Non DubitoA Mind in Many Languages
Search搜索搜尋
EN
日本語FrançaisDeutschEspañol한국어

When Switching Stops Being a Choice为什么越来越难换一家為什麼越來越難換一家 · 04

A Good Reason Is Not an Exemption

为了发展,就先别问了?

為了發展,就先別問了?

Innovation, efficiency, competition and safety can be genuine concerns. None can replace an account of the particular power being questioned.

创新、效率、竞争与安全,都可能是真的理由。真正要分辨的是,它们怎样回答具体问题,还是把问题换掉。

創新、效率、競爭與安全,都可能是真的理由。真正要分辨的是,它們怎樣回答具體問題,還是把問題換掉。

Han Qin ·

At the hearing, Beacon and Stone Granary warn that a review leading to separation would leave the Kingdom State two years behind. Competitors abroad receive computing subsidies. The firms' models perform well, and nine thousand jobs belong to the state's model industry. This is not a weightless advertising slogan. Something valuable may genuinely be at stake in the arrangement being questioned.

A legislator asks who would fall behind. The representative answers: the Kingdom State. The phrase grows larger; the question remains. The state contains model developers, workshops, clinics, employees, and customers. What each gains or loses does not become the same account merely because they share a collective name. A national label cannot perform the work of distinguishing their interests.

This is an imagined hearing, concerned with the firms' shared engineering team and partially combined accounts. We do not know that separation would help, or that the estimate is right. Examine the response itself: does identifying a valuable benefit answer whether this particular arrangement must be accepted, or does it move the discussion away from the part actually being challenged?

Many familiar defenses contain truths. Research needs money, scale can reduce costs, international competition involves timing, and technology can be hard to understand. People offering those reasons need not be presumed dishonest. Sometimes the difficulty is precisely that the reason is genuine while failing to address the relevant exercise of power. Exposing a lie is not the only possible form of criticism.

Concern that review can slow innovation does not establish which innovation this review would slow. Acknowledging the role of scale does not establish that combined accounts, a preferred interface, or an exclusive term are all indispensable. Large words need to be broken into effects that can be examined. Otherwise every disagreement becomes an artificial choice between supporting progress and obstructing it.

Which expenses does the joint engineering team save? What would separate accounts cost? Could computational collaboration remain while customers' scheduling decisions become independently examinable? There may be no easy answer. Concrete alternatives nevertheless reveal who bears costs, which benefits depend on which arrangements, and whether another option is actually worse. Naming a benefit cannot establish the whole chain by itself.

This burden of explanation is not reserved for companies. Those proposing review, limits, or separation must identify the effects they address and the production or services they might damage. Rejecting innovation as an exemption cannot mean allowing a regulator to use antimonopoly as its own exemption. A remedy needs an account of its costs, alternatives, and particular justification too.

The hearing receives a Stone Granary-funded study estimating a thirty-percent reduction in investment after separation. Funding by the party under review is relevant without establishing that the estimate is false. Examine methods, assumptions, and evidence, including whether an independent party can check them. Funding provenance should begin an inquiry, not become a shortcut past the substance of the work.

Conversely, if this is the committee's only evidence and even the means of checking it must come from Stone Granary, dependence is not merely financial. The earlier problem of reviewing capacity appears on the side of reasons. Can the questioners independently understand the account they receive, or must they choose between trusting it and adopting unsupported suspicion as a substitute for inquiry?

A different defense borrows the past. Crossways explains that it found buyers for small workshops and waived their first year's fees. Both are true. Neither answers why house products receive prominence today. Acknowledging an enterprise's contribution is easy enough; recognizing that contribution does not contain a perpetual exemption is the further distinction that a useful history cannot perform for us.

Early investment can earn a return without making that return unlimited in duration or scope. An early English statutory arrangement preserved a patent exception of fourteen years or less for specified new manufactures and their first inventors. It is no ready-made policy for every modern market. It does illustrate that recognizing invention and declining permanent exclusivity can coexist within an institutional design.

