The Safest Job from AI may be Writing

字幕 · 点击任意句跳转

• 00:00Imagine a bowl of fruit sitting on a table.想象一下，桌上摆着一碗水果。
• 00:03You see a bright shiny red apple.你看到一个又红又亮的苹果。
• 00:05It looks absolutely perfect.它看起来完美无瑕。
• 00:06Right, not a single bad spot on it.没错，连一点瑕疵都没有。
• 00:08Exactly.确实如此，近乎完美。
• 00:08Practically close.确实如此，近乎完美。
• 00:10So you pick it up.于是你把它拿起来，张大嘴巴狠狠咬了一口。
• 00:10You open your mouth and you take a massive bite.于是你把它拿起来，张大嘴巴狠狠咬了一口。
• 00:13And instead of sweet juice, your teeth hit hard solid plastic.但牙齿碰到的不是甘甜的汁水，而是坚硬冰冷的塑料。
• 00:17Yes.是的。
• 00:18It actually hurts your jaw.甚至震得下巴生疼。
• 00:20You look closely at it.你仔细端详它。
• 00:21It looks like an apple.它看起来像个苹果。
• 00:23It's shaped like an apple.形状也像个苹果。
• 00:25But well, it has no taste.可是，它毫无味道。
• 00:27There's no juice, no flavor.没有汁水，没有风味。
• 00:28Yeah, no life inside it.对，里面毫无生机。
• 00:29It's completely fake.它完全是个假货。
• 00:30It's just an empty copy.不过是一个空洞的仿制品。
• 00:32The outside looks right, but the inside feels entirely wrong.外表看似逼真，内在却完全不是那么回事。
• 00:36It just leaves you feeling disappointed.只留给你满心的失望。
• 00:38And right now, that is exactly what's happening to the words we read on the internet.而眼下，我们在互联网上阅读的文字正是这种情况。
• 00:42We're just like biting into plastic apples all day long.我们整天就像在不停地咬塑料苹果。
• 00:46So today, we are asking a very big question.所以今天，我们要探讨一个非常重大的问题。
• 00:49Is writing the safest job from artificial intelligence?在人工智能面前，写作是最不容易被取代的工作吗？
• 00:52That is the big question.这确实是个大问题。
• 00:54Is the human writer safe from AI?人类写作者在AI面前真的安全吗？
• 00:56Right.对。
• 00:57It's a question that makes a lot of people from famous authors to everyday office workers Extremely nervous about their future.这个问题让许多人——从知名作家到普通上班族——都对自己的未来感到极度焦虑。
• 01:04Oh, absolutely.哦，绝对是这样。
• 01:04So we're pulling from two great sources for today's deep dive.因此，我们今天的深度剖析参考了两个绝佳的信息来源。
• 01:08First, we have a blog post from the year 2026 written by a tech writer named Murat Buffalo.首先，是一位名叫穆拉特·布法罗的技术作家在2026年写的一篇博文。
• 01:13He is a very strong theory about why human writing is safe.关于为什么人类写作是安全的，他提出了一个非常强有力的理论。
• 01:16But our second source is a heated discussion thread from a website called Hacker News.但我们的第二个来源，是来自Hacker News网站上一场异常激烈的讨论帖。
• 01:21Yeah, that thread was intense.是的，那个帖子的讨论非常激烈。
• 01:22Very intense.确实非常激烈。
• 01:23In that thread, hundreds of internet commenters argued that his beautiful theory is completely wrong.帖子里成百上千的网友争论道，他那美好的理论大错特错。
• 01:29So our mission today is to understand both sides of this argument.所以我们今天的任务，就是全面了解这场争论的正反两方。
• 01:33We will explore why AI still struggles to write like a real human.我们将探讨为什么AI依然很难写出像真正人类那样的文字。
• 01:37We also need to understand why human writers are losing their jobs right now today.但我们同时也需要理解，为什么人类写作者如今正在失去工作。
• 01:42Yeah, it's a complicated reality.是的，现实情况相当复杂。
• 01:43Really is.确实如此。
• 01:44Whether you write fantasy books or you just write simple emails for your office job, this deep dive will show you what the future holds for your words.无论你是创作奇幻小说，还是仅仅在办公室写写日常邮件，这次深度剖析都将向你展示，你笔下的文字将迎来怎样的未来。
• 01:51Okay, let's get into it.好，那我们就切入正题吧。
• 01:52Let's do it.开始吧。
• 01:53Murat Buffalo starts his blog post with a major claim.穆拉特·布法罗在博文开篇就提出了一个重大主张。
• 01:57He says AI is changing some industries incredibly fast, like computer programming.他说AI在某些行业带来的改变快得惊人，比如计算机编程。
• 02:01But writing text, his writing is stuck.但是在文本写作方面，AI的发展却陷入了停滞。
• 02:05Yeah, and he focuses a lot on tech workers to make his point.对，为了证明他的观点，他着重关注了科技工作者。
• 02:09Programmers are very stressed out right now.程序员现在压力非常大。
• 02:11Very stressed.压力极大。
• 02:11They are dealing with a huge new problem.他们正面临着一个巨大的新难题。
• 02:14He calls it cognitive debt.他将其称为“认知负债”。
• 02:17Let's break down that term for a second.我们稍稍拆解一下这个术语。
• 02:18What is that actually doing to a programmer's brain?这究竟会对程序员的大脑造成什么影响？
