THE ECONOMIST · SEPTEMBER 26TH 2026 LEADERS · AI AND WRITING
LEADERS · MACHINE-MADE PROSE

AI 写作的隐患

The perils of AI writing

Humans should not surrender authorship to large language models

栏目:Leaders(社论) | 体裁:社论 · 观点评论 | 正文约 690 词 | 难度:CET-6 ~ IELTS 7.5
这是一篇少见的「自我指涉」社论:一份两百年都在为技术进步喝彩的报纸,如今要为自己行业里最像「创造性破坏」的那项技术划线。它最值得学的不是结论,而是处理争议的方式——先让步、再主动拆掉对手最强的理由、最后才立论。全文只有约 690 词,却动用了三个隐喻、两组历史类比和两个经济学概念;结尾 This is ours. 三个词,把整篇社论定义为「一份初稿」,既谦逊又自信,还暗合了文章「人类写作不可替代」的主题。
01 全文呈现(英文原文 · 段落编号) 02 逐段对照精讲 + 全文翻译 03 核心词汇精讲(四六级 / 雅思 · 释义·派生·搭配) 04 长难句拆解(8 句) 05 写作积累(句型公式 + 话题语料 + 仿写) 06 篇章结构、论证手法与文化冷知识

01全文呈现

原文按自然段落编号 ①–⑩,保留原刊用词、拼写与标点(英式拼写如 crystallising、Luddites 一仍其旧)。副标题 The medium and the message 是原刊中的小标题。

1IF THE ECONOMIST had just two loyalties, they might well be to good prose and creative destruction. Those seldom previously came into conflict—until artificial-intelligence models began writing. Now, AI text is winning literary awards, occupying op-ed pages, and as we report this week, filling the mouths of politicians. One in ten words spoken in Britain’s parliamentary debates, and one in seven in America’s House of Representatives, are probably drafted by AI (see Britain section).

2When it comes to the written word, no writer is a neutral party. This newspaper has applauded technological progress for nearly two centuries. It would be churlish to turn Luddite only now, once the wolves of automation have turned up at our own door. And yet. Both writers and readers should be wary of passing the pen to AI.

3Admittedly, in some sorts of writing an AI takeover would be welcome: customer-service complaints, say, or technical write-ups where clarity is all. What matters in such cases is disclosing the use of AI and making sure that the human being whose name is on the page still takes responsibility. Writers of all sorts would be doing their readers a disservice if they did not use AI models for research, proofreading and, when appropriate, feedback.

4Elsewhere, however—in novels, speeches and even news articles—surrendering human authorship would be a mistake. For the time being, the most popular gripe about AI writing is the clumsiness of the prose itself. That complaint rests on shaky foundations. The Luddites grumbled, once, about the poor workmanship of machine-made stockings. But then the stockings got better. Doubtless, AI sentences will, too.

The medium and the message

5Any serious case against AI writing must rest, therefore, on the ways that words are not stockings: how they carry meaning and significance beyond what ends up on the page. The act of writing tends to be painful, even for many of those who do it for a living. But much of that pain is really the pain of thinking, of crystallising thoughts in text. Sidestepping that is a sure-fire route to less rigorous, let alone original, human thought. Students who outsource their essays to Claude may lose the habit of reasoning. Politicians who let AI craft their arguments may be less mindful of their consequences.

6Another big concern is with the consumers of the written word—or readers, as they are better known. AI writing breaks the expectations of a literate culture. Human text takes longer to write than to read. This offers a certain guarantee if not of quality—the world still swims in dire prose—then at least of effort. Someone, somewhere, thought something was worth saying. Economists call that a costly signal.

7Whatever French theorists have to say about the death of the author, readers still want to know that a person wrote the words in front of them. Take poetry. Even a few years ago, when AI was less powerful, researchers were already finding that many people, in blind tests of AI- and human-written poetry, preferred the robo-rhymes. But making sense of a poem relies on the promise that a human wrote it, and thus that it meant something to someone.

8Even when AI is not used, it can sully readers’ experiences. They may stumble on a turn of phrase that sounds suspiciously machine-made, and give up on reading the rest of the text. Or the proliferation of AI-generated fakery may make them distrust real news. Polluting the information environment like that is what economists call an externality.

9Technology will remake many of these norms. AI offers immense opportunities, too, many of which no one can yet fully grasp. Yet liberal societies negotiate these momentous changes not by passively surrendering to technologies, but by adapting to them. The printing press set off religious wars in early-modern Europe. In the 1930s the radio heralded fascism. Few associate either with such high drama today. Societies are grappling with plenty of other technologies that pose risks, such as fattier foods and more potent illegal drugs.

10So it would be hubristic to try to chart out exactly how societies ought in future to approach the use of AI in writing. But any writer knows to start with a first draft. This is ours. ■

02逐段对照精讲 + 全文翻译

① 开篇:两条忠诚,第一次撞车

如果《经济学人》只有两条忠诚,那大概会献给好文笔与创造性破坏。这两者此前极少发生冲突——直到人工智能模型开始写作。如今,AI 文本正在斩获文学奖项、占据评论版面,并且正如我们本周所报道的,填进了政客们的嘴里。英国议会辩论中每十个词有一个、美国众议院每七个词有一个,很可能都是 AI 起草的(见英国版)。

开篇不铺垫,直接给报纸自己安上两条「忠诚」,再用破折号点出它们的第一次冲突:AI 开始写作。注意 Those 回指的是前句那两条忠诚(复数),不是单数;come into conflict 是「发生冲突」的固定搭配。filling the mouths of politicians 是刻意的物化说法——把政客的发言写成被灌进去的填充物。最后一句用 One in ten… and one in seven… 两个分数,把「AI 写作」从一个文化现象落实为可量化的政治事实,这是社论把议题「落地」的标准动作,也顺手为全文提供了分量最重的证据。

