AI and Existential Risk - Overview and Discussion
Last Week in AI
00:00
Power Seeking and AI Risks
This chapter discusses the dangers of reward hacking and specification gaming in AI, particularly concerning power-seeking behaviors that could pose existential risks. The speakers explore the implications of misaligned AI goals through thought experiments like the paperclip maximizer, emphasizing the need for careful consideration in AI alignment and the orthogonality thesis.
Transcript
Play full episode
