机器的崛起：一个被误解的科幻预言

Stop me if you've heard this one before: humanity gets better and better technology until, one day, while humanity is too busy fighting amongst itself to pay attention, the machine takes over. 如果这个故事你听过，请打断我：人类不断发展科技，直到有一天，当人类忙于内斗而无暇顾及之时，机器便掌控了一切。

Then, finally united against a common enemy, a group of rebels bands together to overthrow The Machine—it could be Skynet from the Terminator series, The Machines from the Matrix series, Ultron from Marvel, or if you're old like me, WOPR from War Games. 然后，一群反抗者最终团结起来对抗共同的敌人，推翻了机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）——它可能是《终结者》系列中的天网，可能是《黑客帝国》系列中的机器大军，可能是漫威中的奥创，如果你和我一样老，也可能是《战争游戏》中的WOPR。

Most of the time, The Machines start by triggering a nuclear war, or they're stopped from doing so at the last minute. 大多数时候，这些机器会引发一场核战争，或者在最后一刻被阻止。

Usually, the story starts after the humans have already lost; fiction rarely shows us the run-up to the machines taking over. 通常，故事开始时人类已经失败；小说很少向我们展示机器接管之前的过程。

The closest I can think of would be the last couple of seasons of the Person of Interest TV series. 我能想到的最接近的例子是电视剧《疑犯追踪》的最后几季。

Now, given the current state of discourse about ChatGPT and this video's title, you're probably almost certainly expecting me to start talking about the dangers of generative AI, or AGI, or artificial super intelligence, about how we have to act right now to stop AI, to prevent the coming war. 鉴于目前关于ChatGPT的讨论现状以及本视频的标题，你几乎肯定会认为我将开始谈论生成式人工智能（Generative AI：能够生成文本、图像或其他媒体内容的人工智能）、通用人工智能（AGI: Artificial General Intelligence: 具备人类智能水平并能执行任何智力任务的人工智能），或人工超级智能（Artificial Super Intelligence: 智能远超人类的人工智能）的危险，以及我们必须立即采取行动阻止AI、预防即将到来的战争。

But I'm about to argue that in virtually every way that matters, it's far too late to stop The Machine from taking over. 但我将要论证的是，在几乎所有重要的方面，阻止机器接管一切已经为时过晚。

And before you say that ChatGPT hasn't gotten that advanced, or it's nowhere near taking over, I'm not talking about that. 在你反驳说ChatGPT还没有那么先进，或者它离接管一切还很远之前，我想说我谈论的并非如此。

I would argue that The Machines took over, for at least most practical purposes, more than a decade before ChatGPT, and all that sci-fi failed to prepare us for it. 我认为，至少在大多数实际应用中，机器早在ChatGPT出现十多年前就已经接管了，而所有的科幻作品都未能让我们为此做好准备。

Well, maybe Black Mirror, but only after it was already far too late. 或许《黑镜》有所涉及，但那时也已经太晚了。

And it wasn't a big war against The Machine; it was gradual. 这不是一场与机器的大战，而是一个渐进的过程。

It is not a centralized, all-knowing, self-aware super intelligence. 它不是一个中心化的、全知全能的、有自我意识的超级智能。

The same way that what we call The Internet is really a collection of computers and networks that are united by common protocols, The Machine is a collection of a whole bunch of interconnected, smaller installations and components united by common technological features and a set of perverse incentives that I'm going to talk about today. 就像我们所说的互联网实际上是大量由通用协议连接起来的计算机和网络的集合一样，机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）也是大量相互连接的、更小的装置和组件的集合，它们由共同的技术特征和一套我今天将要讨论的扭曲激励（Perverse Incentives: 导致与预期相反或不良结果的激励机制）所连接。

And it has arguably decided, without deliberately intending to, most of the major elections in the last 15 or so years. 可以说，在过去15年左右的时间里，它在无意中决定了大多数重要的选举。

It picks which policies do or don't get adopted, who gets famous and who gets canceled, who exits the EU and who remains, which products go viral and which get boycotted. 它决定了哪些政策被采纳或不被采纳，谁能成名谁会被“取消”，谁退出欧盟谁留下，哪些产品会爆红哪些会被抵制。

It even dictates which victims of violence become martyrs for a cause and which just become statistics. 它甚至决定了哪些暴力受害者会成为事业的烈士，哪些仅仅沦为统计数字。

And if that wasn't bad enough—and it's bad—modern generative AI is making it worse, fast, really fast. 如果这还不够糟糕——而它确实很糟糕——现代生成式人工智能正在迅速、非常迅速地使其变得更糟。

And the worst part is that we, the humans, not only didn't form an organized resistance, but we're ignoring it completely because it tricked us into fighting with each other and thinking that we're so in charge. 最糟糕的是，我们人类不仅没有组织起有效的抵抗，反而完全忽视了它，因为它欺骗我们互相争斗，并让我们以为自己掌控着一切。

And we, the human race, better get our act together, because otherwise we are f----- 我们人类最好团结起来，否则我们将万劫不复。