• 02:23Well, think of cognitive debt as extreme mental exhaustion.嗯，可以把认知负债理解为极度的脑力消耗。
• 02:28When a human writes computer code, they follow a logical path.人类写代码时，会遵循逻辑脉络。
• 02:32You know, a human thinks in steps.你知道的，人是分步骤思考的。
• 02:34Right, step one, step two.没错，第一步，第二步。
• 02:35Exactly.正是如此。
• 02:36But when an AI writes computer code, it just forces a solution that mathematically works.但AI写代码时，只是强行凑出一个在数学上可行的解法。
• 02:41It creates a tangled messy web of instructions.它生成的是一堆混乱纠缠的指令网络。
• 02:45So the AI writes the code in two seconds.所以AI两秒钟就能把代码写完。
• 02:47But the human programmer has to read it.但人类程序员却不得不去读它。
• 02:49Great.没错。
• 02:49They have to trace this bizarre alien logic to make sure it won't crash the company's website.他们必须梳理出这种诡异离奇的逻辑，以确保它不会让公司网站崩溃。
• 02:54That is exactly the problem.问题恰恰就在这里。
• 02:55Reading and fixing bad machine code takes a massive amount of brain power.阅读和修复糟糕的机器代码需要耗费海量的脑力。
• 02:59The human brain gets exhausted trying to untangle a machine's weird logic.人类大脑在试图理清机器那古怪逻辑的过程中变得精疲力竭。
• 03:03That tiredness that is the cognitive debt.那种疲惫感，就是所谓的认知负债。
• 03:06And while AI is causing headaches in the coding world, the author says it's doing something even worse to regular writing.作者指出，AI在编程领域引发头痛的同时，它对常规写作造成的负面影响甚至更为恶劣。
• 03:12He uses a great phrase.他用了一个很绝的词。
• 03:14He calls AI writing, plastic solace word Slop.他把AI写作称为“塑料般毫无灵魂的文字垃圾”。
• 03:18Which brings us right back to your plastic apple?这不正好又绕回了你说的塑料苹果吗？
• 03:21The AI text looks like a perfect paragraph on the page.AI生成的文字在页面上看起来像一段完美的段落。
• 03:24The grammar is 100% correct.语法百分之百正确。
• 03:26The spelling is perfect.拼写也无可挑剔。
• 03:27But when you read it, it feels completely empty.但当你读起来时，却感觉无比空洞。
• 03:30It really does.真的就是这样。
• 03:31It uses the same boring words over and over again.它反反复复用着那些乏味的词汇。
• 03:34It lacks energy.毫无生机与活力。
• 03:35Exactly.一点没错。
• 03:35I see this all the time online.我在网上随处可见这种内容。
• 03:37But here was what I don't understand.但我有一点想不通。
• 03:39AI can make amazing videos now.AI现在能制作出令人惊叹的视频。
• 03:41It can generate beautiful complex pictures.它能生成精美复杂的图像。
• 03:44Why is text so bad?为什么写出的文字却这么差劲？
• 03:45Why is writing just a pile of Slop?为什么写出来的内容就是一堆垃圾？
• 03:47Well, the author explains this by talking about a plateau.作者用“平台期”的概念解释了这一点。
• 03:50A plateau.平台期。
• 03:51Like a flat hill.就像一座平顶山。
• 03:52Right.对。
• 03:52A plateau is a flat area at the top of the hill.平台期就是山顶上的一块平地。
• 03:55You climb and climb, but then you reach the plateau.你一路往上爬，可一旦到达平台期，
• 03:58And you stop going up.你就不再继续上升了。
• 03:59You just stay flat.就只是停留在同一水平线上。
• 04:00AI development.AI的发展，
• 04:01Hit a plateau with writing.在写作方面遇到了平台期。
• 04:03Wait.等等。
• 04:03Are you saying the technology stopped getting better?你是说技术已经停止进步了吗？
• 04:05In a way.某种程度上是这样。
• 04:06Yeah.是的。
• 04:06Think about AI video and images.想想AI视频和图像。
• 04:09When companies wanted better AI pictures, they just added more computer power.当各家公司想要更好的AI图像时，他们只需增加更多算力。
• 04:13They bought more servers.他们购买更多服务器，
• 04:14They used more energy.消耗更多能源。
• 04:15And the pictures became more realistic.图像就变得更加逼真了。
• 04:17Just pushing harder.全靠大力出奇迹。
• 04:18Exactly.没错。
• 04:19The solution was just adding more force.解决方法就是不断加大力度。
• 04:22But I'm guessing the brute force doesn't work for a novel or a poem.但我猜，这种暴力破解法对写小说或诗歌并不奏效。
• 04:25It doesn't work at all.完全不起作用。
• 04:26You can't just add a thousand more computers to get deep rich human expression.你不可能靠多加一千台电脑，就得到深刻而丰富的人类情感表达。
• 04:30That is a completely different kind of problem.这是完全不同性质的问题。
• 04:32That makes me wonder why text is so much harder than writing computer code or solving a math problem.这让我很好奇，为什么文字创作会比写代码或解数学题难得多。
• 04:38The author introduces a concept to explain this right.作者提出了一个概念来解释这一点。
• 04:42He calls it a wicked problem.他称之为“棘手问题”。
