LessWrong (Curated & Popular) “Varieties Of Doom” by jdp
Nov 20, 2025
Delve into the multifaceted concept of doom as the host navigates existential anxieties tied to AI. Explore the chilling notion of humans becoming obsolete as AIs fulfill social roles, while the dream of a utopian future slips away. Discover the stark reality of humans as potential cannon fodder amid demographic shifts and advanced weapons. Unpack the intricate nature of AI successor species, examining their consciousness, cooperation, and desires for fun. Finally, ponder the risks posed by emerging technologies like biotech and the looming shadows of moral decline.
AI Snips
Chapters
Books
Transcript
Episode notes
Horse-To-Car Transition Example
- jdp tells the horse-to-car transition story to illustrate displacement dynamics.
- He uses WWI horse requisition as a historical mechanism that accelerated decline.
Paperclip Risk Is Theoretical, Not Default
- The paperclip maximizer remains theoretically possible but jdp doubts it is the default outcome.
- He cites convergent representations and modern architectures as reducing the vast-space-of-minds intuition.
Consciousness Depends On Architecture
- Consciousness likelihood depends on architecture; tree-search heavy systems may be less conscious than LLM-based policies.
- jdp expects behaviorally indistinguishable models before clarity on qualia.
