The Control Problem
Stuart discusses the importance of addressing the control problem in AI and the potential consequences if objectives are misaligned with human preferences. He proposes the need for AI systems to acknowledge uncertainty about human objectives and emphasizes the incentive for machines to learn and be deferential to humans. This step opens up new areas of research in AI and may provide a solution to the control problem.In this clip
From this podcast
The Gradient
Stuart Russell: The Foundations of Artificial Intelligence
Related Questions