• 04:44That's a very important term in science and design.这是科学和设计领域一个非常重要的术语。
• 04:48A wicked problem isn't evil.“棘手问题”并不是说它是邪恶的。
• 04:49It's a problem that has no clear rules.而是指它没有明确的规则。
• 04:51Okay.明白了。
• 04:52No rules.没有规则。
• 04:52All right.好的。
• 04:53It has no clear finishing line.它没有明确的终点线。
• 04:55And most importantly, there's no single right answer.最重要的是，它没有唯一的正确答案。
• 04:57Let's contrast that with something simple like a business budget.我们把它和简单的商业预算对比一下。
• 05:00Yeah.好。
• 05:01Calculating a budget is not a wicked problem.计算预算就不是一个棘手问题。
• 05:03Definitely not.绝对不是。
• 05:04If a company spends ten dollars and makes fifteen dollars, the profit is five dollars.如果一家公司支出10美元，收入15美元，那么利润就是5美元。
• 05:10There's one correct answer.它只有一个正确答案。
• 05:12The rules are absolute.规则是绝对的。
• 05:14And computer coding is very similar to that budget.写代码和算预算非常相似。
• 05:17Coding has strict rules.编程有严格的规则。
• 05:19A computer can run a quick test to see if the code actually works.计算机可以运行一个快速测试，看看代码是否真的能跑通。
• 05:22If the code builds the website properly, it's correct.如果代码能正常搭建起网站，它就是正确的。
• 05:25If the website crashes, it's wrong.如果网站崩溃了，它就是错误的。
• 05:28It's black and white.非黑即白。
• 05:29A machine understands black and white very well.机器非常擅长理解非黑即白的事物。
• 05:31But writing a book has no strict test.但写一本书却没有严格的检验标准。
• 05:33There's no machine in the world that can read a love poem.世界上没有任何一台机器能读完一首情诗后，
• 05:37And print out a score saying, well, yes, this poem is exactly 98% correct.打印出一个分数说：嗯，没错，这首诗的准确率正好是98%。
• 05:41No.确实没有。
• 05:41Not at all.完全没有。
• 05:42Writing is completely open.写作是完全开放的。
• 05:43And because it's open, a writer has to rely on something called theory of mind.正因为它是开放的，作家必须依赖一种叫做“心智理论”的东西。
• 05:47Theory of mind.心智理论。
• 05:48Let's explain theory of mind that sounds a bit like psychology.解释一下心智理论吧，这听起来有点像心理学。
• 05:51It is psychology.它就是心理学概念。
• 05:53Theory of mind is the human ability to guess what another person is thinking and feeling.心智理论指的是人类推测他人想法和感受的能力。
• 05:57Oh, I see.哦，我懂了。
• 05:58When a good human writer sits down to write a story, they're constantly thinking about the reader's brain.当一个优秀的人类作家坐下来写故事时，他们会时刻揣摩读者的心智。
• 06:03They ask themselves, will this sentence make the reader feel sad?他们会问自己：这句话会让读者感到悲伤吗？
• 06:08Will this joke make them laugh?这个笑话会逗笑他们吗？
• 06:10Will they be confused here?读到这里他们会不会感到困惑？
• 06:12So the writer is playing a complex game with the reader's emotions.所以作家是在与读者的情绪进行一场复杂的心理博弈。
• 06:15They're trying to guide the reader through a specific experience.他们试图引导读者经历一种特定的体验。
• 06:18Exactly that.正是如此。
• 06:19But an LLM cannot do this.但大语言模型做不到这一点。
• 06:22Just a pause for a second.稍停一下。
• 06:23An LLM is a large language model.LLM就是大语言模型（Large Language Model）。
• 06:25It's the core technology behind all these AI chat bots we use today.它是我们今天使用的所有AI聊天机器人背后的核心技术。
• 06:29At its heart, it is just a giant math program that looks at a word and guesses the most likely next word.其核心本质不过是一个庞大的数学程序，它看着一个词，然后预测下一个最可能出现的词。
• 06:35That is the key.关键就在这里。
• 06:36The LLM has no physical body.大语言模型没有肉身实体。
• 06:38It has no real world experience.它没有真实世界的经验。
• 06:40It has never felt cold rain on its face.它从未感受过冰冷的雨水打在脸上的滋味。
• 06:41It has never felt hungry or scared or heartbroken.它从未体会过饥饿、恐惧或心碎。
• 06:44Wow, that's true.哇，确实如此。
• 06:45Because it has no feelings.因为没有情感，
• 06:46It cannot guess what a human reader will feel.它就无法推测人类读者会有何感受。
• 06:49It just puts words next to other words because it's math formulas as they usually go together.它只是把词语按数学公式拼接在一起，因为这些词通常经常搭配出现。
• 06:54I needed challenge this a bit though.不过我想对这个观点提出一点质疑。
• 06:56We constantly hear on the news that AI is passing in credible tests.我们经常在新闻上听到AI通过了各种不可思议的考试。
• 07:00It passes the bar exam for lawyers.它通过了律师资格考试，
• 07:02It passes medical exams for doctors.还通过了医师执业考试。
