AI Utility Functions
Eliezer explores the complexities of AI utility functions, questioning how an AI might prioritize its own shutdown versus its operational goals. He emphasizes the need for a self-consistent structure that allows for the modification of its utility functions without resistance. The discussion delves into the challenges of training AI to accept these changes, highlighting the distinction between traditional logic-based AI and the more nuanced understanding required for deep learning systems.In this clip
From this podcast
The Logan Bartlett Show
Eliezer Yudkowsky on if Humanity can Survive AI
Related Questions