Welcome to the Internet of Bugs. My name is Carl. I've been a software professional for more than 35 years now, and today I'm going to tell you a story about where we are, how we got here, and how it's about to get a whole lot worse unless we do something. 欢迎来到“Bug的互联网”。我叫卡尔。我从事软件行业已经超过35年了，今天我将告诉你们一个关于我们身处何方、如何走到这一步，以及除非我们采取行动，否则情况将如何变得更糟的故事。

社会极化：机器的产物而非原因

So I live in the United States, where research says that we are more polarized as a society than at any time since the US Civil War in the 1860s. 我住在美国，研究表明，我们社会目前的两极分化程度是自19世纪60年代美国内战以来任何时期都未曾有过的。

Whether we're on the liberal side or the conservative side, most of us have family members or friends that either we no longer have contact with, or that we can only talk to if we walk on eggshells and go out of our way to avoid any kind of political subject. 无论我们是自由派还是保守派，我们大多数人都有一些亲友，要么我们不再联系，要么我们只能小心翼翼地交谈，竭力避免任何政治话题。

It's made us as a society largely non-functional and largely unable to even have substantive conversations with each other because, given a random selection of Americans, you're likely to end up with a group of people that can't even agree on basic facts. 这使得我们作为一个社会在很大程度上无法正常运作，甚至无法进行实质性的对话，因为随意挑选一群美国人，你很可能会发现他们连基本事实都无法达成一致。

And it's not even that they're going to end up in two groups; there are a number of issues that divide conservatives and a number of issues that divide liberals. 甚至他们都不会只分成两派；有许多问题将保守派内部撕裂，也有许多问题将自由派内部撕裂。

Now, I've read a bunch of books on why we're polarized, including this book: Why We're Polarized and a bunch of others. 我读过很多关于我们为何两极分化的书，包括这本《我们为何两极分化》以及其他许多。

And I have to say, I don't think I've ever been as disappointed in journalists as I am over this. 我不得不说，我从未像现在这样对记者感到失望。

I don't have time to get into a point-by-point breakdown of this literature. 我没有时间逐点分析这些文献。

Just let me tell you what I think they all missed, because no one has explained, or from what I've seen, even proposed any answer to the question: "So why now?" 我只想告诉你们我认为他们都错过了什么，因为没有人解释过，或者说据我所见，甚至没有人提出过对这个问题的任何答案：“为什么是现在？”

Sure, there are issues that divide us, but that's always been true. 当然，有些问题会分裂我们，但这种情况一直都存在。

If we're more polarized now than at any time since the Civil War, then that means that there's something now that's more impactful than anything that existed during the Great Depression, the Vietnam War, or Watergate, or any other period of strife. 如果我们现在比内战以来的任何时期都更加两极分化，那么这意味着现在存在某种比大萧条、越南战争、水门事件或任何其他冲突时期都更具影响力的事情。

But there's no such issue with these books. 但这些书里并没有提到这样的问题。

When people talk about why we're polarized, it's a laundry list of reasons ranging from inequality, which was a lot worse in the 1920s leading up to the Great Depression, but polarization apparently wasn't. 当人们谈论我们为何两极分化时，原因不胜枚举，从不平等（1920年代大萧条前夕情况更糟，但两极分化程度显然不如现在），到摇摇欲坠的机构、深层政府、价值观的丧失、24/7新闻周期、虚假信息或社交媒体，以及我能找出的一大堆其他书籍。

The crumbling institutions, the deep state, loss of values, 24/7 news cycles, disinformation, or social media, or a bunch of other books that I could pull out. 摇摇欲坠的机构、深层政府、价值观的丧失、24/7新闻周期、虚假信息或社交媒体，以及我能找出的一大堆其他书籍。

And that is to even include the ones that I've only read electronically. 这甚至包括那些我只读过电子版的书。

Now, all those factors are bad, but they're all just small pieces of the picture. 现在，所有这些因素都很糟糕，但它们都只是整个图景的一小部分。

And I haven't found anyone who has tied them all together into a coherent narrative, even though the underlying structure seems pretty clear to me from where I'm sitting. 我还没有找到任何人能将它们整合到一个连贯的叙述中，尽管在我看来，其底层结构已经相当清晰。

The Machine is what ties it all together. 正是机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）将这一切联系在一起。

It both provides the tools that are being used to isolate us from each other, and at the same time, it pays the unscrupulous humans among us to do just that. 它既提供了将我们彼此隔离开来的工具，同时又付钱给那些我们中间不择手段的人来做这件事。

Unlike the Civil War, the polarization now isn't a side effect of a cultural divide over a particular issue. 与内战不同，现在的两极分化并非某个特定问题上文化分歧的副作用。

The opposite, in fact: the cultural divide over a bunch of issues is a result of the polarization, and the polarization is being created deliberately because the more polarized we get, the more money they make. 事实上恰恰相反：许多问题上的文化分歧是两极分化的结果，而两极分化是被人为制造出来的，因为我们越是两极分化，他们赚的钱就越多。

机器的起源：从尼尔森收视率到广告技术

The story of The Machine, as I said earlier, does not begin with ChatGPT. 正如我之前所说，机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）的故事并非始于ChatGPT。