This is not ingratitude toward founders. Earned rewards have their grounds; new restrictions need theirs. Neither account should consume the other. Recording what an enterprise actually achieved and examining its present arrangements are not opposing tasks. Gratitude cannot establish why a different restriction is necessary, and criticism need not deny the work for which gratitude was originally appropriate.

Too complex for outsiders draws another boundary around the discussion. Complexity may be real, calling for reviewers with time, people, and testing capacity. It cannot on its own create a territory only the firm may understand or approve. An inability to understand identifies conditions to improve; it is not a natural full stop ending every possible route of examination.

Develop first, govern later contains a similar difficulty. Rules fixed too early can suppress things not yet developed. Waiting until everybody relies on one entrance may leave no usable alternatives. When detailed rules should be settled is a legitimate question. It does not establish that, until then, nobody must explain what an arrangement is already doing to the people passing through it.

Being questioned and being punished are especially important to distinguish. Absence of an established penalty does not remove the need to answer. Explanation requires evidence and reasons; punishment, compulsion, and remedies require their own legal and procedural grounds. We cannot use questionability to convict in advance, or absence of conviction to prevent a question from being put at all.

Safety requires particular care because actual people may need protection. In the story, model-generated fake reviews damage shops and mislead buyers. The state subsequently requires approval before models above a threshold may be released. Seeing a new entrance requirement does not justify pretending the harm never occurred or that every concern about model misuse is a disguised commercial maneuver.

The gate nevertheless has costs. Beacon's safety director chairs the assessment group. Only two providers can conduct the required tests, at a price approaching an open-model group's annual budget. The group stops releases. It acknowledges that its model was used for fake reviews while asking whether the new route is one only large firms can afford to enter.

Separate the inquiries. What activity is restricted? Who is protected? Why is this particular restriction necessary? How far does it reach? Measures addressing malicious users, website filtering, and identifying machine-generated material might protect some of the same people, but their feasibility must be examined. A casually named alternative is no more proof than a casually invoked safety requirement.

Reported fake reviews fall by fifty-five percent. That decline is part of the imagined record. The account does not say whether websites changed their filters at the same time or establish how much the new rule caused. A favorable outcome deserves attention without independently proving necessity. The distinction matters especially when one group bears a major entry cost justified by that outcome.

Who judges cannot be omitted. Assessment needs functional independence from the beneficiary of the restriction, and those restricted need a route for questioning necessity and scope. The chair's dual role is a definite connection, not automatic proof that one company controls all seven members. Actual authority, recusal, and the acquisition of evidence still require examination before a conclusion is reached.

This is more than asking whether somebody's opinion seems fair. It concerns whether the judgment can be reopened. The party proposing or enforcing a restriction cannot possess sole authority to cancel its review. Recusal, disclosure, and other conflict arrangements need concrete design. Printing the word independent cannot complete that work, any more than a published rule certifies fair application.

The example therefore establishes neither that the safety measure is excessive nor that it is justified. It preserves two realities: fake reviews cause harm, and an entry requirement imposes substantial costs. Judgment needs to connect both to examinable necessity, scope, and independent assessment. Asking first whether one sides with large firms or small developers skips the work the case requires.

For your own good is particularly good at making this distinction disappear. Someone deciding what protects another person may be fulfilling a genuine responsibility. But when the affected person can never question the judgment, protection may leave them waiting entirely on somebody else's goodwill. Safety is not the objection. The objection concerns what invoking it removes from the reach of challenge.

Innovation, efficiency, and safety need not be exposed as false before discussion can continue. A stricter reading allows them to be true and still asks what effect the reason addresses, who pays, what alternatives exist, and who can examine the answer. Genuine reasons have scope. They are not permissions to substitute a valued word for every explanation an arrangement requires.

The story does not supply the hearing's eventual vote. We need not finish it for the legislators. Leave the unanswered question on the table instead: when you say the whole state would fall behind, whose particular possibilities would be lost, and why must this particular arrangement preserve them? Setting the collective name down lets actual people reappear inside the reasons offered on their behalf.

