Nmedia - Fotolia
Theoretical researchers have been tinkering with the idea of replacing the logical bits at the heart of modern computers with quantum computing since the late physicist Richard Feynman first proposed the idea in 1982. In theory, quantum computers will enable processing with much smaller elements and enable a new type of processing that may be far more efficient than traditional computer circuits -- a goal that will take on more urgency if Moore's Law indeed hits a wall.
Harnessing the laws of quantum mechanics to build a general purpose commercial quantum computer, however, has proved harder to do than think up. Thirty-seven years after Feynman's bright idea, quantum computing, as they say in the IT business, remains a technology in search of a solution.
But the field is generating a lot of heat if not light.
At this year's Computer Electronic Show (CES,) IBM released the Q System One, a 9-foot airtight glass-paneled, shiny cube billed by CEO Ginni Rometty as the first commercial quantum computer. A striking machine crafted by industrial designers, IBM's quantum computer is limited in the kinds of applications it can run and the performance it delivers compared to traditional computers -- a reality readily acknowledged by the company.
"We are still at an early phase in quantum computing, much like today's classical computers were in the 1950s," said Bob Sutor, vice president of IBM Q strategy and ecosystem at IBM research. "The difference now is that enterprises and individuals alike have access to IBM Q prototypes and are getting 'quantum ready' in terms of general education, and the development of practical applications that will have a quantum advantage over classical computers' capabilities."
Barclays keen on IBM's quantum computer
Include Barclays Bank among enterprises that want to be "quantum ready."
In December, Barclays joined the IBM Q Network, a community of Fortune 500 companies, startups, research labs and universities interested in quantum computing. The membership gives the bank access to IBM's quantum processors to run experiments and access to IBM's technical experts and researchers on quantum computing software.
"We are keen to explore quantum computing by running experiments on actual quantum processors, rather than just using quantum simulators running on a classical processor," said Lee Braine of the Investment Bank CTO Office.
Braine, who has a Ph.D. in computer science and focuses on emerging tech such as blockchain and smart contracts, runs the bank's quantum computing program. Much of the work on quantum computing today has focused on the theoretical aspects of quantum computers, he noted. Algorithms have been designed to do amazing things (like crack the cryptography underpinning blockchains, such as bitcoin). But until now, researchers have had to run these algorithms on simulators.
IBM's quantum computer and community is a start. "We need to collaborate with existing quantum computing experts to help accelerate our learning in this complex field," Braine said.
Barclays' step-by-step approach
Barclays's quantum experimenters have started off exploring several different optimization problems that exist in banking today in order to identify some specific challenges that may be amenable to quantum computing. As a first step, they converted each of the optimization problems into a simple abstract description so that they could classify the algorithmic nature of the problem. For example, this could include algorithms for searching, sorting, factoring or solving linear equations.
Barclays then checked whether a quantum algorithm had already been published to solve that type of problem. A comprehensive list of known quantum algorithms maintained by the U.S.'s National Institute of Standards and Technology (NIST) helped to identify useful algorithms that might form the basis of more complex computations.
The next step included attempting to construct a simplified version of each problem that could potentially be run on a quantum processor. This helped Barclays to focus on a couple of specific challenges. One is related to portfolio optimization in wealth management. A second is related to improving the processes for completing transactions in capital markets -- settlement efficiency. The bank decided to target settlement efficiency for a more extensive experiment and expect to publish a brief report on its findings later this year.
Anticipating a security threat
Barclays is also continuing to assess the long-term threat of quantum computers to potentially crack existing classical cryptography. This is a big issue since banks may have to adopt encryption technology that is more robust against quantum cryptography.
Braine's team is trying to determine the likely timeline for quantum computers to become sufficiently powerful to be able to run quantum algorithms that can factor the large numbers that are the basis of most cryptography. "There are many different opinions regarding this timeline, ranging from five years to thirty-plus years, and so we would welcome more research and industry dialogue on this topic," he said.