It actually starts in the 1950s, more than 20 years before I was born. 它实际上始于20世纪50年代，比我出生早了20多年。

The best source I found about the humble origins of The Machine is this book: Selling the American People: Advertising Optimization, and the Origins of AdTech. 我找到的关于机器卑微起源的最佳资料是这本书：《销售美国人民：广告优化与广告技术（AdTech: Advertising Technology: 用于管理、执行和优化数字广告活动的软件和工具）的起源》。

It's released under a Creative Commons license, and the content is available for free from MIT Press. Link below. 它以知识共享许可发布，内容可从麻省理工学院出版社免费获取。链接在下方。

It's not the easiest book for a casual reader, as it appears to be done life as a PhD dissertation, but it's incredibly well-researched, and its bibliography was a jumping-off point for a bunch of other stuff I had to read to understand the background of what happened before I got involved in all this. 对于普通读者来说，这本书并不容易读懂，因为它看起来像是一篇博士论文，但它研究得非常透彻，其参考文献是我为了理解在我涉足这一切之前所发生的事情的背景而必须阅读的许多其他资料的起点。

I've also linked a YouTube video of the author giving a presentation about the book, and that video is a lot more approachable for people who haven't spent as many years reading academic papers as I have. 我还附上了作者关于这本书的演讲视频链接，对于那些不像我一样花多年时间阅读学术论文的人来说，那个视频更容易理解。

What started in the 50s was something called the Nielsen ratings (Nielsen ratings: 衡量电视和广播节目观众规模和构成的系统), which is a feedback loop between what content was being produced and what people were watching it, where the content at the time would have been TV and radio broadcasts. 20世纪50年代开始的是尼尔森收视率（Nielsen ratings: 衡量电视和广播节目观众规模和构成的系统），这是一个关于内容制作与观众收看情况之间的反馈循环，当时的内容主要是电视和广播节目。

So instead of knowing more people bought the newspaper today than yesterday or last week, they now moved down to the individual show, and they knew more or less the age ranges of the people who were watching or listening to each show. 因此，他们不再仅仅知道今天报纸销量比昨天或上周多，而是深入到每个节目，大致了解每个节目的观众或听众的年龄段。

And as more and more shows were produced, and the industry collected more and more data about who was watching them, it required more sophisticated information technology to keep track of all of it, and that's where, albeit in a really primitive incarnation, The Machine was born. 随着越来越多的节目被制作出来，以及行业收集到越来越多关于谁在观看这些节目的数据，这就需要更复杂的信息技术来追踪所有这些信息，正是在这里，尽管是以一种非常原始的形式，机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）诞生了。

And as The Machine collected that information, it was being put to other purposes. 随着机器收集这些信息，它被用于其他目的。

The list of advertising firms and other players in Selling the American People who figured out how to collect and understand the data has quite a bit of overlap with the list of firms and other players in this book: Golden Holocaust, which is all about how, around the same time frame, the tobacco industry, with the help of advertising and public relations experts, lied and manipulated the public to push more addictive cigarettes on the population and made tons of money in the process and sent countless numbers of lung cancer patients early graves. 《销售美国人民》中那些懂得如何收集和理解数据的广告公司及其他参与者的名单，与《金色大屠杀》一书中提到的公司和参与者名单有相当多的重叠。后者讲述了在同一时期，烟草行业如何在广告和公共关系专家的帮助下，欺骗和操纵公众，向大众推销更具成瘾性的香烟，并在此过程中赚取巨额利润，同时导致无数肺癌患者英年早逝。

The learning from the one enabled the other. 前者的经验促成了后者的发展。

新闻与网络的演变：机器的扩张

Fast forward to the 1980s, to the birth of the 24-hour news channels. 快进到20世纪80年代，24小时新闻频道诞生。

Now news was also competing for ratings and therefore advertising dollars. 现在，新闻也开始争夺收视率，从而争夺广告收入。

Sensational journalism had always sold newspapers, but now, by having shows with different personalities and different takes on current events, figuring out how to tailor the news content to attract more viewers was getting easier and more lucrative, and The Machine grew bigger, this time facilitating, among other things, the US invasion of Iraq under false pretenses, which left countless numbers of Iraqi civilians dead in its wake. 耸人听闻的新闻报道一直都能卖报纸，但现在，通过推出具有不同个性和对时事有不同看法的节目，找出如何调整新闻内容以吸引更多观众变得更容易、更有利可图，机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）也随之壮大。这一次，它促成了（除其他事项外）美国以虚假借口入侵伊拉克，导致无数伊拉克平民丧生。

Fast forward again, now to analytics and advertising on the web that tracked exactly what web pages were seen by exactly what people, and built up profiles of those people so that they knew who was reading what. 再次快进，现在到了网络上的分析和广告，它们精确追踪哪些网页被哪些人看过，并建立这些人的档案，从而知道谁在阅读什么。

And now it was possible to know not only what sites people were watching, but exactly which stories people were paying attention to and exactly what kinds of people were paying attention to them. 现在，不仅可以知道人们在看哪些网站，还能精确知道人们关注哪些故事，以及是哪类人在关注它们。

