Bank of England: Can We Trust AI's Reasons? Exploring Model Fragility in Finance (Explained)

This episode unpacks a recent Bank of England research paper on the often-overlooked 'fragility' of deep learning models in finance. We explore how subtle changes can cause AI to give vastly different *explanations* for its decisions, even when predictions are identical. Discover the significant implications this has for financial stability, risk management, and regulatory practices. This episode explains a real academic paper in plain English for a general audience. Source paper: Deep learning model fragility and - Bank of England https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2023/deep-learning-model-fragility-implications-for-financial-stability-regulation.pdf Keywords: Deep learning, AI in finance, financial stability, regulation, model fragility, explainability, machine learning, risk management, central banking, FinTech