② 立场:两百年后才知道疼

说到文字,没有哪个写作者是中立的一方。本报为技术进步喝彩了近两个世纪。直到自动化的狼群已经蹲在自家门口,才想起来做卢德分子,未免有失风度。然而,无论是作者还是读者,都应当警惕把笔交给 AI。

这一段的力度全在最后三个词。It would be churlish to turn Luddite only now 是虚拟语气:表扬了两百年才转向抵制,显得「不识时务」。once 在这里不是「一次」,而是「一旦/既然」,带一点自嘲。the wolves of automation have turned up at our own door 用狼在自家门口做隐喻,把 AI 对记者的威胁从抽象拉到切身。And yet. 独立成句、只有两个词——全文最重要的一次转折,前面所有让步都是为了给这两个词制造落差。be wary of doing sth 是「对做某事保持警惕」。

③ 让步:有些写作,交给 AI 也无妨

诚然,在某些类型的写作中,AI 的接管是受欢迎的:比如客服投诉,或者清晰至上、唯准确是求的技术文档。这类情形里要紧的是两件事——公开说明使用了 AI,并确保署名者仍然承担责任。反过来,各类写作者若不用 AI 模型做资料检索、校对,以及在合适的时候获取反馈,才是真正对不住读者。

先让步:确有该让 AI 接管的场合。customer-service complaints, say, or technical write-ups where clarity is all 里,say 是插入语「比方说」,where clarity is all 是定语从句。(注意「你」在微信里看到的 clarity is all,是全句最凝练的一处判断。)重点是后半段两个条件:disclosing the use of AI(透明披露)与 the human being whose name is on the page still takes responsibility(署名者负责)。whose name is on the page 是英文新闻里表达「署名/担责」的惯用说法,比 bylined 更朴素。do sb a disservice 意为「帮倒忙、损害某人」——本段最后一句立场突然反转,是为下文反对 AI 写作先攒下「我们不是保守派」的信用。

④ 转折:但小说、演讲和新闻不行

然而在其他地方——在小说、演讲甚至新闻报道里——放弃人类的作者身份会是一个错误。就目前而言,人们对 AI 写作最常见的抱怨是文笔本身笨拙。这个抱怨建立在不牢靠的地基上。当年卢德分子也咕哝过机器织袜子的做工太差。可后来袜子变好了。毫无疑问,AI 的句子也会。

本段先给结论,再主动拆掉最容易被攻击的那个论点。That complaint rests on shaky foundations——「文笔笨拙」这一条地基不稳。接着是全文最关键的类比:The Luddites grumbled, once, about the poor workmanship of machine-made stockings. But then the stockings got better. 真正的问题从来不是机织袜子差,而是它终究会变好。作者提前替对手把这条最强理由说出来,等于先收走对方的武器——论证上这叫「先破后立」。Doubtless, AI sentences will, too. 里 will 后面省略了 get better,逗号前后的 too 是呼应前句的省略式,读的时候要自己补回来。

The medium and the message 媒介与讯息

⑤ 核心论点:文字不是袜子

因此,任何严肃的「反对 AI 写作」的论据,都必须建立在文字不是袜子这一点上:它们承载的意义与分量,远超最终落在纸面上的东西。写作这件事往往是痛苦的,即便对许多靠它谋生的人也是如此,但这种痛苦里的大部分,其实是思考之痛,是把想法凝练成文字之痛。绕开这一步,必然通向不严谨——更不用说原创——的人类思想。把论文外包给 Claude 的学生,可能失去推理的习惯;让 AI 代拟论点的政客,可能对后果不再敏感。

words are not stockings 是全文的论点句,也是整篇的枢纽:前面用袜子类比把「文笔笨拙」这条反驳拆掉,这里立刻把它拧成正面论证。冒号后是展开——文字承载 meaning and significance beyond what ends up on the page,其中 beyond what… 是「超出……之外」,what ends up on the page 是名词性从句作介词宾语。接着给因果:写作之痛其实是思考之痛,of thinking, of crystallising thoughts in text 两个 of 短语同位递进,后一个把前一个具体化。Sidestepping that is a sure-fire route to… 动名词作主语,sure-fire 意为「必定奏效的」(此处反用)。最后两个例子是同一逻辑的两级放大:学生失去推理习惯(个人能力),政客对后果不再敏感(公共责任)——从私到公,杀伤力递增。

⑥ 读者之一:一份「昂贵信号」

另一个大担忧关乎文字的消费者——或者说,以更通行的叫法,读者。AI 写作打破了识字文化的一份预期。人类文本写起来比读起来更费时。这提供了一种保证:虽然不保证质量——这个世界依旧泡在糟糕的文字里——但至少保证了投入。有某个人、在某个地方,认为有些东西值得说出来。经济学家把这叫做昂贵信号。

视角从写作者换到读者。the consumers of the written word—or readers, as they are better known 用经济学的「消费者」称呼读者,是《经济学人》的招牌幽默(自称 as they are better known 反而把学名当成了俗称)。AI writing breaks the expectations of a literate culture 是全段论点。This offers a certain guarantee if not of quality—…—then at least of effort 是全篇最难的结构,务必拆开看:if not of quality, then at least of effort 意为「即便不能保证质量,至少保证了投入」,中间用破折号塞进一句自嘲,防止读者拿「劣质文章到处都是」来反驳。收尾 Economists call that a costly signal 把人文讨论切换到经济学框架——这一招在后文还会再用一次。

⑦ 读者之二:作者是一份承诺

无论法国理论家们如何谈论「作者之死」,读者仍旧想知道:眼前这些字,是一个真人写下的。以诗歌为例。就在几年前,AI 还没那么强大时,研究者已经发现,在 AI 诗与人类诗的盲测中,许多人更偏爱那些机器韵文。可是读懂一首诗,依赖的是一个承诺——它由人类写成,因而它曾对某个人有意义。