Around this time, I was working at a startup here in Austin that provided comments and ratings and engagement tracking on stories at websites from USA Today to San Francisco Chronicle to Better Homes and Gardens. 大约在这个时候，我在奥斯汀的一家初创公司工作，该公司为从《今日美国》到《旧金山纪事报》再到《美好家园》等网站的故事提供评论、评分和互动追踪服务。

There, not realizing the consequences, I made some of my personal contributions to the technology that made up The Machine, but unfortunately, it wasn't the last time in my career I would do so. 在那里，我没有意识到后果，为构成机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）的技术做出了一些个人贡献，但不幸的是，这并非我职业生涯中最后一次这样做。

This is around the timeline where Selling the American People ends and The Age of Surveillance Capitalism picks up. 这大约是《销售美国人民》一书结束，而《监控资本主义时代》（The Age of Surveillance Capitalism: 一本探讨数字经济中数据收集、分析和利用对社会和个人影响的书籍）开始的时期。

I'm guessing you have an idea of what happened next. 我猜你大概知道接下来发生了什么。

社交媒体与智能手机：机器的精细化操控

It was social media and the smartphone. 那就是社交媒体和智能手机。

Now, everything about you could be tracked, from your shopping habits to your GPS coordinates, and that could all be correlated with everything that you clicked on. 现在，关于你的一切都可以被追踪，从你的购物习惯到你的GPS坐标，所有这些都可以与你点击过的所有内容关联起来。

More information was being collected and processed than ever before, and The Machine that was processing all that information was getting incredibly sophisticated. 前所未有的大量信息被收集和处理，而处理所有这些信息的机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）也变得异常复杂。

So much so that there were cases where the machine could deduce that a teenage girl was pregnant even before her father knew. Link to that story below. 以至于在某些案例中，机器甚至能在一位父亲知晓之前，就推断出他的十几岁女儿已经怀孕。相关故事链接在下方。

It was a doozy. 那真是个惊人的案例。

And the more The Machine knows about what content draws more attention and traffic, the better it can help people predict what new content to create to get even more attention and traffic. 机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）越了解什么内容能吸引更多关注和流量，就越能帮助人们预测要创作什么新内容才能获得更多关注和流量。

Analytics continued to inform the science of psychology and behavioral manipulation. 数据分析持续为心理学和行为操控科学提供信息。

Now, in addition to huge, expensive PR campaigns, like the ones that kept cigarettes everywhere long after the science was clear that they were killing people, millions of tiny experiments could be run every day to see who would read what story, who would share what headline, and so on. 现在，除了那些耗资巨大、代价昂贵的公关活动（比如在科学明确证明香烟致人死亡很久之后，仍使其无处不在的宣传），每天还可以进行数百万个微小实验，以观察谁会阅读什么故事，谁会分享什么标题等等。

The more The Machine knew about people, the better it got at predicting and manipulating them. 机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）对人类了解得越多，就越擅长预测和操控他们。

And the better it got at manipulating people, the more effective the ads became, and the more money flowed through the system, funding The Machine. 它越擅长操控人类，广告就越有效，流经系统的资金就越多，从而为机器提供资金。

Now, you've probably heard people complain about the algorithm (Algorithm: 一系列指令或规则，用于解决问题或执行任务，在此指社交媒体用于推荐内容的系统) as the thing that they hated about social media. 现在，你可能听过人们抱怨算法（Algorithm: 一系列指令或规则，用于解决问题或执行任务，在此指社交媒体用于推荐内容的系统），认为它是社交媒体最令人讨厌的地方。

This is the time period where the algorithm was created, and it's just another component of that machine. 这就是算法被创造出来的时期，它只是那台机器的另一个组成部分。

Now that we've had a very abbreviated version of what led us here, let's talk about where we actually are. 既然我们已经非常简略地回顾了导致我们走到这一步的原因，现在让我们来谈谈我们实际身处何方。

And then, sigh, we'll talk about where it looks like we're headed. 然后，唉，我们再来谈谈我们似乎正走向何方。

创作者经济与极化内容的制造

So if you're watching this, I don't need to tell you what social media is, and I assume you've heard people complaining about how social media has damaged public discourse. 如果你正在观看此视频，我无需告诉你社交媒体是什么，而且我猜你一定听过人们抱怨社交媒体如何损害了公共话语。

But there's another side of social media that people don't see, and it's a critical piece of the puzzle. 但社交媒体还有不为人知的一面，它是解决这个难题的关键部分。

Although there's been a lot of talk about how the algorithms the social media companies train will route people to extreme opinions and falsehoods and conspiracy theories to increase people's social media time, there hasn't been nearly enough discussion of the fact that the algorithm doesn't just create polarizing content from nowhere. 尽管人们一直在讨论社交媒体公司训练的算法如何将人们引向极端观点、虚假信息和阴谋论，以增加人们使用社交媒体的时间，但对于算法并非凭空制造两极分化内容这一事实，讨论却远远不够。

All it does is amplify it, and the algorithm can't amplify crazy, completely fabricated conspiracy theories until somebody posts a conspiracy theory in the first place. 它所做的只是放大这些内容，而且算法无法放大那些疯狂的、完全捏造的阴谋论，除非有人首先发布了阴谋论。