Initial applications of quantum algorithms
For traditional IT applications, quantum computers may work like accelerators for certain calculations, much like GPUs do today. However, since quantum computing is completely different from classical computers, they may be able to solve problems that CPUs and GPUs cannot, and will not.
"Quantum computing will likely be used to either speed up computations deep inside current machine learning or deep learning algorithms or provide completely alternative and much more efficient algorithms," said IBM's Sutor.
Industries are just beginning to explore the possibilities, and universities are beginning to develop deeply technical quantum computing curricula, Sutor said. Initial applications for the Q System One will be in chemistry and in solving optimization problems in finance and logistics, he added. For example, IBM Q Network partner JPMorgan Chase is working with IBM scientists on trading strategies, portfolio optimization, asset pricing and risk analysis. In chemistry, IBM is working with Daimler AG on use cases in materials development for manufacturing and batteries.
Difficult to program quantum computers
Enterprises face several practical challenges in using quantum computers to solve practical problems.
"Our greatest challenge occurred when attempting to construct a radically simplified version of candidate problems that are sufficiently complex to retain the essence of each challenge, and yet sufficiently simplified to run on a quantum processor," Braine said.
This included creating better programming abstraction to make it easier for developers to remove unnecessary complexity. It also involved finding ways to separate out potential solutions into a classical computing part and a quantum computing part.
"All of this is a difficult intellectual challenge, and so it is clear that future designers and programmers of quantum programs would benefit from leveraging prebuilt frameworks and components," Braine said.
IBM has built an open source software platform called Qiskit (quantum information science kit) that lets developers program the IBM Q system. In the near future, IBM's Sutor expects developers to work across a hybrid of classical computer tools for developing apps, and quantum computers for processing parts of these apps.
Lee BraineInvestment Bank CTO Office, Barclays
Another big challenge is that the number of qubits is currently limited in general purpose quantum computing processors to a few tens of qubits. A qubit is comparable to a bit in traditional computers but includes the ability to encode more data. Current computers store, process and move individual bits stored as binary 0 and 1 states. In contrast, quantum computers encode information using different physical phenomena to manipulate information that allow each qubit to encode multiple states at the same time. In theory, 300 qubits could represent more data than atoms in the universe.
IBM's quantum computer currently only supports 20 qubits.
"This places limitations on the volumes of data that can be processed, so it is necessary to radically simplify the data sets that the quantum algorithm will operate on," Braine said.
Also, quantum computing systems can operate for only brief periods before they lose information into the environment via a process known as quantum decoherence. This places limitations on the number of operations that can be performed, so it is currently necessary to keep quantum computing programs relatively short.
"For quantum computing to be more practical in banking, the number of qubits will need to increase, and the quantum coherence time will need to increase," Braine said. "Fortunately, ongoing research in quantum computing hardware has been steadily improving those two aspects."
Some scientists not convinced
Much of the work of quantum computing has been conducted by theoretical physicists, but making this work in practice could prove challenging. Many scientists are not convinced that practical quantum computers will pan out in the next decade, if ever.
"The only thing IBM announced was a huge machine made of borosilicate glass, but they say nothing about what is inside this machine," said Professor Michel Dyakonov, of the Laboratoire Charles Coulomb, Universite de Montpellier in France.
Dyakonov said that actually programming the physical representation of qubits in a quantum system is an extremely difficult problem from a practical perspective. A good analogy is a typical bike has three degrees of freedom, and with a little training one can learn to ride a bike. But programming a quantum computer is like trying to ride a bike with a thousand joints in which all the parts can rotate around those joints in a continuous manner. "Could you ride such a bike?" Dyakonov asked.
Enterprises are hopeful
Barclays is optimistic about the future, Braine said. "We expect the capability of quantum computing hardware to continue improving in the coming years -- in both the number of qubits and the quantum coherence times," he said.
"We also expect an increased awareness of quantum computing to increase the number of developers who have knowledge of quantum programming techniques and actual experience of constructing and running quantum circuits."