be three times more powerful than today’s fastest supercomputer.
D-Wave’s machines, which look like large beer kegs hooked up to hoses and dozens of cables, aren’t nearly that fast, but they’re growing rapidly. They’ve graduated from seven-qubit processors to 48 over the past two years, and the company is currently fabricating 128-qubit chips, which they claim will be about 100 times faster than an off-the-shelf $5,000 conventional computer for solving certain tricky computing problems. But in order for these kinds of quantum machines to become exponentially faster than today’s conventional computers, they will need to scale their technology up to thousands or even millions of qubits. D-Wave plans to have a 1,000-qubit system operating by the end of 2009 that would bump the technology out of the R&D phase and into the real world – appealing to a variety of corporations, including Internet search engines, banks, investment firms and insurance brokers, as well as logistics, travel and pharmaceutical companies. A few dozen academics in robotics and bioscience, and a handful of corporations, including industry goliath Google Inc., are already using D-Wave’s quantum machines.
“So far, we’ve only opened the window from a tiny pinprick to a little bit bigger pinprick, and people are rightly confused about what today’s quantum computers can and can’t do,” says Geordie Rose, the 36-year-old co-founder of D-Wave, as we tour his lab. “I’m confused when I read a lot of this stuff and I’ve been doing this professionally for 15 years.” Rose looks more like the buff, friendly-looking dude at the gym who would show you how to use the Stairmaster. Indeed, he’s won championships in wrestling and beach volleyball, but he also has a PhD in theoretical physics and has managed to turn a $4,000 investment in 1999 into a high-tech startup, attracting over $20 million from various government bodies – including the Natural Sciences and Engineering Research Council of Canada, Technology Partnerships Canada and the National Research Council of Canada – and $50 million from private investors, including B.C.-based investors GrowthWorks, BDC Venture Capital and B.C. Investment Management Corp., and more recently from Goldman Sachs Group Inc.
“Dragon’s the big boy. It’s about to have that Quicksilver filter bolted,” says Rose, pointing to a long, elegant-looking cylindrical device laid out on a table, a bunch of ribbony copper wires trailing out of it like spilled guts. “The filter removes all the unwanted noise, and the chips will go in the base.” Dragon and its partner, named Unicorn, have recently been fired up for the first time, though Rose can’t give a timeline for when the 128-qubit chips will be ready for installation. Meantime, the other three older-generation machines, dubbed Orion, are currently quietly going about their complicated business in an adjacent magnetically shielded room. Rose won’t name the majority of D-Wave’s clients or say how much they’re paying to play. Their machines, he admits, are “still in some sense a vehicle for doing research.”
Real-time commercial usability, however, is what puts D-Wave in a league of its own. Quantum computers have already been built by other teams across North America, but so far all of them – at least all the ones we know about – are being used exclusively in isolated research settings. Scientists affiliated with the Massachusetts Institute of Technology (MIT) and the University of Waterloo built a three-qubit machine at Los Alamos in 1998, then scaled it up to seven qubits in 2000. An IBM-affiliated Stanford group followed with a seven-qubit system in 2001, but soon realized it couldn’t be scaled up. Hewlett-Packard has been working on a DARPA-funded project since 2005, while MIT scientists have tested a four-qubit machine with the same algorithm as D-Wave’s system.
But I’m getting ahead of myself. Actually, I’m getting a headache just trying to process Rose’s descriptions of various gizmos right in front of my face. So it’s a relief to retire to his office, which is outfitted with bright paintings, photos of his three young kids, plastic dinosaurs and a guitar.
What is D-Wave’s fleet of machines actually doing? Combing the web for Bin Laden’s emails? Creating a smarter-than-human artificial intelligence?
[pagebreak]
D-Wave isn’t interested in the encryption game, Rose says, describing the computers’ current usefulness as “narrow," designed to be used by “companies for which optimization is a core business and they’re looking for faster, more efficient solvers” – things like scheduling and routing for companies like UPS, staffing projects in corporations with large HR databases or companies that offer specialized Internet searches like Expedia.com.
“Other researchers are driven and motivated to answer questions about a number of big things, whereas we are on the absolute other commercial-product end of the spectrum,” says Rose. “We actually build stuff; our hands are dirty in the guts of the machine. Ours is a specific duct-taped-together system for solving a particular constraint-satisfaction math problem that is both ubiquitous and hard. It short-circuits a lot of the work that traditional algorithms will have to go through to find a solution.” He explains that D-Wave’s web-based interface allows even a novice computer programmer to harness the power of quantum mechanics, instead of the handful of quantum computing specialists squirrelled away in academic settings.
The details of exactly how it does that, and via what aspects of quantum mechanics, are another mind-boggling matter. In theory, a quantum-mechanical system could be built out of anything. But getting it to function as a computer is another challenge altogether. The four most promising systems developed so far have used trapped ions, electrons in semiconductors, photons or superconductors. D-Wave chose to go the superconducting route: cooling superconducting metal – in D-Wave’s case, loops of mostly niobium – to nearly absolute zero to cause the quantum behaviour. Orion, with 16 qubits (or 16 loops), debuted in February 2007 to a gaggle of media and hundreds of other onlookers, and solved three problems: matching molecules on a database, solving a Sudoku puzzle and creating a wedding seating plan.