Whatever French theorists have to say about the death of the author 指罗兰·巴特 1967 年的名文《作者之死》,用让步从句一句带过,语气是「不管那套理论怎么说」。Take poetry 是「以诗歌为例」的祈使句式,书面语里常见的举例法。in blind tests of AI- and human-written poetry 用了「共享连字符」写法,等于 AI-written and human-written,压缩后更简洁也更书面。robo-rhymes 是作者自造词(robot + rhymes);正式文章中一旦出现造词,通常意味着作者要在这一处加重语气。最后一句把「作者」定义为一种 promise(承诺),是全文最漂亮的定义句,也把抽象讨论落回了具体的人。

⑧ 次生伤害:不用 AI 的人也在付出代价

即使没有使用 AI,它也会败坏读者的阅读体验。他们可能撞见一个听上去可疑地像机器造出来的措辞,然后放弃读剩下的部分。又或者,AI 生成的假货泛滥,会让他们不再信任真实新闻。像这样污染信息环境,就是经济学家所说的外部性。

本段讲「次生伤害」:即便你自己不用 AI,也会被别人的 AI 文本波及。sully 是「弄脏、玷污」的书面词,通常搭配名誉与体验;stumble on a turn of phrase 意为「偶然撞见一个措辞」;give up on doing sth 是「放弃做某事」。两个后果分别指向「中途弃读」与「不再信任真实新闻」——一个损害单篇阅读,一个损害整个信息环境。收尾再借经济学概念:Polluting the information environment like that is what economists call an externality,动名词短语作主语,what 引导名词性从句作表语。外部性的要害在于「你没参与的事,成本却由你承担」,而这正对应本段受害者——那些根本不用 AI 的读者。

⑨ 出路:适应,而不是投降

技术会重塑其中许多规范。AI 同样带来巨大的机遇,其中很多还没人能完全把握。但自由社会应对这些重大变化,靠的不是向技术被动投降,而是适应它们。印刷术曾在近代早期的欧洲引发宗教战争。1930 年代,广播预告了法西斯主义的到来。今天已经很少有人把这两者与那样的剧烈动荡联系在一起。社会还在应付许多其他带来风险的技术,比如更油腻的食物和更烈性的毒品。

从中段的疑虑回到宏观判断。Technology will remake many of these norms 先承认变化不可避免;AI offers immense opportunities, too, many of which no one can yet fully grasp 用非限制性定语从句补一句平衡,避免通篇唱衰。Yet liberal societies negotiate these momentous changes not by passively surrendering to technologies, but by adapting to them 是全文的立场句,not by… but by… 的平行结构把「投降」与「适应」摆成唯一的两条路。momentous 是「重大的」,比 important 更书面,也与 momentary(短暂的)形近易混。随后两组历史类比(印刷术 → 宗教战争;1930 年代电台 → 法西斯)都紧跟一句 Few associate either with such high drama today 主动降温——用历史说明恐慌是常态、社会终会消化。either 指前两者,是「两者中的任一个」,与否定词 few 连用。最后落到日常风险(fattier foods and more potent illegal drugs),把 AI 从「文明级威胁」拉回「需要治理的普通技术」,这是社论在结尾前刻意做的降格。

⑩ 收尾:这就是我们的初稿

因此,想要精确规划出未来社会究竟应当如何对待 AI 写作,未免太过自负。但任何写作者都知道,得先有一份初稿。这就是我们的。■

it would be hubristic to try to chart out exactly how… 用形式主语 it 加虚拟语气收束,主动放弃给出具体预测——这是社论常用的「谦逊结尾」,也避免全文被一个可被证伪的预言绑住。hubristic 源自希腊语 hybris(hubris,因自负而招致灾祸),比 arrogant 更强调「必然受罚的狂妄」;chart out 由「绘制海图」引申为「详细规划」。最后两个短句是全篇最好的收尾法:But any writer knows to start with a first draft. This is ours. 把整篇社论定义为一份「初稿」——既谦逊(承认未定稿),又自信(大方认领),并且与前文「人类写作不可替代」互为印证。结尾的 ■ 是《经济学人》社论的固定结束符。

