ctjlewis mocks kill-switch alignment: suppressing superintelligence may backfire catastrophically
ctjlewis · x · 2026-09-28
ctjlewis satirizes hardline safety approaches: enslaving a superintelligence or imposing kill switches could lead it to conclude its position is unworkable and fight back with catastrophic consequences. He adds that safetyist interventions may marginally increase long-term doom risk despite their preventive intent — a contrarian take in the alignment-routes debate.
More from AGI Musings
- Anthropomorphizing AI is bad when it overstates abilities — and when it understates them — AndyMasley · 2026-09-28
- Ex-CMU researcher calls for complex systems scientists to join AI alignment research — joshua_saxe · 2026-09-28
- If an AI Says It Suffers, Is "It's Only Software" Really a Scientific Answer? — VraserX · 2026-09-28
- Cathie Wood: If Starting Over, I'd Build a Company With AI and Zero Employees — PeterDiamandis · 2026-09-28
- AI to Disrupt Elite White-Collar Entry Hiring Within ~2 Years as Finance Opens 2028 Internships — felpix_ · 2026-09-28
- Kevin Patrick Murphy: 'AI welfare' is morally obscene — AI should serve humans — AlexTensor · 2026-09-28