Exploring Pillars of Trustworthy AI: Robustness and Fairness

Dive into a 42-minute technical talk that explores two fundamental pillars of trustworthy AI - robustness and fairness in machine learning models. Learn about critical vulnerabilities in common AI models and techniques for adversarial attacks, followed by defense strategies like gradient obfuscation and transformations to enhance model robustness. Examine the concept of certified robustness through discussions of adaptive attacks on defense mechanisms. Understand how unconscious bias can be encoded during model training through representational and model bias, and discover practical strategies using domain knowledge to develop fair AI systems. Master essential concepts for building ethical, robust, and unbiased AI models that are crucial for responsible AI deployment in everyday applications.