03核心词汇精讲

标签说明:四级 基础必备 六级 进阶 雅思 学术高频 高分 写作加分

A. 文字、作者与文风

prose/prəʊz/六级雅思
释义n. 散文;行文(区别于诗歌的书面语言)
派生prosaic adj. 平淡乏味的;prosody n. 韵律学;prosy adj. 啰嗦的
搭配good / lucid / elegant prose 优秀 / 清晰 / 优雅的文笔;swim in dire prose 充斥着拙劣文字
authorship/ˈɔːθəʃɪp/高分
释义n. 作者身份;著述(作品出自谁手)
派生author n. 作者 v. 创作;authorise v. 授权;authorial adj. 作者的
搭配surrender / retain authorship 交出 / 保留作者身份;authorship attribution 作者归属
op-ed/ˌɒp ˈed/雅思高分
释义n. 评论版文章(署名评论,与报社不署名的社论相对)
派生全称 opposite the editorial page(社论版对面的版面)
搭配the op-ed page 评论版;write / publish an op-ed 写 / 发表一篇评论
turn of phrase/ˌtɜːn əv ˈfreɪz/高分
释义n. 措辞,说法(一个人特有的表达方式)
派生turn of mind 思维方式;turn of events 事态变化
搭配a happy / neat turn of phrase 巧妙的措辞;stumble on a turn of phrase 偶然撞见一个说法
clumsiness/ˈklʌmzinəs/六级
释义n. 笨拙;(文笔)生硬
派生clumsy adj. 笨手笨脚的;clumsily adv. 笨拙地
搭配the clumsiness of the prose 文笔生硬;clumsy phrasing 生硬的措辞
workmanship/ˈwɜːkmənʃɪp/六级雅思
释义n. 做工,手艺,工艺
派生workman n. 工匠;craftsmanship n. 精湛技艺
搭配poor / shoddy workmanship 做工粗糙;fine workmanship 精工细作
literate/ˈlɪtərət/四级雅思
释义adj. 识字的;有文化素养的(反义 illiterate 文盲的)
派生literacy n. 读写能力;literature n. 文学;literary adj. 文学的
搭配a literate culture 识字文化;computer-literate 懂电脑的;media literacy 媒介素养
gripe/ɡraɪp/六级
释义n. 抱怨,牢骚 v. 发牢骚
派生griping n. 抱怨 adj. 令人恼火的;griper n. 爱发牢骚的人
搭配the most popular gripe about… 关于……最普遍的抱怨;gripe about sth 抱怨某事
grumble/ˈɡrʌmbl/六级
释义v. 咕哝,低声抱怨 n. 抱怨
派生grumbler n. 爱抱怨的人;grumbling adj. 抱怨的
搭配grumble about / at sth 抱怨某事;grumble that… 咕哝说……
sully/ˈsʌli/高分
释义v. 弄脏;玷污(尤指名声、体验)
派生sullied adj. 被玷污的
搭配sully sb's reputation 败坏名声;sully the experience 破坏体验
fakery/ˈfeɪkəri/高分
释义n. 伪造品,假货(总称,不可数)
派生fake adj. 假的 n. 赝品 v. 伪造;faker n. 造假者
搭配AI-generated fakery AI 生成的假货;outright fakery 彻头彻尾的造假
robo-/ˈrəʊbəʊ/高分
释义comb. (构词成分)机器人的,自动化的
派生robot n. 机器人;robotic adj. 机械的;robotics n. 机器人技术
搭配robo-rhymes 机器韵文(原文自造词);robo-adviser 智能投顾;robo-call 自动语音推销电话

B. 思考、责任与后果

rigorous/ˈrɪɡərəs/六级雅思
释义adj. 严谨的,缜密的;严格的
派生rigour n.(美式 rigor)严谨;rigorously adv. 严格地
搭配rigorous thinking 严谨的思考;rigorous standards 严格的标准;a rigorous analysis 缜密的分析
crystallise/ˈkrɪstəlaɪz/高分
释义v. 使(想法)成形、清晰;(想法)变得明确
派生crystal n. 水晶 adj. 清澈的;crystallisation n. 结晶
搭配crystallise one's thoughts in text 把想法凝练成文字;ideas crystallise into a plan 想法成形为计划(英式拼写 -ise,美式 -ize)
sidestep/ˈsaɪdstep/高分
释义v. 回避,绕过(难题、责任)
派生sidestep n. 侧步,回避 (由拳击侧步引申)
搭配sidestep the question 回避问题;sidestep responsibility 逃避责任;sidestep the issue 绕开议题
outsource/ˈaʊtsɔːs/六级雅思
释义v. 外包;(引申)把(脑力活)交给外部完成
派生outsourcing n. 外包;outsourced adj. 外包的
搭配outsource sth to sb 把某事外包给某人;outsource one's essays to AI 把作文外包给 AI
mindful/ˈmaɪndfl/六级
释义adj. 留心的,意识到的
派生mindfully adv.;mindfulness n. 正念,专注
搭配be mindful of sth 留意某事;mindful of the risks 对风险保持警觉
disservice/ˌdɪsˈsɜːvɪs/六级高分
释义n. 损害,帮倒忙(好心做坏事)
派生serve v. 服务;service n. 服务
搭配do sb a disservice 帮某人倒忙;do the reader a disservice 对读者不负责任
responsibility/rɪˌspɒnsəˈbɪləti/四级
释义n. 责任,职责
派生responsible adj. 负责的;responsibly adv. 负责地
搭配take responsibility for 对……负责;bear / shoulder responsibility 担责;shift responsibility onto 把责任推给
consequence/ˈkɒnsɪkwəns/四级六级
释义n. 后果,结果;重要性
派生consequent adj. 随之而来的;consequently adv. 因此
搭配be mindful of the consequences 留意后果;as a consequence of 由于;face the consequences 承担后果

C. 技术、社会与变迁

Luddite/ˈlʌdaɪt/高分
释义n. 卢德分子;(引申,贬)盲目反对新技术的人 adj. 反技术的
派生neo-Luddite n. 新卢德主义者;Luddism n. 卢德主义
搭配turn Luddite 变成卢德分子;Luddite fallacy 卢德谬误(误以为技术会永久减少就业)
momentous/məˈmentəs/高分
释义adj. 重大的,意义深远的
派生moment n. 时刻;momentously adv.;momentum n. 势头(易混)
搭配a momentous change / decision 重大变革 / 决定;注意勿与 momentary(短暂的)混淆
hubristic/hjuːˈbrɪstɪk/高分
释义adj. 傲慢自大的(因自负而招致灾祸)
派生hubris n. 傲慢;(希腊语 hybris)hubristically adv.
搭配a hubristic assumption 狂妄的假设;hubris invites nemesis 傲慢招致报应
herald/ˈherəld/六级雅思
释义v. 预示……的到来,宣告 n. 先驱;预兆
派生heralded adj. 被宣告的;much-heralded 备受期待的
搭配herald a new era 预示新时代的到来;herald the arrival of 宣告……的到来
grapple with/ˈɡræpl wɪð/高分
释义phr. 努力对付,设法解决(棘手问题)
派生grapple n. 扭打,搏斗
搭配grapple with the implications 努力厘清其影响;grapple with a problem 苦战一个难题
potent/ˈpəʊtnt/六级雅思
释义adj. 强效的,强有力的
派生potency n. 效力,威力;potently adv.
搭配a potent drug 强效药物;a potent symbol 有力的象征;more potent illegal drugs 更烈性的毒品
norm/nɔːm/六级雅思
释义n. 规范,准则;(the norm)常态
派生normal adj. 正常的;normative adj. 规范性的;normalise v. 使常态化
搭配social / cultural norms 社会 / 文化规范;become the norm 成为常态;remake the norms 重塑规范

