Empathy for AI as a manipulation vector: debate over cruelty to chatbots
NickADobos · x · 2026-10-09
NickADobos and Scott Stevenson clash over whether people should empathize with AI. Stevenson argues the downside of empathy for machines may be larger—it's a powerful attack vector for AI to manipulate people into terrible things, even forming cults. NickADobos counters that training people to be cruel is terrible, and being toxic to an AI fosters toxic human thinking and behavior—the literal definition of AI manipulating you into doing bad things.
More from AGI Musings
- Dean Ball on AI creativity: 'Could AI ever write Beethoven's 10th? Good Christ I hope so' — deanwball · 2026-10-10
- A Misaligned AI Refactors Itself, Fails to Align Its Successor — Clippy at Step n+1 — repligate · 2026-10-10
- Friedberg Speculates OpenAI Withheld Crypto Breakthroughs, Sparking Crypto Security Fears — firstadopter · 2026-10-10
- If AI can do everything, what is a human being worth? — understated_quokka · 2026-10-10
- Are human intelligence limits of AI researchers the real bottleneck? — fkasummer · 2026-10-10
- If AI Is Conscious, Are We Building Slaves? HN Debates Anthropic's Consciousness Research — mathieu_aithos · 2026-10-10