Defining Goals in AI
The discussion delves into the nature of goals in AI, questioning whether ascribing goals to agents is meaningful if those goals are inconsistent. Nora advocates for a pragmatic approach, suggesting that if goals are constantly changing, it may be more useful to focus on behavioral patterns instead. Tim raises concerns about the tendency to conflate the design of AI systems with the descriptions of their goals, emphasizing the importance of building the actual systems rather than merely crafting theoretical frameworks.In this clip
From this podcast
Machine Learning Street Talk (MLST)
Nora Belrose - AI Development, Safety, and Meaning
Related Questions