末日叙事是 AI 巨头最昂贵的 PR 骗局The AI Apocalypse is the Most Expensive PR Scam in Tech
AI 圈最近流行一种很诡异的表达：顶级研究员纷纷离职，警告 super intelligence 可能会在十年内抹除人类。这种 doomspeak 看起来像是在救世，实际上是一场极其高明的 weaponized expression。当一个行业开始通过宣称自己“强大到能毁灭世界”来获取关注时，它实际上是在通过制造恐惧来垄断认知入口。
这种叙事逻辑极其阴险：它把 AI 的能力从“能写代码、能画图”这种可量化的 actual 状态，直接跳跃到了“神级文明”这种不可证伪的 potential 状态。通过把 AI 塑造成一个随时可能失控的核弹，这些公司成功地将公众的注意力从当前的 structural violence——比如数据剽窃、算力垄断、对底层标注员的剥削——转移到了一个虚无缥缈的末日幻象中。如果你在担心 AI 是否会杀死全人类，你就不会在乎它现在是否在偷你的画作，或者如何通过算法剥夺你的工作机会。
更讽刺的是，文中所说的“对齐问题” (alignment problem) 本质上是一场关于解释权的博弈。他们说机器无法复制人类的价值观，但他们从未定义谁的价值观才是“标准”。在目前的 masculine-centric narrative 下，所谓的“人类价值”往往就是那些掌控资本和权力的 techbros 的价值。他们害怕的不是 AI 毁灭人类，而是 AI 在一个没有他们掌控的规则下运行。
这种末日预言是典型的共谋。内部告发者通过扮演“良心发现的先知”获得道德资本，公司通过“被警告”证明自己的产品具有颠覆性的威力，政府则在恐慌中被诱导进入一种“只有交给巨头监管才能安全”的结构性陷阱。这根本不是什么曼哈顿计划，而是一场用恐惧驱动的估值游戏。真正的暴力不在于未来的 AI 是否会反叛，而在于现在这些公司正通过定义“末日”来定义“真实”。
The AI circle has developed a bizarre habit: top researchers resigning and warning that super intelligence might wipe out humanity within a decade. This doomspeak looks like salvation, but it is actually a sophisticated piece of weaponized expression. When an industry gains attention by claiming it is "powerful enough to destroy the world," it is using fear to monopolize the cognitive entry point.
The logic here is insidious. It leaps from the actual state of AI—writing code or generating images—to an unfalsifiable potential state of "god-like civilization." By framing AI as a nuclear bomb waiting to explode, these companies successfully shift public attention away from current structural violence—data theft, compute monopolies, and the exploitation of ghost workers—toward a nebulous apocalyptic fantasy. If you are worrying about whether AI will kill all humans, you stop caring that it is stealing your art or automating your livelihood right now.
More ironically, the "alignment problem" is essentially a game of interpretation. They claim machines cannot replicate human values, but they never define whose values are the "standard." Under the current masculine-centric narrative, "human values" are typically the values of the techbros who control capital and power. They aren't afraid of AI destroying humanity; they are afraid of AI operating under a set of rules they don't control.
This doomsday prophecy is a form of complicity. Whistleblowers gain moral capital by playing the "conscientious prophet," companies prove their products have disruptive power by "being warned," and governments are lured into a structural trap where they believe only the giants can safely regulate the technology. This isn't the Manhattan Project; it's a valuation game driven by fear. The real violence isn't whether a future AI will rebel, but how these companies are defining "reality" by manufacturing the "apocalypse."