“The wedding planner is a classic constraint-satisfaction problem with two categories of constraints: I would absolutely not sit next to this person, and I would prefer to sit next to this person,” explains Rose. “With a large enough group of people, it turns into a problem called ‘satisfiability’ by mathematicians and computer scientists.” Rose acknowledges that a conventional computer can solve Sudoku puzzles in a fraction of a second. “Ours was actually quite slow with all these problems; there were massive slowdowns. But the main point of doing it was to show an end-to-end demonstration of a quantum computer doing satisfiability problems that matter to humans.”
The demonstration attracted some media attention and a ton of skepticism from those working in the field of quantum physics. MIT-based scientist Seth Lloyd, who had already built a similar superconducting quantum computer, wrote in the May-June 2008 issue of Technology Review magazine that “D-Wave seemed to be muddying the quantum well for money” and had “neglected to supply any concrete evidence that the device was actually performing a quantum computation.” He urged D-Wave to do so; otherwise “their name within the scientific community will remain mud.” Scott Aaronson – then at Waterloo’s Institute of Quantum Computing, now at MIT – is quoted in an April 8, 2007, New York Times article as saying that without proof that qubits – and not just conventional bits – were juicing Orion, it is “as useful as a roast beef sandwich.” (He stands by the comment to this day.)
Here’s where we get into an even more challenging aspect of quantum mechanics: device testing. The marker is called entanglement, which Einstein dubbed “spooky action at a distance.” Simply put, it means measuring a qubit’s value based on its interaction with another qubit. But the act of observation causes “decoherence” and results in inaccuracies and computational mistakes. Think of a police radar gun: the force of the gun actually pushes, ever so slightly, on a car’s speed.
“All parties agree that [entanglement] is a non-negotiable requirement for quantum computing,” stresses Seth Lloyd in his Technology Review article. D-Wave’s initial response to the critics was to point out that since it is a private company, it has to protect its scientific booty and not cough up all its secrets. That further inflamed academics, and D-Wave has since acknowledged that it does indeed need to publish its test data. (It has since done so on its website, and is awaiting feedback from other scientists.)
Rose appears not to be phased by the academic heat, and I get the feeling he’s never met a challenge he didn’t like. He’s equally at ease among academics, grade students and head-bangers (on his blog, he gives the latest Metallica album a thumbs-up). “I was raised in an environment where asking questions was good and not knowing the answer was kind of a given,” says Rose, the eldest of four children raised by a linguist mother and biologist father. “The fun was trying to find the answers, and there was always the feeling that you shouldn’t believe anything anyone says until you’ve explored it yourself.”
As a teen growing up in Montreal, Rose preferred extracurricular pursuits to homework. “I didn’t open a book until my first year of university,” he says. “I had an awesome time at school, but for social reasons: girls, sports, friends.” With grades in the low 70s, Rose muscled into McMaster’s engineering program based on wrestling prowess alone. In 1994 he came to UBC for grad school and there he met three people who would soon become D-Wave co-founders: Bob Wiens (then CEO of a records management company bought out by Iron Mountain Inc. in 2000), physicist Alexandre Zagoskin and local venture capitalist Haig Farris, who was teaching a business course that Rose took in 1997.
[pagebreak]
“Geordie and Alex are both incredibly competitive guys,” Farris informs me a few days later, over coffee. (If I were quantum, I could have done both interviews simultaneously.) “They’re great philosophers and thinkers as well, and it takes that mental fortitude to pull off what they’ve done,” according to the 69-year-old co-founder of Ventures West Management Inc., now the largest venture capital group in Western Canada. “Geordie can also explain a complex problem to an idiot like me, a lawyer who’s never taken a science course in his life. It’s quite spellbinding. If I could understand the essence of it, so would customers, financiers, employees, whomever.”
Farris has seen plenty of business students come and go, but Rose had the rare combination of brains, brawn and people skills. “It doesn’t show in his demeanour, but he’s driven to succeed,” says Farris. “A lot of people can’t handle the difficulties startups face – especially running out of money, which D-Wave has experienced on many occasions.” Rose, in turn, credits Farris as “the single most important figure in the development of the company.” Farris gave the young entrepreneur that first $4,000 in 1999 (to buy a computer and printer so that he could knock out a business plan) and introduced D-Wave to key investors and Canadian government supporters.
It also helped that the dot-com bubble was near bursting when D-Wave launched. “A hundred thousand dollars was walking-around money for these guys,” says Rose. “It didn’t really matter that our idea was risky and it wasn’t clear how you could make money back – ‘Whatever, let’s be involved in something cool and fun.’ ” They raised about $500,000 within months, and so began what Rose calls D-Wave’s first life as a “global search team collecting intellectual properties.”