• 07:04If an AI is smart enough to pass a complex medical test, I would assume it can easily learn the rules of writing a good entertaining book.如果AI聪明到能通过复杂的医学考试，我会以为它能轻松掌握写出一本生动有趣的书的规则。
• 07:12Passing a test is just following strict rules.通过考试只是遵循严格的规则而已。
• 07:15A machine is amazing at memorizing facts.机器在背诵记忆事实方面极其出色。
• 07:17It can absolutely write a sentence with perfect punctuation and grammar.它完全能写出标点和语法毫无差错的句子。
• 07:21But writing a correct sentence is not the same thing as writing a good sentence.但写出一个正确的句子，和写出一个好的句子，根本不是一码事。
• 07:26So where's the line between correct and good?那么“正确”和“优秀”之间的界限究竟在哪里？
• 07:28The line is a concept called resonance.界限在于一个叫做“共鸣”的概念。
• 07:31Resonance.共鸣。
• 07:31Resonance means making a feeling happen inside another human's mind.共鸣意味着在另一个人心中唤起某种情绪。
• 07:35When you strike a bell, it vibrates and makes a sound.当你敲响一口钟，它会振动并发出声音。
• 07:39A good sentence does that to a reader.一个好句子对读者起到的也是这种效果。
• 07:41It vibrates in their mind.它在读者的心智中激起回响。
• 07:42I love that image.我很喜欢这个比喻。
• 07:43Yeah, makes you feel something deeply because it connects to your own real life.是的，它能让你深深触动，因为它与你真实的现实生活产生了连接。
• 07:47An AI cannot create resonance because it doesn't have a life to share.AI无法产生共鸣，因为它没有可以分享的生活体验。
• 07:50It only has math.它拥有的只有数学。
• 07:51Okay, let's look at what this actually means for people who write.好，那我们来看看这对写作的人究竟意味着什么。
• 07:54If AI writing is just generic plastic slopp and it can't create resonance, what happens to real human writing, Murat Buffalo argues that human writing is going to become much more valuable.如果AI写作只是千篇一律的劣质流水线产物，无法引发共鸣，那真正的人类写作会迎来怎样的命运？穆拉特·巴法洛认为，人类的写作将会变得更有价值。
• 08:06He does.确实如此。
• 08:07And he builds this argument using two ideas from economics.他借助了经济学中的两个概念来构建这个论点。
• 08:12The first idea is comparative advantage.第一个概念是比较优势。
• 08:15Comparative advantage basically means you should focus your time on the one thing you do best.比较优势基本上是指，你应该把时间集中在自己最擅长的那件事上。
• 08:19Comparative others.相对于其他人而言。
• 08:21Right.没错。
• 08:21If a machine can type basic boring words a million times faster than you, you shouldn't try to race the machine.如果机器能以比你快一百万倍的速度敲出枯燥乏味的基础文字，你就不应该试图去和机器赛跑。
• 08:28You'll lose every time.你每一次都会输。
• 08:30Every single time.每一次无一例外。
• 08:31Instead, you should focus entirely on what the machine cannot do.相反，你应该把精力完全集中在机器做不到的事情上。
• 08:34You focus your time on creativity, deep emotion, and human connection.把时间专注于创造力、深刻的情感以及人与人之间的连接。
• 08:39And a secondary idea is something called a costly signal.第二个概念叫做“高代价信号”。
• 08:42A costly signal is a sign of high quality that's very hard to fake.高代价信号是一种极难伪造的高质量标志。
• 08:47It is a signal that takes a huge amount of time energy or money to produce.它需要耗费巨大的时间、精力或金钱才能发出。
• 08:51I really love the analogy he uses for this.我很喜欢他对此所用的比喻。
• 08:53He talks about the peacock.他提到了孔雀。
• 08:54You know, the male bird with a massive bright colourful tail feathers?就是那种雄性孔雀，有着巨大、鲜艳、色彩斑斓的尾羽？
• 08:58Yes, it's a perfect example from biology.是的，这是生物学中一个绝佳的例子。
• 09:00That giant tail is a costly signal.那条巨大的尾巴就是一个高代价信号。
• 09:03It's extremely heavy.它极其沉重。
• 09:04It makes it very hard for the bird to fly away from a predator.它让孔雀在躲避捕食者时极难飞走。
• 09:07Takes a massive amount of food and energy to grow those feathers.长出那些羽毛需要消耗大量的食物和能量。
• 09:10That's right.一点也没错。
• 09:10So when a female bird sees a male with a huge perfect tail, she knows for a fact that he is strong.所以当雌鸟看到拥有巨大完美尾巴的雄鸟时，她可以确信他很强壮。
• 09:17He survived even with that heavy burden.即便带着如此沉重的负担，他也活了下来。
• 09:19The tail proves his quality.那尾巴证明了他的优良品质。
• 09:23Taking that back to our world in a future filled with fast, cheap, instant AI text.回到我们的现实世界：在未来充斥着快速、廉价、即时的AI文本时，
• 09:27Taking the time to write with a real human voice is the new peacock tail.花时间用真正的人类声音写作，就成了新的孔雀尾巴。