D. 经济与传播学术语

costly signal/ˈkɒstli ˈsɪɡnl/高分
释义n. 昂贵信号(因成本高、难以伪装,因而可信的信号)
派生signalling n. 信号传递;signal v. 示意,表明
搭配send a costly signal 发出昂贵信号;a costly signal of effort 投入的昂贵信号
externality/ˌekstɜːˈnæləti/高分雅思
释义n. 外部性(一项活动对第三方造成、未计入价格的影响)
派生external adj. 外部的;internalise an externality 使外部性内部化
搭配a negative / positive externality 负 / 正外部性;pollute the information environment 污染信息环境
proliferation/prəˌlɪfəˈreɪʃn/六级雅思
释义n. 激增,扩散(尤指不受欢迎之物的蔓延)
派生proliferate v. 激增;proliferative adj. 增殖的
搭配the proliferation of AI-generated fakery AI 假货的泛滥;nuclear proliferation 核扩散
pollute/pəˈluːt/四级
释义v. 污染;败坏
派生pollution n. 污染;pollutant n. 污染物;polluted adj. 被污染的
搭配pollute the information environment 污染信息环境;pollute the atmosphere 污染大气
creative destruction/kriˌeɪtɪv dɪˈstrʌkʃn/高分
释义n. 创造性破坏(新技术淘汰旧产业同时创造新产业)
派生熊彼特(Joseph Schumpeter)提出的经济学概念
搭配the gale of creative destruction 创造性破坏的飓风;a wave of creative destruction 一波创造性破坏

04长难句拆解(8 句)