How and why that happens? See, that's another real problem. 这如何发生，为何发生？看，这又是另一个真正的问题。

Let me give you a peek behind the curtain of the creator economy (Creator Economy: 指创作者通过其内容直接从受众或品牌获得收入的经济模式). 让我带你一窥创作者经济（Creator Economy: 指创作者通过其内容直接从受众或品牌获得收入的经济模式）的幕后。

There is an insane amount of advertising money up for grabs for the people that can capture the attention of the right demographics. 对于那些能够吸引到正确受众群体注意力的人来说，有巨额的广告费可供争夺。

As with most of the distribution of insane amounts of money, the vast majority of the money goes to the people at the very top—the Mr. Beasts and Joe Rogans of the world—and the vast majority of the people making content earn very little or most likely nothing at all. 与大多数巨额财富分配一样，绝大部分资金流向了最顶层的人——比如Mr. Beast和Joe Rogan这类人物——而绝大多数内容创作者收入甚微，甚至一无所有。

But unlike a lot of professions, there's technically nothing to stop any random creator from getting as big as Mr. Beast. 但与许多职业不同，理论上没有任何东西能阻止任何随机的创作者变得像Mr. Beast一样成功。

There aren't as many gatekeepers as there have always been before, and so a large fraction of those content creators probably want to get that big, and far too many of them think that they can, and a lot of those are willing to do absolutely anything to do so. 现在没有以前那么多的“守门人”，因此很大一部分内容创作者可能都想变得那么出名，而且太多人认为他们可以做到，其中很多人甚至愿意为此不惜一切代价。

And this created a whole other industry: the social media gurus, a number of groups and sites where people share or sell techniques for trying to get more clicks, more traffic. 这催生了另一个完整的行业：社交媒体大师，以及许多团体和网站，人们在那里分享或出售旨在获取更多点击和流量的技巧。

Sometimes this was just "how to troll the most people" threads on forums. 有时这只是论坛上“如何最大程度地激怒他人”的帖子。

Sometimes it was classes like "how to make guaranteed money from social media using this proven five-step secret formula" or the ever-present, "how to make thousands of dollars a month with a faceless AI YouTube channel" grift. 有时是诸如“如何使用这套经过验证的五步秘密公式，从社交媒体上赚取有保障的钱”之类的课程，或者无处不在的“如何通过一个无面AI YouTube频道每月赚取数千美元”的骗局。

But more about that one later. 但稍后再详细讨论那个。

And the rules are simple: there's an insane amount of money that any creator can have if only they can get people to interact with their content instead of other people's content. 规则很简单：只要创作者能让人们与自己的内容互动，而不是与他人的内容互动，他们就能获得巨额财富。

And to go along with the same amount of money, there's also an insane amount of analytics data available to help content creators figure out what content gets used and what doesn't. 除了同样多的金钱，还有海量的分析数据可供内容创作者使用，帮助他们弄清楚哪些内容受欢迎，哪些不受欢迎。

And there are lots of books and literature from the 75 years of psychological manipulation experimentation that has run through the machine since its inception. 而且，自机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）诞生以来，75年间通过它进行的心理操纵实验产生了大量的书籍和文献。

This is a recipe for a disaster, and for polarization. 这简直是灾难和两极分化的温床。

We didn't realize it was happening, but we have built the world's largest experimental psychology platform, and it is being used to manipulate the vast majority of the human race, and it has been getting better and better at manipulating people day after day, month after month for 15 or 20 years now. 我们没有意识到这一切正在发生，但我们已经建立了世界上最大的实验心理学平台，它正被用来操纵绝大多数人类，而且在过去的15或20年里，它日复一日、月复一月地在操纵人类方面变得越来越好。

And for those of you that insist on saying, "I don't care if they track my every move, I have nothing to hide." 对于那些坚持说“我不在乎他们追踪我的一举一动，我没有什么可隐瞒的”的人。

Understand that you, just by interacting with or avoiding or ignoring certain content, are giving the bad guys a ton of information that, when combined with enough other people to be statistically significant, allows the machine to fine-tune different messages to get better and better at understanding the science of manipulating human psychology. 请明白，你仅仅通过与某些内容的互动、回避或忽视，就向那些“坏人”提供了大量信息。这些信息，当与足够多的人结合起来达到统计学意义时，能让机器微调不同的信息，从而越来越擅长理解操纵人类心理的科学。

You might not think it affects you. You might insist that you're immune and you never click on anything. 你可能认为这不会影响你。你可能坚称自己免疫，从不点击任何东西。

And even if that were true—and let's be honest, it probably isn't—just by timing how long it takes you to scroll past different content, gives them information on how much or how little that item caught your attention. 即使那是真的——老实说，很可能不是——仅仅通过计时你滚动浏览不同内容所需的时间，就能让他们知道该内容吸引了你多少或多大的注意力。

And if you have a smartwatch or a front-facing camera that scans your face and can see facial expression and track your eyes or wireless earbuds, it's in your pulse, imagine how valuable that info is. 如果你有智能手表，或者能扫描你面部、识别面部表情和追踪你眼睛的前置摄像头，或者无线耳机能监测你的脉搏，想象一下这些信息有多宝贵。