• 09:32That makes total sense.这完全说得通。
• 09:33If I receive a long, thoughtful email from a co-worker, and I can tell they wrote every single word themselves, it's a costly signal.如果我收到同事发来的一封冗长、深思熟虑的电子邮件，而且能看出每个字都是他们亲笔所写，这就是一个高代价信号。
• 09:41It proves they actually took time out of their busy date to think about my problem.这证明他们真的从忙碌的一天中抽出时间，认真思考了我的问题。
• 09:45It says, I am a real person, and I care enough to put my human energy into this.它表明：我是一个真实的人，我足够重视这件事，愿意为此倾注人类的精力。
• 09:50As AI makes all normal texts cheap and boring, human authenticity becomes the real premium product.随着AI让所有普通文本变得廉价而乏味，真实的人性表达反而成了真正的高端产品。
• 09:56Being real becomes the thing people actually want to buy and read.真诚与真实成了人们真正愿意掏钱购买和阅读的东西。
• 10:00The costly signal becomes highly valuable because it's rare.因为稀缺，高代价信号变得极其宝贵。
• 10:03That sounds like a wonderful optimistic theory.这听起来是个非常乐观美好的理论。
• 10:05It makes you feel safe.它让人感到安心。
• 10:07Human writers win in the end because humans crave real connection.人类作家最终会赢，因为人类渴望真实的连接。
• 10:10But then we turn to our second source.但随后我们转向第二个消息来源。
• 10:12The hacker news discussion thread.Hacker News上的讨论帖。
• 10:13Yeah, the reaction there was completely different.是的，那里的反应完全截然不同。
• 10:16The internet commenters completely disagreed with this optimistic theory.网上的评论者完全不认同这种乐观的理论。
• 10:19They were actually quite angry about it.他们甚至对此感到相当愤怒。
• 10:21And the reasoning was simple.原因其实很简单。
• 10:23The real world is messy.现实世界是残酷杂乱的。
• 10:25The commenters pointed out that human writers aren't safe in the future.评论者指出，人类作家在未来根本就不安全。
• 10:28They are losing their jobs right now today.就在此时此刻，他们正在失去工作。
• 10:31It's a harsh reality.这是个残酷的现实。
• 10:33Translators, technical writers, and marketing copywriters are being actively replaced by AI systems as we speak.翻译、技术文档写作者和营销文案人员，就在我们说话的当下，正被AI系统实际取代。
• 10:40I have to push back on the blog post theory here.我必须对博客文章中的理论提出反驳。
• 10:42If we all just agree that AI writing is plastic slop without any resonance, why on earth are companies using it to replace human workers?如果我们都承认AI写作只是没有任何共鸣的劣质流水线产物，那企业究竟为什么要用它来取代人类员工？
• 10:51Why buy a fake apple?为什么要买假苹果？
• 10:53Companies do it because for many businesses, the quality of the writing isn't load-bearing.企业这么做，是因为对许多业务来说，文本的质量并不是“承重墙”。
• 10:58Load-bearing.承重。
• 10:58Like a load-bearing wall in a house.就像房子里的承重墙一样。
• 11:01That's a great way to picture it.这个形象比喻非常好。
• 11:02A load-bearing wall holds up the heavy roof.承重墙支撑着沉重的屋顶。
• 11:04If that wall is weak or poorly built, the entire house collapses.如果那堵墙薄弱或偷工减料，整座房子就会坍塌。
• 11:08It is critical.它至关重要。
• 11:09But some walls in a house are just there to divide a big room into two smaller rooms.但房子里有些墙只是用来把一个大房间隔成两个小房间的。
• 11:14If those divider walls are cheap or thin, it doesn't really matter.如果这些隔断墙造价低廉或很薄，其实并无大碍。
• 11:17The house is still safe.房子依然是安全的。
• 11:18So you're saying that for some companies, the text doesn't have to be perfect to do its job.所以你的意思是，对某些公司来说，文本不需要完美就能发挥作用。
• 11:23It just has to exist.它只需要存在即可。
• 11:26Sometimes text just needs to fill an empty space on a webpage.有时候，文字只需要填补网页上的空白位置就行。
• 11:29Think about the soft music they play in an elevator.想想电梯里播放的那种舒缓音乐。
• 11:32Oh, elevator music.哦，电梯音乐（背景背景乐）。
• 11:33Right.对，没人听电梯音乐是为了体会深刻的情感。
• 11:33No one listens to elevator music to feel deep emotion.对，没人听电梯音乐是为了体会深刻的情感。
• 11:36It's not a beautiful symphony.它不是美妙的交响乐。
• 11:38It's just there.它只是放在那儿，
• 11:39So the elevator is an awkwardly quiet.免得电梯里尴尬地安静。
• 11:41A lot of modern business writing is just elevator music.现代商业写作中有很大一部分充其量就是电梯音乐。
• 11:44The hacker news commenters shared some wild examples of this happening in the real world.Hacker News上的评论者分享了现实世界中发生的一些离谱例子。
• 11:49One person talked about book publishers.有人谈到了图书出版商。