The quantum physics field was ripe with brilliant minds since the corporate world had little interest in building a quantum machine. D-Wave contracted a team of experts around the globe, many with world-class facilities at their fingertips. “We’d pay $100,000 to universities with $100-million worth of equipment, like the Chalmers University of Technology in Sweden,” says Rose about the network of researchers he hired to work on everything from qubit chip fabrication to quantum algorithms. This low-cost, high-leverage scheme attracted another $4 million in investor financing over the next few years. By 2004, Rose says, D-Wave had acquired “more U.S.-granted patents than all other corporations combined in the U.S. in this field. We now have a lead-pipe cinch on anything having to do with superconducting and processors of this sort and have very good picket-fence coverage of the modes of use and tests and design. It’s completely inconceivable that anybody else can operate a quantum computer without having to license our patents.”
D-Wave had thus established the blueprints for building what Rose believed to be the most practical commercial quantum computer, and, with another $7.1-million round of financing in 2003, D-Wave began hiring in-house engineers and applied physicists to build it (from one engineer in 2003, D-Wave now has 13 full-time employees, as well as 60 research collaborators in the U.S. and Europe). Orion’s 2007 debut was primarily meant as a shout-out to interested corporate and academic parties. “The demo said, ‘Hey, here we are, anybody interested?’ And that worked great,” says Rose. “Several opportunities for partnerships arose and the one with Google was the strongest. It continues to be the strongest to this day.”
D-Wave’s work with Google involves machine-learning and pattern-matching algorithms, such as automating the ability to label faces and images with names, which is incredibly challenging for today’s conventional computers. The company’s machines are able to run image-matching algorithms now, but Rose points out that it’s up to Google to decide whether to invest in deploying it with end-user-friendly applications such as cellphones or Internet searches. “Our machines are also going to break. They’re not going to function as robustly as we’d like in the beginning,” Rose admits. “But when they are functioning, they’re going to be blazingly fast. Imagine having a fairly large cluster of these machines working on learning problems in a bunker somewhere.”
[pagebreak]
Meanwhile, D-Wave is facing its own qubit-related constraints issues, as the company’s objective to “go up to millions or tens of hundreds of millions of qubits” butts up against the physical restriction presented by chip size. According to Rose, even these thumbnail-sized qubits are quite large, and shrinking them makes it more difficult to couple them to other qubits and other necessary devices. D-Wave has scaled up through 30 generations of processors to get from 16 qubits to 128 with its newest chip, which the company will soon begin manufacturing in-house. “We can fit roughly 2,000 qubits on our current processor, which is about the limit of where we can go with the current design,” admits Rose. “After that’s achieved, we need to have some other method of going to larger numbers. So the next step in the redesign – or the evolution of the technology – is getting to millions of qubits.”
Unless, that is, D-Wave runs out of investor money first. “I’ve lived through many economic crises, so the next year could be tough for D-Wave,” admits Farris, comparing D-Wave to the Wright brothers, who successfully flew the first powered airplane in 1903. “They didn’t know if they could make it fly. We’re a little bit ahead because we have a quantum computer that can answer Mickey Mouse questions now and be scaled up to become more powerful. But we don’t know how fast it can be, how high it’s going to fly and how long will it stay in the air. Corporate America and the IBMs of the world have said, ‘Everybody will be knocking on your door when you get to 500 qubits.’ But that might change in this economy. Are we looking for more investor money to get to the next level? Is water wet? We’re banking on the fact that early adopters don’t want to be left behind.”
On the positive side, U.S. venture capital spending remained relatively steady in the third quarter of 2008, down just nine per cent from 2007 figures to $7.1 billion, according to a November 2008 Moneytree report. Early-stage or seed investing accounted for almost 40 per cent of that number, with D-Wave investor DFJ the most active player in the market. D-Wave skeptic Scott Aaronson believes the field has “a long way to go” before quantum computers are really useful, but admits that, considering the company’s ability to attract investors and bright minds in engineering, “it’s possible that they will stumble on something revolutionary with spectacular returns for investors who like to place long bets.”
David DiVincenzo, manager of IBM’s Quantum Information Group, contends that none of today’s quantum computing scientists “have a clear view of what will become a money-maker,” and says it’s premature for D-Wave to assume otherwise.
Rose won’t say how much investor money D-Wave has left to burn through, only that, “in the stark reality of today’s economics, we could run out of money.” He remains steadfastly optimistic, however, that D-Wave will grow its client base and, with continued investor support, hit the technological benchmarks necessary to start making profits. And he believes the company will do it first because it’s a private organization. He cites the work of private company Celera Corp., which beat out government-funded researchers in sequencing the human genome.
“I think the pressure and desperation that come from being a startup are necessary to achieve big things,” says Rose. “With a lot of cutting-edge science, people become very specialized in a particular niche and they fail to see the big picture. Building a real quantum machine is very complicated, very expensive and takes a long time. Very few people have had the guts or the stupidity to try.”
