Thursday, 29th of February 2024, 13:00 to 14:00 UK time, join the British Hydrological Society and Young Hydrologic Society for the first 2024 episode where Dr Ziad G. Ghauch from The Alan Turing Institute will show how to find an escape in the AI labyrinth and use it to protect our critical infrastructure.

Get your free online seat at: https://events.teams.microsoft.com/event/fe8eaf2e-4096-4515-91d6-5c4111b5feb9@b2e47f30-cd7d-4a4e-a5da-b18cf1a4151b. Looking forward to seeing you there!

Abstract: In the rapidly evolving field of AI, the need for interpretable and explainable models has become paramount, especially in certain critical domains. This presentation aims to shed light on the concepts of interpretability, explainability, and transparency in AI systems. We begin by defining these key terms and outlining the importance of explainability in AI. A deep dive into the current state of AI reveals the challenge posed by “black box” models, particularly in complex deep learning systems. To address this, we explore the concepts of feature importance and model-agnostic methods, with focus on popular tools like LIME and SHAP. These methods offer insights into model decisions, fostering a deeper understanding of AI outputs. We further highlight the multifaceted benefits of explainability in AI, including enhanced trust and transparency for stakeholders, compliance with regulatory requirements, and addressing ethical concerns like bias and fairness. Finally, we discuss the application of these explainable methods to infrastructure.