January 8, 2019 | Written by: Neil Dodgson
Categorized: AI | Banking | Cloud | FinTech | IBM RegTech Innovations
Share this post:
At school, we were taught classical physics. The laws of Newtonian gravity dominated our studies and life seemed so simple – everything moved in a straight line unless acted on by an external force and F=ma, mass x acceleration. Then it all changed. Strange effects were observed, the photoelectric effect and the wave particle duality of light. The idea of quantum mechanics was born and is perhaps the greatest scientific theory of the 20th century. Einstein, Planck and Niels Bohr revolutionized the way we see the world and our understanding of the basic laws of physics.
Quantum theory has been proved and led to significant advancements in many scientific fields – quantum electrodynamics, quantum chromodynamics, quantum gravity, quantum optics, quantum chemistry and of course, yes, you guessed it, quantum computing. One of the most important implications from the quantum theory of particle physics is that the concept of discrete values is relaxed. So, imagine a ‘thing’ that could exist as a 0 or 1. Well now this can exist as a 0 or a 1, as a 0 and 1 at the same time or any value in between with some probability.
As you all know, Moore’s law is reaching its limit, possibly as early as 2020. Moore’s Law refers to the number of transistors within computer chips – doubling nearly every 2 years. Chips nowadays contain several billion transistors; modern day ones being counted in size of tens of atoms rather than thousands. So, are we reaching the end of increasing computing power? The billions of transistors exist in on or off state, or as 1 or 0 in computer science logic. This represents a bit of information. So, if we introduce quantum theory into the old tin boxes, and instead of bits use quantum bits, qubits to his friends, then in theory the qubit can take on any value between 0 and 1. The more complicated technical term is superposition, but let’s not worry about it at this stage.
IBM 50Q: An IBM cryostat wired for a 50 qubit system
Quantum computing is real, even if still in the infancy stage. But what has this got to do with risk management I hear you ask? Well, one of the biggest potential uses of quantum computers is a simulation. When we talk about the ability to perform large, complex mathematical calculations and simulations in real time on big data (forget the words nanoseconds and microseconds) in quantum computing we mean real time (Einstein did place some restrictions on us with the speed of light limit for transferring information). The regulations now being imposed on banks by the regulations demand a significant increase in the number of simulations. The old days of historical VaR and even Monte-Carlo VaR pale into insignificance when we look at CVA sensitivity calculations for the FRTB CVA capital charge.
It doesn’t just stop there. In order to try and reduce the numerous disputes, our friends at ISDA published the ‘looks SIMple but isn’t so SIMple’ SIMM methodology. In case you’re not familiar, SIMM is the Standard Initial Margin Model for non-cleared derivatives. It requires calculating sensitivities of the non-cleared trades to a large set of market risk factors on a daily basis. Now, the issue is that non-cleared derivatives tend to be complex in nature – otherwise they would be cleared through the CCP’s like LCH. While margining and the posting of collateral reduces some xVA measures like CVA/DVA, it introduces new ones like FVA (funding value adjustment) and MVA (margin value adjustment). Unfortunately, risk follows the second law of thermodynamics – it either stays the same or increases! The issue is when calculating MVA, the sensitivities need to be evaluated in many thousands of Monte-Carlo scenarios for multiple simulation dates. We are talking about n sensitivities, over MC scenarios, over simulation dates for x number of trades. That is n*m*t*x calculations in layman’s terms just to calculate MVA.
In practical terms, quantum computers will need hundreds of qubits. However, claims in the press recently suggest that a quantum computer is 100 million times more powerful than current computers (of course we can always create virtual supercomputers through grid distribution but then what about virtual quantum supercomputers?). Current research labs, including our very own IBM, have machines with tens of qubits. Maybe we will see more of Moore’s Law to be applied to qubits? But then imagine a world where quantum computing is applied to key tasks with cognitive analytics to analyse millions of scenarios and optimize our pension portfolio asset allocation across tens of thousands of securities, minimizing transaction costs all in real time, 24/7 through our own personal roboadvisor.
I leave you with a quote from one of the greatest theoretical physicists of the 20th century, Richard Feynman:
‘What new modifications of our thinking will be required to permit us to analyze phenomena occurring within quantum dimensions?’
Or, maybe a better one is from Mr Spock:
“It’s computing, Jim, but not as we know it! Live long and prosper.”
Global Head, Customer Solutions Group, Risk & Compliance, IBM Watson Financial Services