A Statistical Journey through Trustworthy AI

Explore the statistical perspectives on trustworthy AI in this 46-minute seminar presented by Guang Cheng at the USC Probability and Statistics Seminar. Delve into the challenges of privacy, robustness, and fairness in AI development, focusing on machine un-learning for privacy protection, utilizing artificially generated data to enhance adversarial robustness, and establishing fair Bayes-optimal classifiers. Gain insights into the unique contributions of statisticians in advancing trustworthy AI through empirical, methodological, and theoretical approaches, as the speaker argues that the next generation of AI will be driven primarily by trustworthiness rather than performance alone.