The first point Dilan Rajasingham makes about the importance of quantum computing has nothing to do with qubits, optimization or machine learning.
“Quantum computing is all about the depth of our relationship with our customers,” he told DigFin.
Which is an odd thing to say, considering no customer has a quantum computer, and few could explain what one is. (See DigFin’s Glossary for help on this one.)
Rajasingham is head of emerging technology at Commonwealth Bank of Australia in Sydney. His job is to lead the bank’s dive into tech that will be commonplace in five, ten, twenty years – and understand how they will intermesh. Deep learning, the internet of things, virtual reality and blockchain are all part of this, but perhaps the most complex, and important, is quantum computing.
Basically, the amount of data that will be generated by sensor technologies, distributed ledgers and other gizmos will soon overwhelm the ability of existing, ‘classical’ computers to make sense of it all.
The applications being used today in finance, from augmenting portfolio decisions to executing trades to straight-through processing, will compete on escalating types and volumes of data. Using information to understand customers is still at an early stage of development. Only quantum computers will be able to crunch numbers fast enough to derive value from this.
Use cases, part 1
CBA sees quantum as necessary over the next decade to remain relevant to its clients – so it is trying to master the technology now, and play a role in setting protocols and determining use cases.
It has invested A$14 million in Silicon Quantum Computing, a research company spun out this summer from University of New South Wales; Telstra and the federal government are also investors. SQC is developing hardware, with the aim of building a 10-qubit quantum computer.
On the software side, CBA has been working with QxBranch, an Australian-U.S. company that provides simulators – prototypes that still rely on classical computing to generate models of how quantum computers will function, with a goal of making applications for the real thing.
Rajasingham says use cases are going to include turbocharged versions of machine learning and optimization tools already in play today (see below).
But the other focus is on cyber-security. Quantum computers have the potential, in theory at least, to decrypt even blockchain. The term for when this happens is ‘quantum supremacy’, the moment at which all encryption is broken. (The theory of this has been around since the 1990s; the best known example is Shor’s Algorithm, which has predicted how quantum computers could break all public-key cryptography.)
For a bank protecting customer assets and accounts, this is a scary proposition. “Our customers expect us, as a custodian, to safeguard these things,” Rajasingham said. The only solution is to use quantum computers to develop new forms of privacy and security.
Training day
CBA has begun training its tech people how to work in a quantum environment, via hackathons and simulations with QxBranch (see main picture).
QxBranch was founded three years ago by people involved in aerospace and defense to apply predictive analytics to pricing, risk and machine learning. With data scientists in Washington, D.C. and software engineers in Adelaide, Australia, it is helping customers such as CBA and UBS develop prototype apps for quantum. (UBS officials didn’t respond to messages requesting comment.)
Amazon, Google, IBM and Microsoft are all supporting budding quantum efforts on their clouds, racing to develop hardware for meaningful uses.
“It’s like when the iPhone first came out,” says Michael Brett, the company’s D.C.-based (Australian) CEO. “How could you have apps ready for day one?” The answer was that Apple worked quietly with app developers on simulated versions of the iPhone before the real thing was launched.
QxBranch is working to develop similar apps as big internet companies develop the hardware – themselves in partnership with startups such as SQC, among others.
Just as internet companies host a range of different hardware and specialized electronic circuits – basic central processing units, graphical processing units, and tensor processing units (for machine learning) – they are going to add QPUs. “It’s about directing the right algorithm to the right piece of hardware,” Brett said.
Use cases, part 2
Brett says there are three basic use cases for quantum: quantum chemistry, machine learning, and optimization. Quantum chemistry refers to the physics of gas and molecules, which are governed by quantum theory (as opposed to Newtonian physics for large-scale phenomena), which has massive implications for science but not for finance.
Machine learning, which is relevant to finance, will be revolutionized. Today, data scientists develop cool algorithms, but then spend hours and hours training computers how to identify, say, an image of a cat, or a number. Quantum should allow computers to take over this rote work, freeing data scientists to do something more interesting.
Optimization refers to large-scale, multi-variable activities that require huge number crunching. Brett gives the example of planning the NBA’s tournament schedule: conflicting travel needs, division matchups, special events. And you thought Kevin Durant was the guy working hard; quantum can keep it that way.
In finance-speak, says Rajasingham, a similar problem could be: what country or market is the next area of interest? This could help a portfolio manager identify coming trends, or a business manager target the next product or market expansion.
All together now
CBA’s Rajasingham says the immediate challenge, besides coming to grip with the weirdness of quantum physics, is the need to collaborate in order to make progress.
“It takes a mix of skills to work on a quantum simulator: big data scientists, software engineers, researchers,” he said. “To progress, you need a team,” which includes outside specialists, academics, vendors and regulators. “This is a paradigm shift,” he said, given banks have traditionally had a bias toward building things in-house.
Today, big banks shun public clouds in favor of on-premise ones, using their own server network. That may not be possible with quantum computing.
While Google and other big tech companies are working to add quantum to their cloud data centers, quantum computers remain expensive, specialist devices. “It’s better to buy from Goole or Amazon rather than build your own,” QxBranch’s Brett said.
But banks getting an early start developing apps have the chance to shape the language and protocols of quantum. Rajasingham likens this era to early computers that had to be fed punch cards. Software engineers are working on fundamental programming.
The internet today is driven by coders working in abstract languages such as Python that make it easy for people with only basic programming skills to develop websites and algos.
“This is the opportunity to experiment early, and not just wait and see,” he said.