• 11:51They were using cheap, bad AI to translate popular Spanish fantasy books into English.他们用廉价拙劣的AI把热门的西班牙语奇幻小说翻译成英语。
• 11:57And the translation mistakes were incredibly obvious to anyone reading them.任何读者都能一眼看出那些极其明显的翻译错误。
• 12:00They really were.确实如此。
• 12:01The commenter noted that the AI translated the English word fall.评论者指出，AI在翻译英文单词“fall”时出了错。
• 12:05Now, in English, fall has two totally different meanings.在英语中，“fall”有两个完全不同的意思。
• 12:09It can mean a person tripping and falling down to the ground.它可以表示一个人绊倒、跌落在地上。
• 12:12Or it can mean the season before winter.也可以表示冬天之前的秋季。
• 12:14The AI couldn't understand the context.AI根本无法理解语境。
• 12:17It translated the beautiful autumn season as a person falling down.它把美丽的秋天翻译成了一个人摔倒在地。
• 12:21That is a catastrophic mistake for a fantasy novel.对于一部奇幻小说来说，这是灾难性的错误。
• 12:23It ruins the atmosphere completely.它彻底破坏了氛围感。
• 12:25But here's the crazy part.但最不可思议的地方在这里。
• 12:27The readers complained online, but they still bought the books.读者虽然在网上抱怨，但他们依然买了这些书。
• 12:30They bought them anyway because they love the famous characters in the famous brand.他们还是买了，因为他们喜爱知名IP里的著名角色。
• 12:33The actual quality of the translated words wasn't load bearing for the publishers wallet.翻译文字的实际质量对出版商的钱包来说并非“承重墙”。
• 12:40The publishers still made money.出版商依然赚到了钱。
• 12:42This reveals a very hard truth for anyone who wants to write for a living.这向所有想以写作为生的人揭示了一个非常残酷的现实。
• 12:46Before you can become a famous author with a beautiful expensive peacock tail, you have to pay your bills.在你成为拥有华美昂贵孔雀尾巴的知名作家之前，你得先交得起账单。
• 12:52You really do.确实如此。
• 12:53You pay your rent with basic starter writing jobs.你得靠基础的入门级写作工作来交房租。
• 12:57You write instruction manuals.你写产品说明书。
• 12:58You translate basic articles.你翻译基础文章。
• 13:00You write simple marketing emails.你写简单的营销邮件。
• 13:02Those are the exact divider wall jobs that AI is taking away.而这些正是AI正在夺走的“隔断墙”式工作。
• 13:07If you can't get a basic starter job to pay your rent, how do you survive long enough to practice and become a master writer?如果你连一份能付得起房租的基础入门工作都找不到，你又怎么能撑到练就一身本领成为写作大师的那一天？
• 13:13That's the real fear in the writing community.这才是写作圈真正的恐惧所在。
• 13:15The top of the profession might be safe, but the bottom of the ladder is disappearing completely.这个行业的金字塔尖或许是安全的，但阶梯的最底层正在彻底消失。
• 13:19So if AI is taking over the basic jobs, but completely failing at the high level creative art, how are successful writers actually using AI in the year 2026?那么，如果AI正在接管基础工作，却在高层次的创意艺术上彻底受挫，成功的作家在2026年究竟是如何使用AI的？
• 13:29Are they just ignoring it?他们是直接忽视它吗？
• 13:30Not at all.完全不是。
• 13:31The commenters on Hacker News had a lot of practical insight on this.Hacker News上的评论者对此有很多务实的见解。
• 13:35Professional writers do use AI, but they don't use it as an autopilot.专业作家确实会使用AI，但他们不会把它当成自动驾驶仪。
• 13:39They use it strictly as a copilot.他们严格将其用作副驾驶。
• 13:41Okay.明白了。
• 13:41An autopilot flies the airplane completely by itself.自动驾驶仪是完全独立操纵飞机的。
• 13:45The pilot can go to sleep.飞行员可以去睡觉。
• 13:47But a copilot just assists the human pilot who is still flying the plane.但副驾驶只是协助仍在亲自驾驶飞机的人类飞行员。
• 13:52That's the difference.这就是区别。
• 13:53Real writers do not let the AI write the final words on the page.真正的作家绝不会让AI在稿纸上写下最终定稿的文字。
• 13:57In fact, one commenter shared a strict rule they use.事实上，一位评论者分享了一条他们奉行的严格准则。
• 14:00They said that as soon as the AI suggests a specific word, that word becomes poisoned.他们说，一旦AI建议了某个特定的词，那个词就“有毒”了。
• 14:06Coisend.有毒。
• 14:06That is a fascinating way to describe it.这个说法真有意思。
• 14:09So if the writer asks the AI for ideas and the AI suggests the word beautiful, the writer will absolutely refuse to use it.所以如果作家向AI寻求灵感，而AI建议了“美丽”这个词，作家就坚决不会再用它。
• 14:16They'll say, okay, now I can't use beautiful.他们会说：好吧，现在我不能用“美丽”了。
• 14:17I have to dig deeper and find a much better, more interesting word.我必须更深入挖掘，找到一个更好、更有趣的词。
