From data to culture and back: Building trustworthy learning systems
Learning systems are increasingly built upon data that reflect our interactions, intentions, and choices, effectively transforming these digital footprints into cultural artifacts. The learned models themselves become cultural artifacts, embedding significant information about the individuals and the societal context from which they originate. This relationship has profound implications, particularly concerning the impact of AI on choices, narratives, and society. This seminar will address these complexities by discussing strategies for training AI systems to align with desired goals, relating legal frameworks and lived experience, mitigating unintended consequences, and robustly protecting sensitive information. Solving these critical issues will allow us to develop trustworthy learning systems for the good of society.