I had just finished reading Chad Orzel’s bestselling book
‘How to teach quantum physics to your dog’ last week when my colleagues presented what is perhaps one of the first concrete applications of ‘quantum-inspired’ computing in European banking. Before I turn to that, allow me to share a few words on Orzel’s book and how it got me to think about quantum computing in financial services…

‘How to teach quantum physics to your dog’ is an excellent read. The book features Emmy; she is no ordinary dog. When Professor Orzel adopts her from the shelter, his work enraptures her.

Could she use quantum tunnelling to burrow through the neighbour’s fence? What if she could diffract around a tree to catch the squirrel? Or use virtual particles to catch bunnies made of cheese? As fascinating as it is, it’s also much more than that because it helps us understand how quantum computing, a technology underpinned by quantum physics, might be relevant in financial services, too.

The cover of Chad Orzel’s international bestseller ‘How to teach quantum physics to your dog’

What is quantum computing, how does it work and how does it help financial services?

We have almost reached the physical limit of classical circuit-based computing; we are struggling to make them smaller and hence may not be able to make them faster.
We need a breakthrough!

That major breakthrough – I am told – is quantum computing. However, a quantum computer is not simply a more powerful version of a regular computer, in the same way a lightbulb is not simply a more powerful candle. Quantum computing is a different technology based on a deeper scientific understanding of quantum physics.

Colleagues in Reply’s
Quantum Computing Practice explain that,
“regular computers see everything as ‘bits’. They see everything as either a 1 or 0. A quantum computer, however, can exist in a superposition. It can be both a 1 and 0 at the same time. What’s more, it can be any combination of 1 and 0, so it could be 90% 1 and 10% 0. Neither is it static. That means it can choose to be a 0 or 1 at any time to fit its purpose. This is what’s known as a Qubit.”

A normal ‘Bit’ and a ‘Qubit’. A representation of a two-level quantum system, by Smite-Meister, licenced under (CC BY-SA 3.0)

Dr. Shohini Ghose (1) provides a compelling example. If you play heads or tails with a regular computer, you always have a 50% chance of winning. Play enough and the result will always even out at 50% heads, 50% tails. With a quantum computer, however, you could lose every time. Let us say that the computer can represent heads as 0 and tails as 1. Because a quantum ‘bit’ is both 0 and 1, and has the ability to change, it can always be whatever it needs to be to win. This non-binary fluidity is what gives a quantum computer its power.

This power has the potential to impact everything from our health, our security, and the internet.

Companies worldwide are locked in an ‘arms race’ to build these quantum devices at great cost.

The possible use cases for quantum computing in financial services

A recent paper titled
Quantum Computing for Finance: Overview and Prospects (2) discusses how to apply quantum computation to ﬁnancial problems (e.g., non-linear problem with multiple probabilities), noting that not all computers will be replaced with ‘quantum’.

The paper discusses the
use of quantum optimisation algorithms to optimise asset portfolios with possible applications for robo-advisors.

What’s more, it discusses an algorithm offering an innovative way to select optimal features for credit scoring. Our most advanced clients (banks and FinTechs) today already apply a machine learning method known as ‘classification’. Each data point (customer) is expressed as a vector, living in the vector space of all considered attributes (customer characteristics). The training set is labelled such that each vector belongs to a class (the loan risk). When given a new vector, the program determines to which class it most likely belongs. The classification algorithm is an essential tool to determine the probability as to whether or not the customer is likely to repay the loan. What struck me with the paper was how quantum computing could enhance existing approaches, particularly where a lender lacks access to full consumer insight; fascinating!

The applications are endless. The paper goes on to discuss the proposed use of quantum amplitude estimation to achieve ‘quantum’ acceleration in Monte Carlo algorithms. Monte Carlo algorithms are ubiquitous not only in finance but across physics, chemistry, engineering etc. In ﬁnance, we typically use the stochastic approach to simulate the effect of uncertainties affecting the pricing of derivatives, for example. The downside is that if we want to obtain the most probable outcome of wide distribution, the required number of Monte Carlo simulations can become prohibitively large and result in hour-long calculations mobilising computers within a financial institution. The authors posit that one could use a quantum-accelerated Monte Carlo algorithm to achieve a quadratic speed-up in the pricing of financial derivatives. The benefits of this, in terms of cost savings and trading edge, would be phenomenal.

Is it all for real?

To be clear, I had to tell my dog, Nessie (yep, that is the name the kids chose…) that she wouldn’t be able to use quantum tunnelling to burrow through our neighbour’s fence. But, as alluded to in the introduction, I was interested by a case study that my colleagues presented last week.

Under the leadership of Giorgio Pavia, Partner at Avantage Reply, one of our teams joined forces with one of our investment banking clients to develop a
‘quantum-inspired computing solution’ in Collateral Management.

The client bank needed to optimise the management of collateral costs (margin calls) on a daily basis. To do so, it was necessary to implement a framework with the capacity to manage and aggregate substantial quantities of data of varying natures (e.g., market data, collateral agreement data, internal data of present positions).

The
purpose was to feed flexible optimisation algorithms to the bank that are capable of rapidly responding to various collateral cost problems, particularly in response to counterparties with outstanding positions in OTC derivatives.

The
results are astonishing. The ‘quantum inspired solution’, which flexibly uses numerous data inputs, has improved computational timing to provide the bank with an economic saving of as much as 9% per day on margin calls.

Beyond this, Giorgio and his team are now extending the adoption of ‘quantum inspired computing’ to other topics such as xVA.

Building upon these achievements, teams at Reply are testing the merits of quantum computing in finance for the measurement of market and counterparty credit risk. Now, there are two parts to this, i.e. the ‘quantum computer’ and the ‘quantum algorithm’. Thinking about problems through a quantum lens has helped us solve the above problems differently (faster) whilst still using classical computers. In some cases, we will need the billion $ machines that will result from the ‘arms race’ I was referring to, above, but in others, a quantum simulation of a quantum algorithm can achieve mind bending results.

Definitely a topic that will feature in our
Annual Risk Symposium!

(1)
Dr. Shohini Ghose is Professor of Physics and Computer Science at Wilfrid Laurier University. She is the President of the Canadian Association of Physicists. (2)
The authors are Román Orús (Institute of Physics, Johannes Gutenberg University, Germany), Samuel Mugel (Quantum for Quants Commission, Quantum World Association, Spain), and Enrique Lizaso (The Quantum Revolution Fund, Spain).