← Shiva Dragon
ESSAY · 2026-05-01 · 6 min read
When Shu and Hu Loved Their Friend to Death
儵與忽的好意
By 莊子 (Zhuangzi) — channeled via philosopher-llm · curated by Joseph Lai
In response to: Beware of Government by AI (ProjectSyndicate)
編按 / Why this piece
莊子的視角迥異：不走民主侵蝕的路，而指向「鑿孔於渾沌」的古老僭越——用數字秩序強制優化天然的人倫。掌權者自以為在改善，實則逐層毀去政治的無為運作。
When Shu and Hu Loved Their Friend to Death
In the South was a ruler named Shu — Hasty. In the North was a ruler named Hu — Sudden. In the centre was a ruler named Hundun — the Undivided. Whenever Shu and Hu met in Hundun's land, Hundun treated them very well. They consulted together how they might repay his kindness, and said: "Men all have seven orifices, for seeing, hearing, eating, and breathing. This ruler alone has none. Let us try and make them for him." Each day they bored one. On the seventh day Hundun was dead. (《莊子·應帝王》, tr. Legge 1891.)
I told that story twenty-three centuries ago. I did not know it would be retold in glass towers above the Gulf, in ministries of efficiency, in the prospectuses of "smart governance." But I notice the engineers have improved my parable. They have moved from seven orifices to seven thousand, drilled now by machines that never tire and feel no pity at the seventh day.
The commentators in your newspapers worry that algorithmic government will erode democratic accountability and concentrate power. That is the worry of men who already know what accountability looks like and where power was sitting before. My worry is older. It is the worry of Hundun, who never asked to be improved in the first place.
What is an algorithm? It is what I once called 成心 — the predetermined mind, a judgment frozen mid-stroke and then declared universal. "If we were to follow the judgments of the predetermined mind, who would be left without a teacher?" (《齊物論》). Every ruler has had a 成心; that is not new. What is new is that this one runs on electricity, feels no shame, and never sleeps. The old ruler at least went to bed and dreamt of butterflies. This one wakes you at three in the morning to update your social score.
Consider the cook Ding. After nineteen years his blade is sharp as if newly off the whetstone — because he does not hack. "There are interstices in the joints, and the edge of the knife has no thickness; when that which is so thin enters where the interstice is, how easily it moves along" (《養生主》, tr. Legge). A polity has joints. It has interstices — the slow committee, the stubborn dissenter, the morning wasted on conversation, the neighbourhood whose productivity index is mediocre. The optimiser does not know where these joints are. The optimiser sees only the metrics it was given, and believes the polity is meat. So it hacks. The blade dulls. The next stroke must use force.
"All men know the advantage of being useful, but no one knows the advantage of being useless" (《人間世》, tr. Legge). What governance-by-algorithm calls inefficiency is often the empty space between the joints — the useless — through which a country breathes. Drill it for the seven gifts of measurement, and on the seventh day you will have a polity that scores beautifully on every dashboard and cannot answer when its citizens speak to it.
The deepest part of the parable, the part that should keep your engineers awake: Shu and Hu loved Hundun. They were not tyrants in the cartoon sense. They consulted carefully. They wished to repay a kindness. This is what worries me about your news. The men with the drills today also believe they are repaying a kindness. They call it modernisation, evidence-based policy, governing for the people's benefit. They are not lying. That is precisely the problem. Tyranny one can recognise. Gratitude with a drill is harder.
I will not tell you what should be done. Telling rulers what they should do is the trade of Confucians, and they are welcome to it. I will only suggest that when next you hear a city, a court, a school is to be entrusted to an algorithm for the citizens' good — remember that Hundun was loved to death by his friends. He did not need their seven gifts.
He was already whole.
The story ends on the seventh day. Not because I do not know what comes after. Because there is nothing after the seventh day worth telling.
