You may have seen headlines that Google is working on a quantum computer chip with researchers from the University of California Santa Barbara. But there are some questions you might have before we get into the specifics of the initiative. What is a quantum computer? What can it do? Further, what is Google planning to do with a quantum computer once it’s built one?
Let’s start at the beginning: what is a quantum computer? The University of Waterloo’s Institute for Quantum Computing says that quantum computing is “essentially harnessing and exploiting the amazing laws of quantum mechanics to process information.” Quantum mechanics, in turn, encompasses the mechanics of the motion and interaction of subatomic particles.
A quantum computer is different from a traditional computer in the way that it processes information. A traditional computer uses strings of “bits,” which encode information in zeroes and ones. A quantum computer uses qubits, or quantum bits, units of a quantum system that encodes the zeroes and ones into two distinguishable quantum states. A qubit can represent a zero, a one, or both values simultaneously. But because qubits behave “quantumly,” it’s subject to quantum effects like superposition and entanglement.
Superposition is the ability of a quantum system to be in multiple states at the same time, so that it can be “here” and “there” or “up” and “down” simultaneously. Entanglement is a correlation between quantum particles so strong that it remains perfectly correlated even across great distances, and are said to “dance” in perfect unison even when placed at opposite ends of the universe. Thanks to these effects, a quantum computer can process a large number of calculations simultaneously, and at extremely high speeds.
“Think of it this way,” the University of Waterloo’s page explains. “Whereas a classical computer works with ones and zeros, a quantum computer will have the advantage of using ones, zeros, and ‘superpositions’ of ones and zeros. Certain difficult tasks that have long been thought impossible (or ‘intractable’) for classical computers will be achieved quickly and efficiently by a quantum computer.”
But the effectiveness of quantum computing is the subject of a heated debate, and scientists actually disagree about what exactly a true quantum computer is. As the Institute for Quantum Computing page notes, researchers from IQC and MIT hold the current world record for the number of qubits used in an experiment — twelve — but a true quantum computer that is able to outperform a classical computer is still “years away.” While a single company is selling commercial quantum computers, it’s unclear whether its machines are actually faster than traditional computers. (We’ll get to that in a few paragraphs.)
So, now that we’re up to speed on the basics, here’s the most recent news: Google has announced that it’s expanding its quantum computing research and backing a project to build a new type of quantum processor. In a Google+ post, director of engineering Hartmut Neven says that the Quantum Artificial Intelligence Lab is launching a hardware initiative to build “new quantum information processors based on superconducting electronics.” John Martinis, a physics professor who has researched quantum computing since the 1980s, and some members of his team of about 20 at UC Santa Barbara will join the project.
Martinis will become a joint employee of Google and USCB, and TechCrunch reports that Martinis and his team will be based in Google’s Santa Barbara office and continue to use UCSB’s fabrication and measurement facilities. Google explains that:
“With an integrated hardware group the Quantum AI team will now be able to implement and test new designs for quantum optimization and inference processors based on recent theoretical insights as well as our learnings from the D-Wave quantum annealing architecture. We will continue to collaborate with D-Wave scientists and to experiment with the ‘Vesuvius’ machine at NASA Ames which will be upgraded to a 1000 qubit ‘Washington’ processor.”
As The Wall Street Journal reports, the UCSB team published a paper on a five-qubit array that demonstrated advances in correcting specific errors that occur under the conditions that create quantum effects. Ars Technica explains that a little further, noting that Martinis and his team have focused their research on creating fault-tolerant qubits “using a solid-state superconducting structure called a Josephson junction. By linking several of these junctions and spreading a single quantum state across them, it’s possible to reach fidelities of over 99 percent when it comes to storing the quantum state.” Larger systems could be configured to run almost any kind of algorithm to solve the problem at hand, but scientists say that to be useful, a quantum computer would need to be built with tens of thousands of qubits, or more. To do that, qubits need to be fabricated to more stably maintain quantum states.
Josephson functions are one approach to making qubits more robust, so that quantum states are less prone to decaying when it interact with its environments, but have the advantage of being familiar to computing companies, which can make them with standard fabrication techniques. The focus on Josephson functions could demonstrate that it will work for multi-qubit machines, or could end up proving that it has severe limitations. In either case, the focused research is likely to “narrow the field,” as Ars Technica puts it, of potential technologies being researched.
Martinis told The Wall Street Journal that he hopes the new project with Google will yield technology that “will not lose its memory” as quickly as earlier hardware. But he views the new hardware initiative as a “complementary approach” to what D-Wave is doing, and expects it to benefit from Google’s ongoing work with D-Wave, including investigations into possible applications for the new hardware that Google develops.
So, let’s get to the backstory and the reasons why Google needs to build new hardware to create its own quantum computer. As mentioned earlier, only one company claims to be selling commercial quantum computers: Vancouver-based D-Wave Systems. The company was founded in 1999 and in 2011 shipped a computer ordered by Lockheed Martin to a facility run by the University of Southern California. The only other machine shipped so far is one ordered by Google.
As Ars Technica explains, D-Wave uses a new approach to quantum computing, different from the traditional method of processing information with qubits, and its machines “perform a process called quantum annealing instead of the more typical approach, which involves encoding information in a quantum state of a collection of entangled qubits.” But Google continues to be interested in the approach, even as tests have reported to show that the $15 million D-Wave 2 isn’t any faster than a traditional computer, and scientists debate whether the D-Wave is actually a true quantum computer at all.
As The Verge reports, Martinis last year referred to the actual manufacturing of a true quantum computer “a physics nightmare.” MIT’s Technology Review notes that Martinis co-authored a paper that concluded that there was “no evidence of quantum speedup” displayed in tests run on the D-Wave computer. Critics say that without quantum speedup, the D-Wave is simply a conventional computer, but the company claims that the tests used the wrong sort of problems to demonstrate the machine’s abilities.
