Responsible Design of Edge AI: A Pattern Approach for Detecting and Mitigating Bias

Watch a 15-minute conference talk from the tinyML Summit 2023 exploring the critical issue of bias detection and mitigation in Edge AI systems. Learn about a pattern-based approach developed from machine learning fairness best practices, specifically adapted for Edge AI applications. Discover how these patterns can be applied to voice activation systems with keyword spotting and speaker verification components. Gain insights from PhD Candidate Wiebke Hutiri of Delft University of Technology on addressing the growing concern of AI bias, particularly in Edge AI where research remains limited despite widespread deployment. Understand the importance of responsible AI design as voice assistants are projected to reach 8.4 billion deployments by 2024. Explore how pattern catalogs can effectively capture and communicate design knowledge, making best practices transferable across Edge AI domains to facilitate the development of fair and trustworthy systems.