Gary Marcus vs. Statistician: Is AI Danger Uniquely Different or Just Amplified Human Capability?
GaryMarcus · x · 2026-09-12
Statistician Kareem Carr argues AI isn't uniquely dangerous — like giving everyone ten million dollars, any boost to human capability increases risks such as bioterrorism, so AI safety arguments must identify what's specifically dangerous about AI.
Gary Marcus counters: 'AI is not UNIQUELY dangerous, it is dangerous in unique ways — and still a serious problem.' The exchange hits a core debate in AI safety: whether AI risk is ordinary amplification of capability or qualitatively different.
Related event: Statisticians and Gary Marcus Debate Whether AI Risks Are Unique(2 posts)→
More from AGI Musings
- AGI race debate: controllable superweapon logic vs default-doom framing, strawmen abound — teortaxesTex · 2026-09-12
- The AI exponential is really many cascading sigmoids, argues researcher — adhamelarabawy · 2026-09-12
- Founder: AI Is the Only Thing Propping Up the US Economy Right Now — bindureddy · 2026-09-12
- A brief history of AI alarm hype cycles: GPT-2 to GPT-6, every wave a departure — StewartalsopIII · 2026-09-12
- Gary Marcus resurfaces Pynchon's 1984 warning on AI, bio and robotics curves converging — GaryMarcus · 2026-09-12
- Automated forecasting system puts AI-caused catastrophe odds at 0.47% by 2030 — emollick · 2026-09-12