Technological advances are changing the landscape for financial firms and regulatory agencies. Researchers including Atlanta Fed economist Larry Wall are exploring how artificial intelligence and machine learning may help financial supervisors make sense of ever-larger pools of data with the goal of enhancing financial stability.
The availability of deeper information—say, data on specific loans or financial instruments—combined with new tools to analyze these data could help supervisors better evaluate the risk of banks and financial systems, Wall pointed out in an article in the Atlanta Fed’s Notes from the Vault.
Techniques such as "deep learning" use neural networks loosely patterned after the operation of neurons in the human brain. These techniques allow computers to solve relatively complex problems such as facial recognition.
There is potential here. Deep learning, Wall said, combined with deeper data could help supervisors better understand not just the operations of individual financial institutions, but also the endless web of links among financial institutions and markets.
But machine learning has limits. First, because it relies on historical data, machine learning is not currently well-suited for predicting things that have never occurred. In addition, while it can help supervisors identify correlations—X happened along with Y—even the cleverest machines can't prove that one action or circumstance caused another. Establishing "causality" is critical to understanding why things happen and how they might be prevented.