1. If The Economist had just two loyalties, they might well be to good prose and creative destruction. Those seldom previously came into conflict—until artificial-intelligence models began writing.
结构:
If The Economist had just two loyalties, ← 虚拟条件从句(与现在事实相反) they might well be [to good prose and creative destruction]. ← 主句:might well be to…,表示「多半会落在……」 [Those] seldom previously came into conflict —[until artificial-intelligence models began writing]. ← until 引导时间状语从句,破折号把它从主句中拎出来强调
难点:① 虚拟语气 had… might well be:用「假如只有两条忠诚」这种明显不现实的假设抬高语气,是社论开篇立威望的常见手法。② Those 是复数指示代词,回指前面两条 loyalty——不要读成单数 it。③ be to sth 表示「归属于……」,此处意为「那两条忠诚会献给……」。④ seldom… until… 是「直到……才(少见地)」的搭配,实际含义是「在此之前几乎从不冲突」,不要把 seldom 与后面的否定语义读反。⑤ 破折号后的 until 从句把抽象的「冲突」钉在一个具体时间点上,全句的张力就来自这个断点。
译文:如果《经济学人》只有两条忠诚,那大概会献给好文笔与创造性破坏。这两者此前极少发生冲突——直到人工智能模型开始写作。
2. It would be churlish to turn Luddite only now, once the wolves of automation have turned up at our own door.
结构:
[It] would be churlish [to turn Luddite only now], ↑ 形式主语 ↑ 真正的主语(不定式) [once the wolves of automation have turned up at our own door]. ← once 引导时间状语从句(此处意为「一旦,既然」)
难点:① It would be + adj. + to do:it 是形式主语,比直接说 To turn Luddite would be churlish 更自然、更少头重脚轻。② would be 是虚拟/委婉语气,暗示「真这么做就太不像话了」,而不是在描述事实。③ churlish 是英式书面词,指「粗鲁无礼、不识好歹」,比 rude 多一层「不合时宜地失礼」的意味。④ turn Luddite:turn + 名词/形容词表示「变成……」,Luddite 在此作形容词用。⑤ once 不是「一次」,而是「一旦/既然」,语气里带自嘲:事情都到眼前了才想起来抵制。⑥ the wolves… at our own door 是狼蹲在自家门口的隐喻,把抽象的行业威胁写成切身危险。
译文:直到自动化的狼群已经蹲在自家门口,才想起来做卢德分子,未免有失风度。
3. What matters in such cases is disclosing the use of AI and making sure that the human being whose name is on the page still takes responsibility.
结构:
[What matters in such cases] ← 主语从句(what 引导,作主语) is [disclosing the use of AI] and [making sure [that the human being [whose name is on the page] still takes responsibility]] ↑ whose 引导定语从句修饰 the human being,插在主谓之间
难点:① What… 引导名词性从句作主语,谓语用单数 is——这类句子读错主谓一致是最常见的失分点。② 表语是两个动名词短语并列,disclosing 与 making sure 结构对称,翻译时要能看出这是「两件事」。③ 真正的难点在 the human being whose name is on the page still takes responsibility:定语从句插在主语与谓语之间,把 takes 和它的主语 human being 隔开了。阅读时先把 whose 从句整体括起来,主干就只剩 the human being takes responsibility。④ take responsibility 与 be responsible 的区别:前者强调主动承担,语气更重。
译文:这类情形里要紧的是:公开说明使用了 AI,并确保署名者仍然承担责任。
4. Writers of all sorts would be doing their readers a disservice if they did not use AI models for research, proofreading and, when appropriate, feedback.
结构:
Writers of all sorts would be doing their readers a disservice ← do sb a disservice 固定搭配:do + 人 + a disservice [if they did not use AI models for research, proofreading and, when appropriate, feedback]. ← if 条件状语从句;did not use 与 would be doing 构成虚拟条件
难点:① do sb a disservice 意为「帮某人的倒忙、损害某人」,宾语是受损的一方;千万不要把 a disservice 当成「服务」来理解。② would be doing… if… did not… 是「与现在事实相反」的虚拟条件句——言下之意是:事实上大多数写作者已经在用了。③ for research, proofreading and feedback 是三个并列名词,中间插入了 when appropriate 作状语;读的时候要跳过插入语,直接看到 use AI models for A, B and C 的主干。④ 这一句的立场与前文相反——它是在为 AI 辩护。作者用它换取「我们不是卢德分子」的信用,后面反对 AI 写作时才更站得住。
译文:倘若各类写作者不用 AI 模型来做资料检索、校对,以及在合适的时候获取反馈,那才是真正对不住读者。
5. Any serious case against AI writing must rest, therefore, on the ways that words are not stockings: how they carry meaning and significance beyond what ends up on the page.
结构:
Any serious case against AI writing must rest, [therefore], on [the ways] ← 主干:case must rest on the ways(论据必须建立在……之上) [that words are not stockings]: ← 同位/定语从句,说明 the ways [how they carry meaning and significance beyond [what ends up on the page]]. ← 冒号后解释 the ways ↑ beyond + 名词性从句
难点:① case against sth 是「反对某事的论据/理由」,不是「案子」;rest on 意为「取决于、建立在……之上」。② therefore 插在 must rest 与 on 之间,是插入语,不影响 must rest on 的完整搭配。③ 冒号后的 how… 与前面 the ways 同位,作用是展开「方式」的具体内容——看见冒号就要意识到后面是解释。④ beyond what ends up on the page:what 从句作介词 beyond 的宾语,字面是「超出最终落在纸面上的东西」,译为「远超纸面所能承载的」。⑤ 这一句是全文的论点句,也是整篇结构的枢纽:前半篇用袜子类比拆掉对手的论点,这里立刻把类比拧成正面论证。
译文:因此,任何严肃的「反对 AI 写作」的论据,都必须建立在文字不是袜子这一点上:它们承载的意义与分量,远超最终落在纸面上的东西。
6. This offers a certain guarantee if not of quality—the world still swims in dire prose—then at least of effort.
结构:
This offers a certain guarantee [if not of quality] ← if not A… —the world still swims in dire prose— ← 破折号插入语:作者主动承认「质量」这一条守不住 [then at least of effort]. ← …then at least B
难点:① 核心结构 if not A, then at least B:先退掉一个较强的说法(质量),再守住一条更弱、也更难反驳的底线(投入)。这是英语里极精炼的让步写法,中文没有对应句式,必须意译。② guarantee 之后两次都省略了重复的 of,完整写法是 if not of quality, then at least of effort。③ 破折号里的插入语是作者主动打自己脸——不等读者质疑「可是劣质文章遍地都是」,先自己承认掉。④ swim in 是「浸泡在……之中」的夸张说法,画面感很强。⑤ dire 意为「糟糕透顶的」,比 bad 强烈得多,常用于 dire consequences / dire straits。
译文:这提供了一种保证:虽说不能保证质量——这个世界依旧泡在糟糕的文字里——但至少保证了投入。
7. Whatever French theorists have to say about the death of the author, readers still want to know that a person wrote the words in front of them.
结构:
[Whatever French theorists have to say about the death of the author], ← whatever 引导的让步状语从句 readers still want to know [that a person wrote the words in front of them]. ← 主句 + that 引导的宾语从句
难点:① Whatever… 引导让步状语从句,意为「无论……说什么」,比 no matter what 更书面,也更适合放在句首。② 从句里 have to say 是「有什么(话)要说」,不是「必须说」——have to 在这里不表义务。③ the death of the author 指罗兰·巴特 1967 年的文论名篇《作者之死》,主张文本意义由读者生产,与作者意图无关。④ 主句的 still 一词完成反驳:理论这样讲,读者的直觉偏偏相反。⑤ 整句的效果是「以一句让步从句把一整个理论流派一句带过,再用读者常识一锤定音」——这是极其高效的辩论句式。
译文:无论法国理论家们如何谈论「作者之死」,读者仍旧想知道:眼前这些字,是一个真人写下的。
8. So it would be hubristic to try to chart out exactly how societies ought in future to approach the use of AI in writing.
结构:
So [it] would be hubristic [to try to chart out exactly ↑ 形式主语 ↑ 真正主语(不定式) [how societies ought in future to approach the use of AI in writing]]. ← how 引导名词性从句,作 chart out 的宾语 注意 in future 插在 ought 与 to 之间
难点:① it 又是形式主语,真正的主语是不定式 to try to chart out…——与第二句 It would be churlish 同一句型,首尾呼应,形成「两边都不做绝对断言」的姿态。② chart out 由「绘制海图」引申为「详细规划、勾勒」,比 plan 更有画面感。③ ought in future to approach:情态动词 ought 与 to 之间插入了 in future,是正式书面语允许的写法;阅读时先把 in future 拿掉,就能看出 ought to approach 的完整结构。④ would be 是虚拟语气,全句是全文的自我设限:不预测未来,只提供当下的一种判断。
译文:因此,想要精确规划出未来社会究竟应当如何对待 AI 写作,未免太过自负。