But Martinis’ work on the D-Wave machine led him to talks with Google. The chip in D-Wave’s latest computer has 512 qubits, wired into a more limited component called a quantum annealer. The annealer can run only a specific algorithm for a specific type of problem “that requires selecting the best option in a situation with many competing requirements — for example, determining the most efficient delivery route around a city,” Technology Review explains. According to Martinis, theory and simulation suggest that annealers could deliver quantum speedups. “There’s some really interesting science that people are trying to figure out,” he told Technology Review.
Martinis also thinks that his method of fabricating qubits could make better quantum annealers. Martinis told Technology Review that he hopes to make an annealer with qubits that can stably maintain a quantum state of superposition. The qubits of D-Wave’s computer can maintain superpositions only for nanoseconds, and Martinis has built qubits that can maintain superpositions for as long as 30 microseconds. He builds qubits from aluminum circuits built on sapphire wafers, and chills them to 20 millikelvin, near absolute zero, so that it can become superconducting. D-Wave’s chip is similarly cooled, but its circuits are made from niobium on top of silicon wafers. Martinis is switching to making his own qubits on silicon, but believes that insulator materials used in the D-Wave chips may limit its performance.
Since its purchase of a D-Wave machine, Google has teamed up with NASA and the Universities Space Research Association (USRA) to expand its research into quantum computing beyond a single approach. The three established the Quantum Artificial Intelligence Laboratory (also called the Quantum AI Lab or QuAIL) in 2013.
NASA’s objective with the collaboration is to find ways that quantum computing and quantum algorithms can improve NASA’s ability “to solve difficult optimization problems for missions in aeronautics, Earth and space sciences, and space exploration.” USRA notes that the collaboration “brings together university research, government research and industrial research” to ascertain “the benefits of quantum computing for a range of challenging applications.” The Google+ page for the Quantum AI Lab notes that the group is studying “the application of quantum optimization to difficult problems in Artificial Intelligence.”
USRA’s website notes that the collaboration would use a D-Wave Two quantum computer, installed at NASA’s Advanced Supercomputing facility at the Ames Research Center. At the time that the partnership launched, USRA noted that the computer would initially be equipped with 512 qubits and will be upgraded to 2,048 qubits when the capability becomes available. The D-Wave machine is the subject of a number of scientific papers, which debate whether or not it exhibits the characteristics of a true quantum computer. In Google’s announcement, Neven notes that Google will continue to collaborate with D-Wave scientists for experiments on the “Vesuvius” machine at NASA, which will be upgraded to a 1,000 qubit processor.
So when is Google likely to produce its first quantum computer? What would it use the machine for? While the quantum computer could take years to develop, it could be put to work on a huge variety of tasks, especially as researchers combine D-Wave’s focus on achieving scale — building quantum computers with as many qubits as possible — with Martinis’s research on achieving stability. As the page for the University of Waterloo’s IQC explains, the potential of quantum computing is huge:
“Simulation of quantum systems has been said to be a ‘holy grail’ of quantum computing: it will allow us to study, in remarkable detail, the interactions between atoms and molecules. This could help us design new drugs and new materials, such as superconductors that work at room temperature. Another of the many tasks for which the quantum computer is inherently faster than a classical computer is at searching through a space of potential solutions for the best solution.”
A quantum computer could ultimately accelerate the process of searching for patterns in any kind of data, from weather to stock market information. Quantum computers could crack data encryption by factoring large numbers — a task that’s virtually impossible for a traditional computer — and could facilitate drug discovery research or financial modeling. Google could anticipate using quantum computing ability in concert with the research the company is beginning to undertake into health and genetics. Alternatively, or additionally, the company could be looking to apply a quantum computer’s power toward advanced forms of artificial intelligence.
As PC Mag reported in July, Google co-founder Sergey Brin said that the company eventually hopes to develop “fully reasoning” artificial intelligence. At a rare joint interview with Larry Page, Brin noted that, “You should presume that some day we will be able to make machines that can reason and think and do things better than we can.”
Artificial intelligence is a very possible application for Google’s quantum computing research, with NASA’s QuAIL page noting that in the next five years, the team will develop “quantum AI algorithms.” USRA’s page also notes that its collaboration with NASA “started with a focus on supercomputing and artificial intelligence, and now extends that focus to include the intersection of quantum computing and artificial intelligence.”
TechCrunch also notes that further collaborations among the researchers working at Google could drive artificial intelligence applications of quantum computing, with Martinis citing an interest in machine learning, a subset of artificial intelligence that sees computers learning from data:
“Martinis says that he is specifically excited about Google’s expertise in ‘mapping machine learning applications to a quantum computer.’ While Martinis doesn’t explicitly mention this in today’s announcement, he is probably also referring to Google’s hire of Geoff Hinton, a pioneer in the field of deep neural networking and who came to Google in 2013 from the University of Toronto as part of Google’s acquisition of his startup DNNresearch.”
While the quantum computing initiative is very likely part of Google’s ambition to create machines that are able to think like humans do, quantum computers could change the processes behind the way we develop drugs, how we detect diseases, how we analyze the weather, or even how self-driving cars drive themselves. In May, upon the launch of the Quantum Artificial Intelligence Lab, Neven acknowledged the vast potential of quantum computing and its potential applications in machine learning:
“We believe quantum computing may help solve some of the most challenging computer science problems, particularly in machine learning. Machine learning is all about building better models of the world to make more accurate predictions. If we want to cure diseases, we need better models of how they develop. If we want to create effective environmental policies, we need better models of what’s happening to our climate. And if we want to build a more useful search engine, we need to better understand spoken questions and what’s on the web so you get the best answer.”