Understanding robots and the governance of automated technology
In a recent speech on the governance of artificial intelligence (AI) in financial institutions James Proudman, Executive Director of UK Deposit Takers Supervision at the Prudential Regulation Authority (PRA) highlighted what the boards of financial institutions need to consider before using AI or machine learning.
The type of data to be used needs to be decided. From that should follow decisions on how that data should be modelled and tested, and the outcomes also tested for correctness. Where humans are overseeing processes there must be proper incentives to good performance and a clearly understood accountability structure. This must, where necessary, be supported by appropriate training and systems.
Any new technology comes with both advantages and risks. Where AI in financial institutions is concerned many of the methods and skills used to counter the risks develop from already familiar safeguards. No technology should be adopted unless it comes thoroughly tested, everyone understands the outcomes to expect, and systems are constantly monitored to ensure that those outcomes are being delivered. That is not unique to AI, and in many cases will have been standard practice when adopting any new technology in the past. Similarly it should come as little surprise that where training is required it may be needed at all levels within an institution. Anyone with any degree of responsibility must understand what a system’s outcomes should be, why, and how to solve problems when they arise.
But AI brings some new challenges, not the least of which is the potential for a machine accidentally to ‘learn’ previously unnoticed biases and errors and magnify them. Rigorous systems of governance and a healthy level of wariness of AI may be something of a brake on its widespread adoption, but they are sensible and necessary precautions nonetheless. The aim should not be to resist new technologies, but to adopt them with appropriate caution and ensure that they are functioning as they should.
The world is changing at a staggering rate and big data, advanced analytics, smartphone technology and ever more advanced methods of communication are having profound effects on economies. Financial institutions need to change and adapt to keep pace with these developments, so AI and machine learning must be welcomed rather than resisted. There are risks, certainly, but early engagement with such systems in order fully to understand them should ensure that the appropriate safeguards are in place and functioning as they should.