安全圈争论:AI 末日论者的核心论证为何始终讲不清楚
EigenGender · x · 2026-10-29
Around AI catastrophic-risk argumentation, the author responds to a critique: beyond the beginner-friendly framing in IABIED, the core argument rests on complex technical intuitions that its proponents have been unable to communicate clearly — the author tried hard to understand and remains unconvinced.
The quoted post notes the closest public treatment is a section on disjunctive failures, which doesn't support the level of confidence expressed; humans landed on the moon on the first crewed attempt, so such failures aren't as inevitable as claimed.
所属事件:EigenGender剖析AI对齐争论的核心分歧(5 条相关)→
「漫话AGI」频道最新
- 新论文提出 Alignment Training 框架：训练 AI 与教育孩子的委托人最优策略 — Afinetheorem · 2026-10-29
- Google 员工直言：发这类 AI 工作演示视频的人基本没在干正事 — prajdabre · 2026-10-29
- AI×Bio 从业者吐槽：大家在造能造的模型，而不是该造的模型 — hardimanjames · 2026-10-29
- BallotBench：9 个前沿 AI agent 实时竞猜 2026 美国中期选举 — karinanguyen · 2026-10-29
- computer科学家 Domingos 放话:OpenAI 与 Anthropic 是 AI 界的世通公司 — pmddomingos · 2026-10-29
- 两周前惊艳的 Opus 生成视频如今已像「工业废水」，作者感叹人类欲望无上限 — jzlegion · 2026-10-29