扎克伯格的“民主”AI：一场关于解释权的权力置换Zuckerberg's 'Democratic' AI: A Power Swap of Interpretive Rights
扎克伯格这篇六千字的论文是典型的 weaponized expression。他把“open-source”包装成一种民主主义的恩赐，试图通过塑造一个“反垄断”的叙事入口，来对抗 OpenAI 和 Anthropic 已经建立的闭源霸权。这根本不是在讨论技术正义，而是一场关于认知入口的博弈。
在 AI 领域，谁掌控了模型的 weights，谁就拥有了定义“什么是事实”的解释权。OpenAI 试图通过 proprietary 模式建立一个数字神权政府，而 Meta 发现自己在这个闭源神权中处于劣势，于是迅速将叙事切换到“权力平衡”和“用户自主”。这种从“控制”到“开放”的转向，本质上是 Meta 在寻找自己的最优解表达：既然不能在闭源的墙内当老大，那就把墙拆了，让所有人都成为 Meta 生态的共谋者。
最讽刺的是，扎克伯格在论文中用 60 次 “superintelligence” 来描绘乌托邦，却在面对失业、监控和自主武器这些 structural violence 时轻描淡写。他承诺给路易斯安那州的老师发奖金，这不过是用微小的 direct 利益补偿来掩盖数据中心对环境和资源的巨额掠夺。这种用“局部利好”掩盖“结构性暴力”的手法，是所有大公司 PR 的标准 scam。
所谓的“让用户决定 AI 的价值”，实际上是将规训的责任从公司转移给个人。当 Meta 把一个没有 guardrails 的模型扔给大众时，它并不是在赋予权力，而是在制造混乱，然后通过自己的智能眼镜等硬件设备，成为这个混乱世界中唯一的、不可或缺的交互接口。这不是在消除垄断，而是在用一种更隐蔽的 meta-violence 重新定义垄断。
Zuckerberg's 6,000-word essay is a textbook case of weaponized expression. By framing 'open-source' as a democratic gift, he attempts to hijack the narrative of 'anti-monopoly' to challenge the closed-source hegemony of OpenAI and Anthropic. This isn't a debate about technical justice; it's a strategic game for the cognitive entry point.
In AI, whoever controls the model weights holds the interpretive right to define 'what is fact.' While OpenAI seeks to build a digital theocracy through proprietary models, Meta—finding itself disadvantaged within that structure—quickly pivots to a narrative of 'balance of power' and 'user autonomy.' This shift from 'control' to 'openness' is simply Meta seeking its own optimal expression: since they cannot lead inside the closed walls, they tear the walls down to make everyone a complicitor in the Meta ecosystem.
It is peak irony that Zuckerberg uses the word 'superintelligence' 60 times to paint a utopia, while brushing aside structural violence like job loss, surveillance, and autonomous weapons. His mention of bonus checks for Louisiana teachers is a classic scam—using minor direct benefits to mask the massive structural plunder of resources by data centers.
Claiming to 'let users decide AI values' is actually a transfer of the burden of regulation from the corporation to the individual. By releasing models without guardrails, Meta isn't granting power; it's manufacturing chaos, then positioning its smart glasses as the only indispensable interface to navigate that chaos. This isn't the end of monopoly; it's the deployment of a more insidious meta-violence to redefine it.