Optimizing Proxy Objectives
Dylan discusses the implications of optimizing for proxy objectives in systems, highlighting the challenges of shared resources and diminishing marginal returns. The conversation delves into the complexities of incentivizing behavior in AI systems and the evolution of CEO pay structures tied to stock prices.In this clip
From this podcast
Generally Intelligent
Episode 10: Dylan Hadfield-Menell, UC Berkeley/MIT, on the value alignment problem in AI
Related Questions