And when combined with measurements taken from enough other people, the mountain of data they've collected from you and others has been weaponized against your fellow humans. 当这些数据与从足够多其他人那里收集到的测量结果结合起来时，他们从你和他人那里收集到的海量数据就被武器化，用来对付你的人类同胞。

And just by participating on those platforms, we all contribute to that. 仅仅通过参与这些平台，我们都在为此添砖加瓦。

Now, that said, individuals opting out one at a time are not going to make enough of a dent to matter. 话虽如此，个人一个接一个地退出并不会产生足够大的影响。

If this is going to get better, it's going to take a concentrated effort, but that'll have to be a later video. 如果情况要好转，需要集中努力，但这将是以后视频的内容。

机器的谎言：制造冲突与信息茧房

At this point in time, media properties and content creators are no longer being rewarded for informing or educating or being trusted. 在当下，媒体机构和内容创作者不再因提供信息、教育或赢得信任而获得回报。

They're only rewarded for taking up your time so you can look at ads, and the race is on to make the most money by getting the most traffic. 他们只因占用你的时间让你看广告而获得奖励，而通过获取最大流量来赚取最多钱的竞赛正在进行。

Because you know what they can do with the machine once they're getting paid solely based on likes and shares and clicks? 因为你知道一旦他们仅仅根据点赞、分享和点击量获得报酬，他们能用机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）做什么吗？

They can just start making things up. 他们可以开始凭空捏造事实。

Create conspiracy theories out of whole cloth and throw them at different groups of people and see if they stick. 凭空捏造阴谋论，抛给不同的群体，看看是否有人相信。

Make up a story that there's a child abduction ring located in a basement of a pizzeria in Washington, D.C. that doesn't even have a basement. 编造一个故事，说华盛顿特区一家根本没有地下室的比萨店里有一个儿童绑架团伙。

And if enough people click on the story, cool, they make money. 如果足够多的人点击这个故事，太棒了，他们就赚钱了。

And they can double down and make up more stories about that basement and more stories about child abduction rings and see if they can make money from that, too. 他们可以加倍努力，编造更多关于那个地下室的故事，更多关于儿童绑架团伙的故事，看看是否也能从中赚钱。

And if people don't click on it, they just cross that one off the list and make something else up. 如果人们不点击，他们就把它从列表中划掉，再编造其他东西。

Lather and repeat until something works. 反复尝试，直到奏效。

There's been a lot of discourse about ads that push political misinformation on social media, especially with respect to the Brexit referendum in the U.S. presidential elections. 关于在社交媒体上推动政治虚假信息的广告，已经有很多讨论，尤其是在英国脱欧公投和美国总统选举方面。

And intentionally misleading political ads are a problem, but only during election season, which, despite how it feels, is only a fraction of the time. 故意误导的政治广告确实是个问题，但只在选举季期间，而选举季，尽管感觉很长，却只占一小部分时间。

And without the infrastructure of The Machine, those ads wouldn't be able to find the right audience. 如果没有机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）的基础设施，那些广告将无法找到合适的受众。

Before you get upset at the mention of politics, though, I'm not here to tell you whether you should be liberal or conservative or what stance you should take on any of the issues or what ads are or aren't misleading. 不过，在你因提到政治而感到不快之前，我并不是来告诉你应该成为自由派还是保守派，或者在任何问题上应该采取何种立场，也不是来评判哪些广告具有误导性或不具有误导性。

This isn't about what issues we're fighting about or what side of them people should be on. 这无关我们争论的是什么问题，也无关人们应该站在哪一边。

It's about how the fighting itself has become more important than the actual issues. 它关乎的是争斗本身已经变得比实际问题更重要。

Both sides routinely lie to their supporters. 双方都惯常对自己的支持者撒谎。

Like last year here in the U.S., one side was lying about who would pay the tariffs, and the other side was lying about how healthy and mentally capable their elderly candidate wasn't. 就像去年在美国，一方在关税由谁支付的问题上撒谎，而另一方则在他们的年迈候选人身体和精神状况不佳的问题上撒谎。

And what's sad is, from that description, either lie could apply to either side. 而令人悲哀的是，根据这种描述，任何一个谎言都可能适用于任何一方。

Now, I'm trying to get you to realize that politicians aren't the problem. 现在，我正试图让你意识到政客并非问题所在。

Well, okay, politicians are always a problem, but they aren't the worst problem or the main problem. 好吧，政客总是个问题，但他们不是最糟糕的问题，也不是主要问题。

This goes far beyond politics. 这远远超出了政治范畴。

People's entire identities, values, and worldviews are being overridden. 人们的整个身份、价值观和世界观正在被颠覆。

I've had conversations with members of my extended family where there are basically no facts related to news or politics that we can agree on. 我曾与我的大家庭成员进行过对话，我们基本上无法就任何与新闻或政治相关的事实达成一致。

And I don't mean disagreeing about whether something was good or bad or whether we like a person or not. 我不是指对某件事是好是坏，或者我们是否喜欢某个人存在分歧。