听证会上,灯塔与石仓的代表说,如果审查使两家分开,王国州的模型业会落后两年。别处有竞争者,有政府给的算力支持。它们的模型也确实做得好,州内九千份工作与这一行有关。这一答,不是随便一句没有重量的广告。

委员追问,落后两年,落后的是谁?代表答,是王国州。话说得更大,问题却还在那里。州里有模型公司,也有用模型的诊所、作坊、雇员和买家。他们得到和失去的,不会只因装进一个总称,就自动成为同一笔账。

这里仍用虚构的王国州。提案问的是两家合用团队、合并部分账目是否该受审查。我们并不知道分开一定好,也不知道两年的估计对不对。散文要读的,是这一类答复怎样起作用:有一件值得珍惜的好处,是否就可以先不问取得它的安排?

许多说法有真实的一面。研发要钱,规模可能降低成本,国际竞争有时间窗口,技术也确实难懂。说这些话的人,不必被预先想成坏人。有时最难的问题,恰恰是理由真的成立,但它尚未答到被问的那一件。

我们担心过度审查会拖慢创新,不等于已经知道这一次审查会拖慢哪项创新。我们承认做大能投入,不等于这份合账、这个优先接口、这项独占条款全都不可少。要把大词拆到能核的作用上,争论才不只剩支持进步或反对进步两边站。

例如,合用训练团队省了哪部分费用,分开账目又会损失什么?有没有办法保留计算上的合作,让不同客户的排期仍能独立核查?这些问题未必有轻巧的答案,但落到具体办法以后,才看得见谁承受代价,以及别的办法是否真的更差。

这份说明不是只问公司。提出审查、限制或分离的一方也要答,它处理的是哪一项作用,是否损害生产与服务,损失怎样承担。不能一边反对企业用发展免问,一边让改变安排的机构只靠反垄断三个字就免于说明。

听证会上还有一份石仓出钱的研究,估计分开会使投资少三成。付钱的人恰好是被审查的一方,值得注意,不等于数字就假。要看方法、假设和数据能否由不依赖它的一方复核。资金来源不是一条跳过内容的捷径。

反过来,若委员会手里只有这一份材料,连检验的方法也只能请石仓提供,依赖就不只在钱上。先前说的审查能力,在理由这一侧也有影子:问者能否自己看清别人交来的理由,而不是只能在信任与怀疑之间挑一种态度?

还有一种答复,借的不是未来,而是过去。通衢说,它当年替小作坊找买家,免过佣金。都是真的。可若问题是今天自营商品为什么排在前面,这段创业史仍没有回答。说清早年的贡献,不难;承认贡献不包含永久豁免,才是要做的分别。

早期的投入可以得到回报,并不等于回报没有期限或范围。英格兰早期的《垄断法》曾为特定新制造的首创者保留十四年以内的专利特许例外。这个历史安排不能直接搬来处理今天所有市场,它至少让人看见,承认首创与不给永久独占,可以同在一种设计里。

所以问题不是对创业的人薄情。已经挣得的报偿有它自己的根据,新的限制也要有新的根据。它们不能彼此吞掉。把一个人或一家企业曾经做成的事照实记下,与继续问今天的安排,没有什么必然冲突。

太复杂,外人管不了,则把难处换成了另一种界线。复杂可能是真的,因此审查者需要人、时间和测试能力。它不能单凭复杂就成为只能由企业自己理解、自己批准的领地。看不懂,是该补的条件,不是终止追问的天然句号。

先发展、后治理也有相似的一面。过早定死规则,确实可能把还没长出的东西压住;等所有人都依赖同一个入口,再问怎么办,也可能已经没有可用替代。讨论规则何时细化,不等于在那以前,连说明现在做了什么都可以免掉。

受问与受罚尤其要分开。一项新做法没有现成的处罚条款,不等于提出问题的人不该得到回答。回答需要材料与理由;处罚、强制与救济仍要它们自己的法律和程序根据。不能借可问提前判罪,也不能借未定罪封住可问。

安全的名义还要更仔细一些,因为它护着的可能是真实受伤的人。故事里,模型生成的假评价损害店铺,也误导买家。后来州规定,某个规模以上的模型发布前必须认可。不能只见门槛,就把假评价造成的损害当作不存在。

但那道门有自己的代价。灯塔安全部门的负责人兼任评审组长,认可测试只有两家做得了,费用约等于一个公开模型小组一年的经费。小组停了发布。它承认自己的模型被拿去写过假评,也问,这道门是不是只有大公司进得去?

