By Larry D. Wall, Executive Director, Center for Financial Innovation and Stability, Federal Reserve Bank of Atlanta.
Larry, can you please tell the Risk Insights readers a little bit about yourself, your experiences and what your current professional focus is?
I have a Ph.D. is in business administration from the University of North Carolina at Chapel Hill. For college basketball fans, my last year in Chapel Hill was Michael Jordan’s championship year. I moved from Chapel Hill to Atlanta where I have spent my career the Fed studying various banking and financial issues. My current research interests lie in the areas financial innovation and stability.
Disclaimer: The views expressed here is that of the author and not necessarily those of the Federal Reserve Bank of Atlanta, or the Federal Reserve System.
At the Risk Americas 2018 summit, you will be speaking on your insight regarding ‘Some financial regulatory implications of Machine Learning: An economist’s perspective’. Why is this a key concern right now? And what are the essential things to remember?
A combination of more powerful computers, lower cost data storage and improved algorithms have fueled impressive developments in the ability of machine learning and artificial intelligence in the last few years. We are now at the point where Andrew Ng could state that “Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don’t think AI will transform in the next several years.” Whether the “next several years” is the right time horizon remains to be seen. However, what is clear is that firms in almost every industry are either using artificial intelligence/machine learning (AI/ML) in production or actively exploring its use.
What is the essential thing to remember about AI/ML is that it represents a set of tools that have some very valuable strengths but also some notable weaknesses. The AI/ML of science fiction is still just science fiction. The strength of available AI/ML algorithms is that they can be extraordinarily good at finding correlations given sufficient good data. That weakness arise from its need for lots of good data and its inability to go very far beyond identifying and responding to correlations.
What are the main considerations of further integration of machine learning into businesses?
Identifying good use cases that play to its strengths and where its weaknesses can be mitigated.
Without giving too much away, how could machine learning impact the markets for financial services?
My biggest long run concern about ML in financial services arises from the combination of two issues: (a) ML will be a major source of competitive advantage in many areas and (b) ML works best with lots of data—the more the better. The combination of these two could be a virtuous cycle for the winners, more market share yields better ML which leads to more data and even better ML. On the other hand, once a financial firm starts losing market share it could be very difficult to reverse the trend. The long-run result could be a few big winners with most of the rest exiting the business. This could lead to a significant reduction in competition in some areas of finance and an increase in the scale of the too-big-to-fail financial firms.
Where can I find some of your writings on financial stability, financial innovation and machine learning?
You can find the paper that is the basis for my talk at the 2017 CEAR/CenFIS Conference here.
I also have a paper with Julapa Jagtiani from the Federal Reserve Bank of Philadelphia and Todd Vermilyea in Banking Perspectives titled The Roles of Big Data and Machine Learning in Bank Supervision here.
Hear insights like this and more at the Annual Risk Americas Congress…