Real-World Applications of Artificial Intelligence in Investment Banks

Google, IBM, universities, and several other companies are throwing money at quantum computing but the question remains if they will be able to see a return on this money.  We are just in the infancy of the quantum computing space, and for most, it is truly not even understood, let alone known how it could be applied to the finance industry.  We would like to make a brief introduction to this complex topic, and a few ways that are being studied by which quantum computing may be utilized in the finance and banking sector.

Regular computing is binary, zeros, and ones.  Using these we are capable of tackling many problems.  The issue lies in the size of these problems and the computing power.  As each level of complexity increases with a problem, the number of bits needed is squared. 

Quantum computing uses qubits these are similar to the position of an electron.  They can be a one or zero or both or a combination of the two at the same time called super position.  This is a very simple description but according to University Of Texas Austin professor Scott Aaronson, quantum computing is based on “amplitudes” of positive and negative and a qubit has an amplitude of being zero or one, kind of like a weather forecast that has a chance of rain 0% to 100%.  Quantum computers want to have the amplitudes of qubits leading to wrong answers to cancel each other out.  With qubits the number of positions allow the complexity to increase to the power of 2.  Two qubits have 4 positions (4=22) and three have eight total positions (8=32).

Currently, there are computers with 53 qubits built by Google and IBM, this would take 253 classical bits.  To give some perspective there are 2300 atoms in the universe.  The expectation according to Martin Reynolds Managing VP Gartner of is that computers with tens of thousands of qubits will be required to tackle real-world problems.  Reynolds thinks that the doubling of qubits every year will be seen over the next decade.

So, the next question is how this type of computing power can be harnessed to help the finance sector.  There are three areas where the potential of quantum computing seems to be possible. 

IBM has been testing a few areas which they lay out in a report, stating they see three main categories where they see as the future of quantum computing in finance, targeting and prediction, trading optimization, and risk profiling.  These categories can potentially help an organization with front-office and back-office decisions regarding client management for “know your customer,” credit origination, and onboarding, as well as trading and asset management, and business optimization, including risk management and compliance.  These are all currently in the theoretical space.

In a paper by Román Orús in the Reviews in Physics Journal, he outlines these three areas that through some small scale tests show promise for quantum computers.

  1. Portfolio optimization– in an experiment to answer the questions, “Which assets should be included in an optimum portfolio? How should the composition of the portfolio change according to what happens in the market?” The goal was to “find the optimal trajectory in the portfolio space, while taking into account transaction costs and market impact.”  This problem was solved on two D-Wave chips with 512 and 1152 simulated qubits.  Only small instances were implemented but were done so successfully.
  2. Optimal arbitrage opportunitiesmaking profit from differing prices in the same asset in different markets. In a white paper by Rosenberg, optimal arbitrage opportunities can be detected using a quantum annealer(a process using quantum fluctuations to find a system optimization).  These results were implemented on the D-Wave 2X quantum annealer, for a small-size example with five assets and correct results were obtained.
  1. Optimal feature selection in credit scoring– financial institutions that specialize in lending money to estimate the level of risk associated with a loan. Credit scoring is a textbook machine learning problem.  Banks want to determine which data on past applicants can provide useful information in determining the creditworthiness of new applicants.  In a paper by Andrew Milne feature selection was shown to be an optimized quantum annealer.  This was implemented as a proof-of-principle on the 1QBit SDK toolkit.  These results show that future quantum annealers can also be used to determine optimal features in credit analysis.

IBM in another paper does discuss the potential of, as they put it “a quadratic speed up” to help with risk analysis which they are currently in the testing phase of.  They see this speeding up of computations helping with creating of “new algorithms to optimize portfolios, price derivatives, analyze risk, or calculate more accurate default probabilities, can have a massive and widespread impact on the long-term success of global financial institutions”  As an example of the quantum benefit, ”a classical Monte Carlo simulation (which)may require millions of samples, (but)only a few thousand quantum samples might be sufficient on a quantum computer”

So far, these small-scale tests are just scratching the surface of the potential that quantum computers will have on the finance industry.  Some other areas that have potential are data classification, regression, support vector machines, and principal component analysis.  Though these breakthroughs are probably still quite far off, as the development of quantum computers continues, with each new qubit added we move ever closer to solving these issues.

Disclaimer:  The author of this text, Robin Trehan, has an undergraduate degree in Economics, Masters in international business and finance, and MBA in electronic business. Trehan is Senior VP at Deltec International The views, thoughts, and opinions expressed in this text are solely the views of the author, and not necessarily reflecting the views of Deltec International Group, its subsidiaries, and/or employees.