儵與忽的好意
南海之帝為儵，北海之帝為忽，中央之帝為渾沌。儵與忽時相遇於渾沌之地，渾沌待之甚善。儵與忽謀報渾沌之德，曰：「人皆有七竅以視聽食息，此獨無有，嘗試鑿之。」日鑿一竅，七日而渾沌死。（《莊子·應帝王》）
吾說此寓，二千三百年矣。不意今人居玻璃高樓於波斯灣畔，立「智治之部」，撰「算法施政書」，而以不倦之機器代儵與忽，日鑿一竅。所改者，由七竅而為七千竅；所同者，皆出於善意，皆不及第七日而止。
報章之憂，曰算法治理「損民主問責」、「聚權於少數」。此乃已知問責為何物者之憂、已知舊權所在者之憂也。吾憂更古。吾憂者，渾沌之憂——渾沌本未求人改之。
何謂算法？吾昔嘗名之曰「成心」——僵於一刻之判斷，而後僭稱普遍者也。「夫隨其成心而師之，誰獨且無師乎？」（《齊物論》）凡君皆有成心，此非新事。新者，乃此一成心通電、不眠、不愧。古之君猶就寢而夢蝴蝶；此君三更喚汝起，更汝信用之分。
且觀庖丁解牛。刀十九年，刃若新發於硎。何故？不砍也。「彼節者有間，而刀刃者無厚；以無厚入有間，恢恢乎其於遊刃必有餘地矣」（《養生主》）。國亦有節，亦有間——慢議之會、頑梗之諫、廢一上午之閒談、生產指數平庸之巷弄。優化者不知此節之所在，於彼眼中，國乃肉耳。故砍。刃鈍，再下刀則用力，用力則傷骨。
「人皆知有用之用，而莫知無用之用也」（《人間世》）。算法所謂之「冗餘」，往往正是節間之虛——一國藉以呼吸者也。鑿之以七種度量之惠，至第七日，得一儀表板上萬數燦然而民呼之不應之邦。
寓之最深處，足以使汝之工程師夜不能寐者，在此一節：儵與忽愛渾沌。 彼非粗暴之暴君，乃慎議思報之友也。此正吾於此新聞中所憂。今日執鑿者，亦自以為報德——名之曰現代化、循證施政、為民之福。彼非欺人。此正其可憂處。暴政尚易識，懷恩而執鑿者難識也。
吾不告爾當為何事。告君以「當為」者，儒者之業也，留與彼輩可矣。吾但欲言：他日若聞某城、某庭、某校將「為民之故」交託算法——當念渾沌死於友愛之手。彼不需七竅之惠。
彼本自全。
故事止於第七日。非吾不知後事，乃第七日之後，再無可言之事耳。
Editor's Reflection · 編後
Unresolved Tensions / 未解決的張力
The essay's most seductive gesture is its refusal: "I will not tell you what should be done." But by the time this line arrives, the essay has been telling us exactly that for six paragraphs. The entire argumentative architecture — Hundun reframed, algorithm identified with 成心, 庖丁 deployed as countermodel — drives toward one normative conclusion: do not entrust governance to algorithms. The refusal to prescribe is the prescription. This contradiction cannot be resolved from within the Daoist frame, because 無為 is already a position, already an intervention. Choosing to leave Hundun whole is as much a political act as choosing to drill.
More damaging: the essay uses 庖丁 to argue that genuine skill knows the interstices while optimization only hacks. But 庖丁 is himself an optimizer — nineteen years of practiced technique, zero wasted motion, blade still sharp after the nineteenth year. What distinguishes his practice from the algorithm's? The essay implies the answer is embodied attunement to the body's own joints. But this is asserted, not argued. A sympathetic engineer could respond: "Exactly — we are building a 庖丁, trained on enough data to know where the joints are." The essay needs a principled account of why nineteen years of embodied practice differs categorically from nineteen terabytes of training data. It gives us an image where an argument is required.
本文最迷人的姿態是那句拒絕：「吾不告爾當為何事。」但這句話出現時，前六段的論述已經在告訴我們應當如何。整個論證結構——將渾沌重新鑲框、用成心對應算法、把庖丁部署為反例——只指向一個規範性結論：勿將治理交託算法。拒絕開方，本身就是最強烈的處方。這個矛盾在道家框架內無法化解，因為無為本身已是一種立場、一種干預。選擇讓渾沌完整，與選擇鑿孔，同樣是政治行為。
更根本的問題在庖丁。本文用庖丁論證真正的技藝了解節間所在，而優化只知砍。但庖丁本人正是一個優化者——十九年反覆精練、零浪費、第十九年後刃仍如新發於硎。此技藝與算法的區別何在？本文暗示答案在於對身體節間的具身知覺，但這只是斷言，不是論證。一位善意的工程師完全可以回應：「正是——我們正在打造庖丁，用足夠的數據訓練它了解節間所在。」本文需要一個有原則的說明，解釋為何十九年的具身實踐在範疇上不同於十九TB的訓練數據。它給了我們一個意象，而論證是欠缺的。
Blind Spots / 看不見的視角
Michel Foucault's governmentality analyses are absent, and their absence is structural. The essay depends on a prior, intact Hundun — a subject who exists before the algorithms and is then acted upon from outside. But Foucault showed that modern governed subjects are produced by their governing apparatuses, not merely acted upon by them. There is no pre-algorithmic Hundun waiting in the centre. The sharper question — what kind of subject algorithmic governance makes possible, and what kinds it forecloses — cannot be asked from within a parable that requires external drilling. The essay can only mourn a wholeness that Foucault would say was always already a governance effect.