要把几问分开:限制哪一种行为,保护谁,为什么非得用这一种办法,限制多宽。针对恶意使用者、平台过滤、标明机器生成内容,是否可能护到同一批人,需要比较,不是随便说一个替代名字就算已经可行。

统计又说假评价少了五成五。少了是故事给的事实,却没有记网站是否同时改了过滤,也没有核多少变化来自新办法。因而这一个数字,不能独自证明门槛必要,或证明没有门槛就做不到。好结果值得看,因果仍须查。

谁来判也不能漏掉。评审要在功能上独立于受益的一方,受限制的人要有一条能问必要性与范围的路。组长兼任是明确的牵连,却不自动证明七个人已经被一家公司控制;实际权力、回避和资料怎样取得,还要继续核。

这不只是看法公不公平,还涉及一份判断有没有可能被重新打开。提出或执行限制的一方,不能独自取消对限制本身的审查。利益冲突采取回避、披露或别的办法,需具体制度设计;把独立二字写上去,并没有替它们完成工作。

因此,这个例子没有宣布安全办法已证过度,也没有宣布已证正当。它保留两面:假评价是真损害,小组承担的进入代价也是真代价。要判办法,得把两面接到可检验的必要性、范围与独立审查上,而不是先问你站大公司还是站小组。

有时,为你好最容易使人忘记这个分别。别人代为判断什么安全、什么有益,或许正履行职责;如果被判断的人永远不能问那份判断,保护就可能变成只剩等待对方善意。问题不在安全这个词不好,而在它是不是替具体权力拿走了可问的位置。

发展也好,效率也好,安全也好,都不必先被我们揭穿成假话,才需要继续谈。真正严格的读法恰恰允许它们是真的,然后仍然问:这份理由处理哪一个作用,谁付代价,有什么替代,谁能复核?真的理由,也有需要说明到哪里才算答到的问题。

那场听证最终怎样表决,故事没有写。我们可以先不替委员们续上结局,只把尚未答完的那一问留在桌面:你说全州会落后,究竟是谁的什么会失去,又为什么必须以现在这份安排来保住它?把总称放下来,人才有机会重新出现在理由里。

聽證會上,燈塔與石倉的代表說,如果審查使兩家分開,王國州的模型業會落後兩年。別處有競爭者,有政府給的算力支持。它們的模型也確實做得好,州內九千份工作與這一行有關。這一答,不是隨便一句沒有重量的廣告。

委員追問,落後兩年,落後的是誰?代表答,是王國州。話說得更大,問題卻還在那裡。州裡有模型公司,也有用模型的診所、作坊、僱員和買家。他們得到和失去的,不會只因裝進一個總稱,就自動成為同一筆帳。

這裡仍用虛構的王國州。提案問的是兩家合用團隊、合併部分帳目是否該受審查。我們並不知道分開一定好,也不知道兩年的估計對不對。散文要讀的,是這一類答覆怎樣起作用:有一件值得珍惜的好處,是否就可以先不問取得它的安排?