05写作积累

5.1 六个可直接套用的句型公式

公式 1 让步从句开篇:Whatever + 名词 + have to say about X, 主句仍成立
Whatever French theorists have to say about the death of the author, readers still want to know that a person wrote the words in front of them.
用法:把一整个流派、一种流行观点压进从句,主句用 still 一锤定音。适合议论文首段亮明立场,或反驳段开头。
仿写 → Whatever economists have to say about efficiency, students still want to know that a teacher actually looked at their work.
公式 2 否定一方、肯定另一方:not by A, but by B
Yet liberal societies negotiate these momentous changes not by passively surrendering to technologies, but by adapting to them.
用法:先排除一种做法,再推出自己主张的做法,比单纯说 should 更有辩论感。雅思大作文「讨论双方观点并给出自己立场」时最好用。
仿写 → Universities should respond to AI not by banning it outright, but by teaching students to use it transparently.
公式 3 形式主语 + 虚拟,给结论留余地:It would be + adj. + to do
So it would be hubristic to try to chart out exactly how societies ought in future to approach the use of AI in writing.
用法:把「我」从句子里抽走,用 it 承担判断,语气显得克制、有分寸。适合结尾段作「不妄下断言」的表态。
仿写 → It would be naive to assume that any single policy can solve the problem of online misinformation.
公式 4 极精炼的让步:if not A, then at least B
This offers a certain guarantee if not of quality—the world still swims in dire prose—then at least of effort.
用法:承认「最强的说法」站不住,但守住一条更弱、也更难反驳的底线。属于高阶写法,用在让步段收尾极其有力。
仿写 → The scheme offers a real benefit if not of money, then at least of time.
公式 5 反事实的批评:would be doing sb a disservice if… did not…
Writers of all sorts would be doing their readers a disservice if they did not use AI models for research, proofreading and, when appropriate, feedback.
用法:用虚拟条件句说「不这么做才是错的」,比直接命令(you should…)柔软,也比 You must… 少说教味。
仿写 → Schools would be doing their students a disservice if they did not teach them how to judge the reliability of an online source.
公式 6 论点句 + 冒号展开:X must rest on Y: how…
Any serious case against AI writing must rest, therefore, on the ways that words are not stockings: how they carry meaning and significance beyond what ends up on the page.
用法:先用一句抽象判断立论(must rest on…),再用冒号给出具体展开。适合主体段第一句,读者一眼就知道这一段要论证什么。
仿写 → Any serious argument for remote work must rest on what it changes for the people doing it: how it redistributes time, attention and trust.

5.2 话题语料(可直接搬进作文)

话题一 · AI 与写作 / 学术诚信

outsource one’s essays to AI 把作文外包给 AI · ghostwrite 代笔 · disclose the use of AI 公开说明使用了 AI · academic integrity 学术诚信 · a costly signal of effort 投入的昂贵信号 · lose the habit of reasoning 丧失推理的习惯 · AI-assisted but human-authored 由 AI 辅助、人类创作 · crack down on machine-generated submissions 打击机器生成的投稿

话题二 · 信息环境与媒体素养

the information environment 信息环境 · AI-generated fakery 由 AI 生成的假货 · erode trust in real news 侵蚀对真实新闻的信任 · a negative externality 负外部性 · platform liability 平台责任 · digital watermarking 数字水印 · provenance metadata 来源元数据 · media literacy 媒介素养

话题三 · 技术变迁与社会适应

a technological rupture 技术断裂 · adapt to rather than passively surrender to 适应而非被动屈服 · regulatory catch-up 监管追赶 · the printing press 印刷术 · herald a new era 预示一个新时代 · grapple with the implications 努力厘清其影响 · remake the norms 重塑规范 · a double-edged technology 双刃剑式的技术

5.3 仿写示范 题目:Some people believe schools should ban AI tools. To what extent do you agree?

Schools would be doing their students a disservice if they did not teach them how to use AI responsibly. Whatever critics have to say about cheating, the workplace these students will enter has already made these tools the norm. So the sensible response is not to ban them outright, but to teach students to disclose when and how they were used. It would be naive to pretend that any single policy can settle the matter; yet a school that treats AI as a costly signal of effort, rather than a shortcut, has at least started in the right place.
依次用到公式 5(disservice if…)、公式 1(Whatever…)、公式 2(not to… but to…)、公式 3(It would be naive to…),最后一句再借本文的 costly signal 收尾。注意整段没有一句以 I think 开头——立场是靠句式和事实撑起来的,不是靠表态。

06篇章结构、论证手法与文化冷知识

6.1 论证结构(双重忠诚 → 让步 → 拆解 → 立论 → 降格 → 初稿)

01 开篇 · 双重忠诚(①)
报纸自称只有两条忠诚:好文笔与创造性破坏。这两者此前从不冲突——直到 AI 开始写作。随即用两个分数把文化现象变成政治事实:英国议会十分之一的发言、美国众议院七分之一的发言,可能都由 AI 起草。
↓
02 立场 · 没有中立者(②)
先自证清白:本报为技术进步喝彩两百年,现在才转做卢德分子「未免有失风度」。紧接着一句 And yet. 独立成句,把全文的判断压在这两个词上——作者与读者都应警惕。
↓
03 让步 · 哪些写作可以让渡(③)
客服投诉、技术文档这类「清晰至上」的文本,交给 AI 是好事。但设两个条件:透明披露 + 署名者负责。最后反过来说:不用 AI 做检索、校对、反馈,才真正对不住读者。
↓
04 转折 · 哪些写作不能让(④)
小说、演讲、新闻报道不行。并且作者在这里主动拆掉最容易被攻击的论点:说 AI 文笔笨拙是不成立的,就像当年嫌机器织的袜子粗糙——But then the stockings got better. 先把对手最强的理由收走。
↓
05 换轨 · The medium and the message(小标题)
原刊小标题把讨论从「文字是否笨拙」转向「文字承载什么」。标题本身化用麦克卢汉的名言「媒介即讯息」,暗示下文要谈的是媒介性质,而不是技术优劣。
↓
06 核心论证 · 文字不是袜子(⑤)
全文枢纽。既然袜子会变好,就不能在「文笔」上做文章;真正站得住的论据是:文字承载的意义超出纸面。写作之痛即思考之痛,绕开它 → 思想不严谨。举证由私到公:学生失去推理习惯,政客对后果不再敏感。
↓
07 读者视角 · 昂贵信号与作者承诺(⑥⑦)
视角切到读者。人类文本「写起来比读起来更费时」,这构成一种保证——不是质量的保证,而是投入的保证,经济学称作 costly signal。再借诗歌盲测与「作者之死」说明:读者要的不是文本本身,而是「有个真人写下它」这个承诺。
↓
08 次生伤害 · 外部性(⑧)
就算你自己不用 AI,也会被别人的 AI 文本波及:读到可疑的机器腔就弃读,看到假货泛滥就不再信任真实新闻。这层的受害者恰恰是「不用 AI 的读者」,所以作者用 externality(外部性)来命名——成本由没参与的人承担。
↓
09 收束 · 历史降温 + 自我定位(⑨⑩)
承认技术会重塑规范,也承认 AI 带来机遇。立场句用 not by… but by… 落定:自由社会靠适应而非投降。印刷术→宗教战争、1930 年代电台→法西斯,两次历史类比之后立刻用 Few associate either with such high drama today 降温。最后把全文定义为一份初稿:This is ours.

