Bank of England Member Describes How AI Could Threaten Financial Stability
Commentary by Simon Lovegrove
A member of the Financial Policy Committee of the Bank of England urged regulators, market participants and AI safety experts to work together to ensure that the behavior of any future algorithms can be constrained and controlled.
In a speech titled "Monsters in the Deep?," external member Jonathan Hall highlighted: (i) the complexity and unpredictability of deep learning models, and (ii) the possibility of collusion or destabilizing behavior that could arise from an unconstrained profit maximizing function. He stated that these factors create performance and regulatory risks for trading firms using neural networks for trading applications. Mr. Hall warned that the adoption of deep trading algorithms could lead to a less resilient and highly correlated market ecosystem and the danger of neural networks "learning" the value of actively amplifying an external risk, e.g. spoofing.
Mr. Hall suggested three areas of focus:
- Training, monitoring and control: Any deep trading algorithms will need to be trained extensively, tested in multi-agent sandbox environments and constrained by tightly monitored risk and stop-loss limits.
- Alignment with regulations: Any deep trading algorithms must be trained in a way that ensures that their behavior is consistent the regulatory rule book.
- Stress testing: Stress scenarios should be created using adversarial techniques, as managers and regulators cannot rely on neural networks behaving in a smooth manner. Stress tests should also be used not just to check performance and solvency, but also to better understand the reaction function of deep trading algorithms. Testing must be ongoing to ensure that the reaction function has not changed due to forgetting or opponent shaping.
Commentary
Global regulators are grappling with the same core questions about risks potentially created by AI and deep learning along with regulators in the U.S. In his speech, Mr. Hall described some of those risks and urged more collaboration on ways to control the behavior of future algorithms. From a U.S. perspective, readers may recall Gary Gensler and Lily Bailey’s "Deep Learning and Financial Stability" article and related speeches on five ways AI could lead to future financial crises.
These comments are important markers in the ongoing debate about the future of regulation. Although deep learning is incredibly powerful, it can go sideways, so careful thought needs to be given as to how to proceed. It’s important to note, as well, that the risks described in Mr. Hall’s speech arise from potential future implementation of deep trading agents, rather than anything Mr. Hall is seeing at the moment.