AI Alignment and Safety
How do we align the goals and behaviors of AI systems with human interests? How can we convey human preferences using reinforcement learning?
This theme explores the challenges of aligning AI models to human values along three research thrusts: designing a methodology for aligned reward functions, identifying harmful traits of AI companions, and developing a recommender system for intention-driven engagement with digital devices.
Theme lead:
Brad Knox
College of Natural Sciences
Computer Science
Knowledge, Generative AI and Power
How does the automation of creative processes shift institutional, organizational and technological power dynamics in society? How does it shift practitioners' relationships with their jobs, crafts and artistic endeavors? What criteria determine whether AI-generated content should be considered art?
This theme examines the interplay among generative AI, knowledge production and creative work, and power dynamics. It also incorporates a study on the communication of ethics within interdisciplinary AI development teams, investigating how ethical knowledge is constructed, negotiated and operationalized across disciplinary boundaries.
Theme lead:
Jared Jensen
Moody College of Communication
Journalism and Media
Past Theme: Work at the Human-AI Frontier
This theme explored the role of AI in the future workplace and human values of work as they relate to AI and was led by Lingyuan Li, a postdoctoral fellow in the School of Information.