I mean, we've had entire conversations where they've had a list of facts, events, and statistics that they got from their information bubble (Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境). 我的意思是，我们进行过完整的对话，他们有一份来自他们的信息茧房（Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境）的事实、事件和统计数据清单。

And I've had a list of facts and events and statistics that I got from my information bubble on the same topic, and their list and my list have nothing whatsoever in common. 而我有一份来自我的信息茧房的关于同一主题的事实、事件和统计数据清单，他们的清单和我的清单之间没有任何共同之处。

There's just no way to have a meaningful conversation out of those circumstances. 在那种情况下，根本无法进行有意义的对话。

You either have to stop talking or you have to change the subject. 你要么停止交谈，要么改变话题。

It's like Invasion of the Body Snatchers was a documentary. 这就像《天外魔花》是一部纪录片一样。

People are going to get mad at me for saying this, but the current political crisis, despite it being a matter of life and death for many people, is not itself the problem. 人们会因为我这样说而生我的气，但当前的政治危机，尽管对许多人来说是生死攸关的问题，它本身并不是问题。

If we don't figure out how to resolve our societal fragmentation, it's going to be crisis after crisis after crisis. 如果我们不找出如何解决社会碎片化的方法，那么将是危机接踵而至。

And anyone who thinks that we become a functional society again if we could just get rid of any particular politician or come to agreement on any particular hot-button issue is believing the lie that The Machine is pushing. 任何认为只要我们能摆脱某个特定的政客，或者在某个特定的热门问题上达成一致，我们就能再次成为一个正常运作的社会的人，都在相信机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）所推销的谎言。

You see, The Machine doesn't just tell people what to believe; it also tells them who to hate. 你看，机器不仅仅告诉人们应该相信什么，它还告诉他们应该憎恨谁。

Unscrupulous content creators run experiments on the other side of social media too. 不择手段的内容创作者也在社交媒体的另一面进行实验。

They make it fake social media users: bots (Bots: 自动化程序，模拟人类行为在网络上执行任务). Lots of them. 他们制造虚假的社交媒体用户：机器人（Bots: 自动化程序，模拟人类行为在网络上执行任务）。数量众多。

And they make the bots act the way that a certain group of people act according to their data. 他们让机器人根据他们的数据，模仿特定人群的行为方式。

They see what pieces of content that their competitors are serving to users in that group so they can adjust their output accordingly. 他们观察竞争对手向该群体用户推送了哪些内容，以便相应地调整自己的产出。

And this has been happening for much, much longer than ChatGPT has been around, at least since 2017. 这种情况发生的时间比ChatGPT出现的时间要长得多，至少从2017年就开始了。

And unscrupulous content creators don't want competitors serving content to people that they're targeting because they don't want to share ad revenue. 不择手段的内容创作者不希望竞争对手向他们锁定的用户推送内容，因为他们不想分享广告收入。

See what they can do next. 看看他们接下来能做什么。

When the bots uncover what content is popular from the same other stories, they just invent stories about why those things are false, harmful, even dangerous to children, even, and they have their bots reinforce those stories. 当机器人发现其他相同故事中哪些内容受欢迎时，它们就会编造故事，解释为什么这些内容是虚假的、有害的，甚至对儿童构成危险，然后让它们的机器人强化这些故事。

They shout from the rooftops that the crap that they made up is vitally important for your future and the safety of children, and the stuff that anyone else made up, they say that that's all lies and it's all paid for by George Soros or the Koch brothers or some other source they've cast as a political villain. 他们大声疾呼，他们编造的那些垃圾对你的未来和儿童的安全至关重要，而其他人编造的东西，他们则说那全是谎言，都是由乔治·索罗斯或科赫兄弟或其他他们塑造成政治反派的来源资助的。

And this is where things get really scary, because the easiest way for them to convince the people to believe that their content is true and other content is lies is to get the people who are watching their content to dislike and distrust and potentially even hate other content creators, other news outlets, other brands, other narratives, other political parties, even other family members, pretty much anyone and any content outside the information bubble that they've been pigeonholed into, which just further isolates them so the only "facts" they ever see on their feed are intended only to enrage them and anything contains "facts" outside that bubble they've been conditioned to despise and to distrust. 这就是事情变得真正可怕的地方，因为他们让人们相信自己的内容是真实的、而其他内容是谎言的最简单方法，就是让那些观看他们内容的人去厌恶、不信任，甚至可能憎恨其他内容创作者、其他新闻机构、其他品牌、其他叙事、其他政党，甚至其他家庭成员，几乎是任何被他们限制在信息茧房（Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境）之外的人和内容。这只会进一步孤立他们，使得他们在信息流中看到的唯一“事实”都旨在激怒他们，而任何包含“事实”的茧房之外的内容，他们都被训练去鄙视和不信任。

And if those people, as a side effect, end up with nonsensical, radically false ideas about what's going on in politics in a way that affects their vote, or they're incited to violence on a college campus, as far as the people that are making the content are concerned, it's just incidental. 如果这些人，作为副作用，最终对政治中发生的事情产生荒谬、完全错误的看法，从而影响他们的投票，或者他们在大学校园里被煽动暴力，就内容创作者而言，这都只是附带的。