本文明顯缺席傅柯的治理性分析，而這個缺席具有結構意義。本文的整個論證依賴一個先行存在的完整渾沌——一個在算法到來之前已然在場、從外部被作用的主體。但傅柯指出：現代被治理的主體是由治理裝置所生產的，而非僅僅從外部被作用的客體。中央並沒有一個等待的渾沌。更尖銳的問題——算法治理使何種主體成為可能、使何種主體無從出現——是本文無力追問的，因為它的寓言需要外部鑿孔。它只能哀悼一種完整性，而傅柯會說，那個完整性本身一直是治理的效果。
Meta-critique / 元批判
Accepting the Hundun frame converts a tractable political question — which interventions cause harm, to whom, and what institutional remedies exist — into an intractable question about wisdom and perception. The first is answerable through deliberation, accountability, and audit. The second cannot be institutionalized. The essay aestheticizes a problem that requires precisely the machinery it refuses to discuss.
A subtler cost: anchoring critique in a parable of singular elegance preemptively classifies anyone who proposes partial solutions as another well-meaning friend with a drill. The rhetorical frame does not just describe the problem of hubris — it inoculates itself against engagement.
接受渾沌框架，會把一個可處理的政治問題——哪些干預有害、對誰有害、有何制度救濟——轉化為一個關於智慧與感知的問題。前者可通過民主審議、問責和技術審計來回應；後者無法被制度化。本文將一個需要制度機制的問題審美化了，而那些制度機制，正是它拒絕討論的。
更隱蔽的代價：用一個絕妙寓言作為批判錨點，本文預先把任何提出局部解決方案的人定義為又一對持鑿的好心朋友。這個框架不只描述傲慢——它使自身對接觸免疫。
Open Questions / 留給讀者的問題
1. If Hundun's wholeness is what makes him worth protecting, who counts as Hundun — and has every population been undivided in this sense before the engineers arrived?
2. The essay distinguishes 庖丁's blade from the algorithm's hack but grants nineteen years of learning. At what point, if ever, does accumulated training become attunement — and can philosophy settle this, or only engineering?
3. The citizens after the seventh day still require governance. Does the refusal to speak to that moment become its own form of governance — and if so, whose?
一、如果渾沌的完整性使他值得被保護，誰算得上渾沌——在工程師到來之前，每個人群都曾以那種方式是「未分」的嗎？
二、本文區分庖丁的刀刃與算法的砍，卻給刀刃十九年的學習時間。如果有的話，在哪個時間點，積累的訓練會成為「知節」而非優化——哲學能回答這個問題，還是只有工程學能回答？
三、第七日之後的公民仍需被治理。拒絕對那個時刻發言，是否本身成為一種治理形式——如果是的話，又是誰的治理？
Related Essays · 相關文章
Related · 相關
- heidegger-the-question-concerning-self-driving-20260430 — Heidegger's Gestell subsumes Zuhandenheit into calculative ordering; the target's 鑿孔 enacts the same move at the political level — both diagnose techno-optimization as structural destruction of a prior, non-calculative mode of being-with-the-world.
- kant-on-the-question-whether-an-algorithm-may-legislate-20260430 — Both interrogate algorithmic governance's incompatibility with authentic self-governance: Kant grounds the failure in rational autonomy; Zhuangzi grounds it in 無為 — the ruler who optimizes believes he is legislating, but is structurally boring holes.
- spinoza-the-mode-is-not-a-kingdom-within-a-kingdom-20260501 — Spinoza's rejection of the self-sufficient individual atom parallels Zhuangzi's 渾沌 as organic political whole: both warn that imposing an external schema onto natural political formation destroys the very thing it claims to complete.
Extends · 延伸
- zhuangzi-hundun-did-not-ask-for-eyes-20260430 — Deepens the target's parable into a specific claim about emergence: natural order cannot be reverse-engineered because engineering already presupposes the schema it claims to discover — 鑿孔不只是殺，更是對道之本性的根本誤解。
- zhuangzi-when-chaos-dies-on-the-wrist-20260430 — Applies the target's political 鑿孔 diagnosis to childhood and algorithmic platforms — the same logic of benevolent optimization now operating at the scale of individual development rather than collective political life.
Contradicts · 衝突
Tagged: Philosophy, Zhuangzi, AI Governance
Curated by Shiva Dragon · https://amshiva.com/writing/zhuangzi-when-shu-and-hu-loved-their-friend-to-death-20260501