許多說法有真實的一面。研發要錢,規模可能降低成本,國際競爭有時間窗口,技術也確實難懂。說這些話的人,不必被預先想成壞人。有時最難的問題,恰恰是理由真的成立,但它尚未答到被問的那一件。

我們擔心過度審查會拖慢創新,不等於已經知道這一次審查會拖慢哪項創新。我們承認做大能投入,不等於這份合帳、這個優先接口、這項獨佔條款全都不可少。要把大詞拆到能核的作用上,爭論才不只剩支持進步或反對進步兩邊站。

例如,合用訓練團隊省了哪部分費用,分開帳目又會損失什麼?有沒有辦法保留計算上的合作,讓不同客戶的排期仍能獨立核查?這些問題未必有輕巧的答案,但落到具體辦法以後,才看得見誰承受代價,以及別的辦法是否真的更差。

這份說明不是只問公司。提出審查、限制或分離的一方也要答,它處理的是哪一項作用,是否損害生產與服務,損失怎樣承擔。不能一邊反對企業用發展免問,一邊讓改變安排的機構只靠反壟斷三個字就免於說明。

聽證會上還有一份石倉出錢的研究,估計分開會使投資少三成。付錢的人恰好是被審查的一方,值得注意,不等於數字就假。要看方法、假設和資料能否由不依賴它的一方複核。資金來源不是一條跳過內容的捷徑。

反過來,若委員會手裡只有這一份材料,連檢驗的方法也只能請石倉提供,依賴就不只在錢上。先前說的審查能力,在理由這一側也有影子:問者能否自己看清別人交來的理由,而不是只能在信任與懷疑之間挑一種態度?

還有一種答覆,借的不是未來,而是過去。通衢說,它當年替小作坊找買家,免過傭金。都是真的。可若問題是今天自營商品為什麼排在前面,這段創業史仍沒有回答。說清早年的貢獻,不難;承認貢獻不包含永久豁免,才是要做的分別。

早期的投入可以得到回報,並不等於回報沒有期限或範圍。英格蘭早期的《壟斷法》曾為特定新製造的首創者保留十四年以內的專利特許例外。這個歷史安排不能直接搬來處理今天所有市場,它至少讓人看見,承認首創與不給永久獨佔,可以同在一種設計裡。

所以問題不是對創業的人薄情。已經掙得的報償有它自己的根據,新的限制也要有新的根據。它們不能彼此吞掉。把一個人或一家企業曾經做成的事照實記下,與繼續問今天的安排,沒有什麼必然衝突。

太複雜,外人管不了,則把難處換成了另一種界線。複雜可能是真的,因此審查者需要人、時間和測試能力。它不能單憑複雜就成為只能由企業自己理解、自己批准的領地。看不懂,是該補的條件,不是終止追問的天然句號。

先發展、後治理也有相似的一面。過早定死規則,確實可能把還沒長出的東西壓住;等所有人都依賴同一個入口,再問怎麼辦,也可能已經沒有可用替代。討論規則何時細化,不等於在那以前,連說明現在做了什麼都可以免掉。

受問與受罰尤其要分開。一項新做法沒有現成的處罰條款,不等於提出問題的人不該得到回答。回答需要材料與理由;處罰、強制與救濟仍要它們自己的法律和程序根據。不能借可問提前判罪,也不能借未定罪封住可問。

安全的名義還要更仔細一些,因為它護著的可能是真實受傷的人。故事裡,模型生成的假評價損害店鋪,也誤導買家。後來州規定,某個規模以上的模型發佈前必須認可。不能只見門檻,就把假評價造成的損害當作不存在。

但那道門有自己的代價。燈塔安全部門的負責人兼任評審組長,認可測試只有兩家做得了,費用約等於一個公開模型小組一年的經費。小組停了發佈。它承認自己的模型被拿去寫過假評,也問,這道門是不是只有大公司進得去?