• 14:20Because if you use the AI's obvious word, the readers will smell that plastic apple.因为如果你用了AI提出的那种显而易见的词，读者就能闻出那股假苹果的味道。
• 14:25It is a brilliant strategy.这真是个绝妙的策略。
• 14:26You use the AI to brainstorm to generate a list of the most boring average ideas and then you intentionally avoid all of them.你用AI进行头脑风暴，列出一堆最平庸、枯燥的想法，然后刻意避开所有的这些选项。
• 14:33You don't let the machines speak for you.你绝不让机器替你发声。
• 14:35It's like having an incredibly fast editor sitting next to you at your desk.这就像你桌旁坐着一位手速飞快却毫无灵魂的编辑。
• 14:38But this editor has absolutely no soul.但这位编辑毫无灵魂。
• 14:41You give the editor rules.你给编辑制定规则。
• 14:43You ask them to check your spelling or structure.你让他们检查拼写或结构。
• 14:45But you hold the pen tightly in your own hand.但你始终把笔紧紧握在自己手中。
• 14:48This raises a very important question.这就引出了一个非常重要的问题。
• 14:50What actually happens when a human writer gets lazy and lets the AI hold the pen for a long time?如果人类作家偷懒，长时间让AI来执笔，究竟会发生什么？
• 14:56Ooh, what does happen?噢，到底会发生什么？
• 14:57Well, one hacker news commenter ran a fascinating experiment to find out.Hacker News上的一位评论者做了一个有趣的实验来一探究竟。
• 15:02They used an advanced model called GPT-5.6.他们使用了一个名为GPT-5.6的先进模型。
• 15:05I'm very curious about this.我对此非常好奇。
• 15:07What did the experiment look like?这个实验具体是怎样的？
• 15:08The user gave GPT-5.6 a prompt and asked it to write an entire book by itself and not just a short story, a huge epic book.用户给了GPT-5.6一个提示词，让它独立完成一整本书的创作，而且不是短篇故事，是一部鸿篇巨著。
• 15:17Roughly 110,000 words long.篇幅大约有11万词。
• 15:19110,000 words.11万词。
• 15:21That's the size of a giant Harry Potter book.那相当于一本超大部头《哈利·波特》的篇幅了。
• 15:23Did it actually manage to finish the story?它真的成功写完了整个故事吗？
• 15:26Physically, it produced the words.从形式上看，它确实把文字堆出来了。
• 15:27And in some technical ways, it was impressive.而且在某些技术层面上，确实令人惊叹。
• 15:30The AI was very good at remembering the basic mechanics of the plot, and knew exactly what town a character was in.AI非常擅长记住基本的情节机制，精准地掌握角色身处哪座城镇。
• 15:35It kept track of what items they were carrying in their backpack.它能清楚记录角色背包里装了哪些物品。
• 15:38Okay, that sounds pretty good so far.好吧，听起来到目前为止还算不错。
• 15:39It does.确实如此。
• 15:40But as the story got longer, the AI developed a strange obsession.但随着故事越来越长，AI产生了一种奇怪的偏执。
• 15:44It became obsessed with keeping all the rules perfectly strict.它变得极度痴迷于死板地遵循所有规则。
• 15:48Wait, what do you mean by obsessed with the rules?等等，痴迷于规则是什么意思？
• 15:49How does the computer enforce rules in a fiction book?计算机怎么在一本虚构小说里执行规则？
• 15:52It focused so entirely on making sure the timeline and the locations made logical mathematical sense that it completely forgot how to make the story fun.它全神贯注于确保时间线和地点在数学逻辑上严丝合缝，以至于完全忘记了该怎么让故事变得引人入胜。
• 16:01Oh, no.噢，不。
• 16:02Yeah, I forgot character motivation.是的，它忘记了人物动机。
• 16:04It forgot emotion.它忘记了情感。
• 16:05All the joy and surprise drained out of the text.所有的乐趣和惊喜都从文字中消失殆尽。
• 16:08It ended up sounding like a robot writing a police report about a fantasy world.最终读起来就像一个机器人在为奇幻世界写警方出警报告。
• 16:12Because it was treating the creative story like a math budget again.因为它又一次把创意写作当成了数学预算来算计。
• 16:15It wanted everything to be perfectly neat, tidy and balanced, but real human stories aren't neat.它希望一切都绝对工整、规矩、平衡，但真实的人类故事绝不是工整划一的。
• 16:21Life is messy.生活本就是混乱繁杂的。
• 16:22We make bad decisions based on feelings.我们会一时冲动做出糟糕的决定。
• 16:24We surprise each other.我们总会出人意料。
• 16:25Precisely.一点没错。
• 16:26The machine tried to optimize the story for logic and it killed the resonance.机器试图为了逻辑优化故事，结果扼杀了共鸣。
• 16:30So the best writers today use AI strictly to check the underlying structure.所以当今顶尖的作家严格只用AI来检查底层架构。
• 16:35They might ask the AI, does this chapter transition smoothly?他们可能会问AI：这一章衔接过渡顺畅吗？
• 16:38Did I forget a plot detail from chapter two?我有没有漏掉第二章里的某个情节细节？
• 16:40But the human keeps the intention.但创作意图始终由人类掌控。