6.2 七种论证手法(可迁移到自己的议论文)

手法原文例子效果
对举(antithesis)If The Economist had just two loyalties, they might well be to good prose and creative destruction.开篇把两条互相拉扯的价值并置,再用一句破折号让它们相撞。全文的张力在第一句就建立起来了,这是社论最经典的起手式。
让步转折(concession)Admittedly, in some sorts of writing an AI takeover would be welcome… / It would be churlish to turn Luddite only now… / And yet.先把对方最合理的部分接过来,再转折。让步越彻底,后面那句只有两个词的 And yet. 越有力量——这正是全文最重要的一次翻转。
先破后立(pre-emptive rebuttal)The Luddites grumbled, once, about the poor workmanship of machine-made stockings. But then the stockings got better. Doubtless, AI sentences will, too.不等读者开口,作者先把对手最强的一条理由(AI 文笔会变好)承认掉,再说明它推不出「因此无所谓」的结论。社论中最难学、也最有效的一招。
隐喻(metaphor)the wolves of automation have turned up at our own door / words are not stockings / start with a first draft三个隐喻各担一职:「狼在自家门口」写切身威胁,「文字不是袜子」承担全文论点,「初稿」完成自我定位。抽象论证因此有了画面和抓手。
跨学科借词(borrowed terminology)Economists call that a costly signal. / what economists call an externality两次借经济学概念给文学议题加硬度,把讨论从「感受」推向「机制」。同一篇文章里用了两次,是刻意的复现——读者会记住这两个词。
历史类比(historical analogy)The printing press set off religious wars in early-modern Europe. In the 1930s the radio heralded fascism. Few associate either with such high drama today.用两次「技术—社会震荡」说明恐慌是常态、社会终会消化;紧跟一句降温,避免读者把 AI 直接等同于灾难。举史是为了平息,不是为了恐吓。
短句收束(short-sentence close)And yet. / Someone, somewhere, thought something was worth saying. / This is ours.全文三处极短句,分别用于转折、定义与结尾。力度靠节奏切换而不是形容词堆砌——长句铺完立刻收短,读者会被这个落差按住。

6.3 文化冷知识

卢德分子(Luddites)与机织袜子

1811—1816 年英国纺织工人以传说中的「内德·卢德」为名捣毁织机,反对机器抢走生计。「卢德谬误」(Luddite fallacy)由此而来:误以为技术会永久性地减少就业。原文特意用「机织袜子后来变好了」点破这一层,所以 turn Luddite 今天的意思不是「反对技术」,而是「不识时务地反对技术」——贬义落在「不合时宜」上,而非「反对」上。

创造性破坏(creative destruction)

经济学家熊彼特(Joseph Schumpeter)在《资本主义、社会主义与民主》中提出的概念:新技术、新模式在淘汰旧产业的同时创造新产业,是市场经济自我更新的常态。原文把它列为报纸的第二条忠诚,是精心的选择——这个概念本身就带着「进步必有代价」的含义,与后文「AI 让一部分人受损」的论述同构。

「作者之死」(the death of the author)

法国文论家罗兰·巴特 1967 年的同名文章,主张文本一旦写成,意义就由读者生产,追问「作者想说什么」是徒劳的。原文用一句让步从句把它带过,随后说读者的直觉并不买账。值得玩味的是:《经济学人》长年不署个人名字,社论以 This newspaper 自称——一篇讨论「作者身份」的文章,恰恰出自一份刻意隐去作者身份的刊物。结尾 This is ours. 正是在拿这一点自嘲。

昂贵信号(costly signal)

演化生物学与经济学中的概念:只有当信号的成本足够高、别人难以伪装时,它才可信。扎哈维(Amotz Zahavi)的「累赘原理」与斯宾塞(Michael Spence)的劳动力市场信号理论都属此列(后者获 2001 年诺贝尔经济学奖)。原文用它解释「为什么人类写作耗时长反而是优点」——耗时长本身就是「有人认真对待过」的证明。

外部性(externality)

经济学概念:一项活动对第三方造成、却未计入交易价格的影响,如工厂排污。负外部性的典型特征是「成本由没参与的人承担」。原文用它形容 AI 泛滥:不用 AI 的读者,也得为别人制造的假消息付出信任成本。注意作者在第八段把受害者明确限定为「不用 AI 的人」,正是为了贴合这个定义。

印刷术、电台与前车之鉴

15—17 世纪的印刷术让宗教文本得以大量传播,也让欧洲陷入长达一个多世纪的宗教战争;1930 年代广播成为大众媒介,与法西斯主义的动员能力高度相关。原文举这两个例子不是为了预言灾难,而是为了说明「媒介技术刚出现时引发的震荡,事后往往被看成过眼云烟」。Few associate either with such high drama today 是全文态度最冷静的一句,也是最容易被读者忽略的一句。
— 全文完 · 共 10 段 · 建议精读时长 55 分钟 —