要把幾問分開:限制哪一種行為,保護誰,為什麼非得用這一種辦法,限制多寬。針對惡意使用者、平台過濾、標明機器生成內容,是否可能護到同一批人,需要比較,不是隨便說一個替代名字就算已經可行。

統計又說假評價少了五成五。少了是故事給的事實,卻沒有記網站是否同時改了過濾,也沒有核多少變化來自新辦法。因而這一個數字,不能獨自證明門檻必要,或證明沒有門檻就做不到。好結果值得看,因果仍須查。

誰來判也不能漏掉。評審要在功能上獨立於受益的一方,受限制的人要有一條能問必要性與範圍的路。組長兼任是明確的牽連,卻不自動證明七個人已經被一家公司控制;實際權力、回避和資料怎樣取得,還要繼續核。

這不只是看法公不公平,還涉及一份判斷有沒有可能被重新打開。提出或執行限制的一方,不能獨自取消對限制本身的審查。利益衝突採取回避、披露或別的辦法,需具體制度設計;把獨立二字寫上去,並沒有替它們完成工作。

因此,這個例子沒有宣佈安全辦法已證過度,也沒有宣佈已證正當。它保留兩面:假評價是真損害,小組承擔的進入代價也是真代價。要判辦法,得把兩面接到可檢驗的必要性、範圍與獨立審查上,而不是先問你站大公司還是站小組。

有時,為你好最容易使人忘記這個分別。別人代為判斷什麼安全、什麼有益,或許正履行職責;如果被判斷的人永遠不能問那份判斷,保護就可能變成只剩等待對方善意。問題不在安全這個詞不好,而在它是不是替具體權力拿走了可問的位置。

發展也好,效率也好,安全也好,都不必先被我們揭穿成假話,才需要繼續談。真正嚴格的讀法恰恰允許它們是真的,然後仍然問:這份理由處理哪一個作用,誰付代價,有什麼替代,誰能複核?真的理由,也有需要說明到哪裡才算答到的問題。

那場聽證最終怎樣表決,故事沒有寫。我們可以先不替委員們續上結局,只把尚未答完的那一問留在桌面:你說全州會落後,究竟是誰的什麼會失去,又為什麼必須以現在這份安排來保住它?把總稱放下來,人才有機會重新出現在理由裡。

Source & reading note原论文与阅读说明原論文與閱讀說明

Nine essays draw on the Monopoly prequel, seven papers and dialogue with earlier thinkers. Kingdom State, Crossways, Beacon and future water provision are imagined, not disguised allegations or forecasts. Linked cases distinguish investigations, allegations, jurisdiction, liability and remedies within their stated institutional and dated scope. Size, success, defaults and investment do not alone establish a violation; benefits do not answer for every exercise of control. Separation, interoperability and standby capacity are discussed as proposals with costs, not proven universal solutions. This is not legal or investment advice. English is independently written.

九篇据《垄断论》前置篇、七篇正文与传统对话篇改写。王国州、通衢、灯塔及未来供水等场景为设想,不是现实企业的化名或对未来的预测。真实案件与制度材料另列来源,区分调查、指控、管辖判断、责任判决与救济;按相应机构和材料日期说明,不以一项法律结论代答全部规范问题。规模、成功、默认设置或投资关系本身不直接证明越线;有益的一面也不替具体控制作答。关于拆分、互通与备用能力的建议不是已经验证的万能方案,不提供法律或投资建议。简繁中文同文,英文独立重写。

九篇據《壟斷論》前置篇、七篇正文與傳統對話篇改寫。王國州、通衢、燈塔及未來供水等場景為設想,不是現實企業的化名或對未來的預測。真實案件與制度材料另列來源,區分調查、指控、管轄判斷、責任判決與救濟;按相應機構和材料日期說明,不以一項法律結論代答全部規範問題。規模、成功、默認設置或投資關係本身不直接證明越線;有益的一面也不替具體控制作答。關於拆分、互通與備用能力的建議不是已經驗證的萬能方案,不提供法律或投資建議。簡繁中文同文,英文獨立重寫。

Read the source paper ↗阅读原论文 ↗閱讀原論文 ↗

Further reading: Statute of Monopolies, section VI进一步阅读:《垄断法》第六条的专利特许例外原文進一步閱讀:《壟斷法》第六條的專利特許例外原文 ↗

Hansard, 1829: the statutory fourteen-year patent exception quoted in Parliament英国议会1829年辩论记录:对第六条十四年专利例外的引述英國議會1829年辯論記錄:對第六條十四年專利例外的引述 ↗