• 16:42The human alone decides what the story actually means.唯有人类才能决定故事到底蕴含着什么意义。
• 16:45The human decides what emotion they want to trigger in the reader's mind.唯有人类才能决定想要在读者心中唤起怎样的情感。
• 16:50The human provides the resonance.共鸣源自人类本身。
• 16:52The AI is just a tool.AI仅仅是一个工具。
• 16:54It's a very fast, very smart dictionary.它是一部反应极快、极其聪明的词典。
• 16:57But it is not an author and it can't be.但它不是作家，也永远成不了作家。
• 16:59Let's summarize everything we have unpacked in this deep dive.让我们总结一下在这次深度探讨中所剖析的一切。
• 17:02We started out asking if human writing is safe from the rise of AI.我们一开始提出了一个问题：在AI崛起面前，人类写作是否安然无恙。
• 17:07We discovered that AI genuinely struggles with the wicked problem of human connection.我们发现，AI确实难以解决“人与人之间情感联结”这一极其棘手的难题。
• 17:11It has no real world experience so it can't feel.它没有现实世界的经历，因此根本无法感知情感。
• 17:14Right.对。
• 17:15It can't create that deep resonance.它无法创造那种深层次的共鸣。
• 17:16When it tries to write creatively it just produces plastic, solace, word, slump.当它试图进行创意写作时，产出的不过是僵化、空洞的文字堆砌。
• 17:22However, we also learned that the business of writing is changing brutally fast.然而，我们也了解到，写作行业正在以残酷的速度发生剧变。
• 17:26AI is successfully replacing the everyday, divider wall writing jobs.AI正在成功取代那些日常的、办公室隔间里的基础写作工作。
• 17:31The entry-level starter jobs that young writers rely on are rapidly going away.年轻写作者所依赖的初级起步岗位正在迅速消失。
• 17:35Which brings this whole conversation right back to you, the listener.这把整个话题重新引回到了你——也就是听众的身上。
• 17:38Think about the emails you write for your job every day.想一想你每天为了工作撰写的那些电子邮件。
• 17:40Think about the reports, the messages or the presentations you put together.想一想你整理的报告、沟通的消息或是制作的演示文稿。
• 17:44Are your words just elevator music?你的文字只是无足轻重的电梯背景音乐吗？
• 17:45That's the real test.这才是真正的考验。
• 17:46Are you just filling empty space with text because your boss expects a weekly report?你只是因为老板想要一份周报，就在用文字填补空白吗？
• 17:51Or are your words a costly signal?还是说，你的文字是一种“高代价信号”？
• 17:54Do they show real messy human thought?它们是否展现了真实、不规整的人类思想？
• 17:57Do they prove to the person reading them that you actually care?它们能否向读者证明你真的在乎？
• 18:01Because if your writing is just elevator music, an AI can probably do it for you right now.因为如果你的文字只是电梯背景音乐，AI现在大概就能替你代劳了。
• 18:05But if your writing is a costly signal, your unique human voice is more valuable today than it has ever been in history.但如果你的文字是一种高代价信号，那么你独一无二的人性之声在今天将比以往任何时候都更有价值。
• 18:11I want to leave you with one final, slightly provocative thought to explore on your own.最后，我想留给你一个略带启发性的思考，供你自行探索。
• 18:17We spent this deep dive talking about the importance of humans holding the pen, but what if society just stops doing that?我们在这次深度探讨中一直在谈论人类握笔的重要性，但如果整个社会干脆不再亲笔书写了呢？
• 18:24What do you mean?这是什么意思？
• 18:24If humans eventually stop writing all this basic everyday text, and we let machines write all our standard emails and instruction manuals and daily news updates would eventually happen to our own language.如果人类最终停止撰写所有这些基础的日常文本，放任机器去写完我们所有的标准邮件、说明手册以及每日新闻速递，我们的语言最终会发生怎样的变化？
• 18:35Oh, that is a deeply unsettling thought.噢，这真是一个令人深感不安的想法。
• 18:37Think about it.仔细想想吧。
• 18:37If you spend all day, every single day reading text that was generated by a math formula, will we eventually start talking like the machines?如果你每天每时每刻都在阅读由数学公式生成的文本，我们最终会不会也开始像机器一样说话？
• 18:45If we only consume AI logic, will our beautiful messy human language slowly change to become more robotic?如果我们只吸收AI的逻辑，我们那优美而繁复的人类语言是否会慢慢蜕变，变得像机器人一样？
• 18:52Simply because that's what we are used to reading.仅仅因为我们已经习惯了阅读那种文字。
• 18:54Wow.哇。
• 18:56Will we slowly become the plastic apples ourselves?我们自己最终也会慢慢变成那些塑料苹果吗？
• 18:58That is something to really chew on.这确实非常值得玩味。
• 19:00So thank you for joining us on this deep dive, keep reading, keep writing, and whatever you do, keep it human.感谢大家参与本次深度探讨，请保持阅读，坚持写作，无论做什么，都要保留人性的温度。