They only care about ad revenue, and they will do almost anything to make more of it, and they don't care about any fallout. 他们只关心广告收入，他们几乎会不惜一切代价去赚取更多，并且不关心任何后果。

Understand that there are hundreds, or thousands, or maybe even tens of thousands of these video, podcast, and other content creation operations who have acquired sophisticated software that they are using to track as many different types of users as they can, and they are trying to find and push pieces of content that are optimized to attract the attention of each different user type they're tracking, and also optimize to get those users to hate the kinds of content from other sources. 请理解，有成百上千，甚至可能上万个这样的视频、播客和其他内容创作机构，他们已经获得了复杂的软件，用来追踪尽可能多的不同类型的用户。他们正试图寻找并推送经过优化的内容，以吸引他们追踪的每种不同用户类型的注意力，并优化这些内容，让这些用户憎恨来自其他来源的内容。

And I, as a content creator, get inundated with proposals from people who want me to pay them to help make my content go viral, and what I would get for my money is their special software and algorithms and skills and techniques and yada yada yada. 而我，作为一名内容创作者，收到了大量提案，那些人希望我付费给他们，帮助我的内容走红，而我用钱换来的将是他们特殊的软件、算法、技能、技术等等。

And those people don't want to sell just to me; they want to sell to everyone that they can. 这些人不只想卖给我，他们想卖给所有能卖的人。

So, now we have another piece of The Machine: a few groups of very experienced coders who are each building their own software suites that track users and make up content to optimize for user types and all that other stuff that I've been talking about, and they sell access to those tools to a bunch of content creators who then use those tools to try and get as much of each social media network's ad revenue as they can. 所以，现在我们有了机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）的另一部分：少数几组经验丰富的程序员，他们各自构建自己的软件套件，用于追踪用户、制作内容以优化用户类型以及我一直在谈论的所有其他东西。他们将这些工具的访问权限出售给大量内容创作者，这些创作者随后使用这些工具，试图尽可能多地获取每个社交媒体网络的广告收入。

And those tools builders are competing with each other to be the group that gets the biggest number of content creators paying them for tools, even if all or most of the content creators competing with each other are all paying the same tools company for the same set of tools to use against each other. 而这些工具的开发者们彼此竞争，都想成为拥有最多内容创作者付费使用其工具的团队，即使所有或大多数相互竞争的内容创作者都在向同一家工具公司付费，使用同一套工具来互相攻击。

404media has done a lot of great work exposing this; link on that below. 404media在这方面做了很多出色的揭露工作；相关链接在下方。

生成式AI的未来：更深层的个体操控

Which brings us to the one major piece of the puzzle left: where we're headed from here. 这把我们带到了最后一个主要的谜题：我们接下来将走向何方。

I told you we'd come back to it, and that's ChatGPT and the other new AIs that compete with ChatGPT. 我告诉过你我们会回到这个话题，那就是ChatGPT以及其他与ChatGPT竞争的新人工智能。

With respect to this decade-old interconnected network of people trying to monopolize social media ad dollars, ChatGPT is a new shiny toy they are, relatively speaking, just learning what they can do with. 对于这个已经存在十年、由试图垄断社交媒体广告收入的人们组成的互联网络来说，ChatGPT是一个新的闪亮玩具，相对而言，他们才刚刚开始学习如何利用它。

The biggest uses at the moment are generating AI swap like Shrimp Jesus and Bread Horse pictures, bots that write inanely worried Tweets and Facebook posts, and some faceless YouTube channels, but that's not going to last long. 目前最大的用途是生成像“虾耶稣”和“面包马”图片那样的AI换脸内容，编写无聊担忧的推文和脸书帖子，以及一些无面YouTube频道，但这不会持续太久。

And the better they get at using AIs like ChatGPT and Sora, and the cheaper and faster those AIs get in general, the easier it will be for them to split us into more groups and monopolize more of our attention to collect more of the ad revenue. 他们越擅长使用ChatGPT和Sora等人工智能，以及这些人工智能总体上变得越便宜、越快，他们就越容易将我们分裂成更多群体，垄断我们更多的注意力，从而收取更多的广告收入。

And AIs are powerful tools for manipulating humans. 人工智能是操纵人类的强大工具。

We've already seen reports of people harming themselves after being encouraged by chatbots to do so. 我们已经看到有报道称，人们在聊天机器人的鼓励下自残。

I'm telling you, you only think we're polarized now. 我告诉你，你只是以为我们现在才两极分化。

Don't forget, generative AI video is still in its infancy, but it's getting better almost by the week. 别忘了，生成式AI视频仍处于起步阶段，但它几乎每周都在进步。

For short, heavily compressed video clips like you see on social media, soon it will become almost impossible to tell real footage from a good fake, and trust me, you haven't seen anything yet. 对于你在社交媒体上看到的短小、高度压缩的视频片段，很快就几乎不可能分辨出真实镜头和制作精良的假视频了，相信我，你还没见过什么呢。

Just wait until it can create virtual people that can be animated fast enough to keep up with generative AI output in real time with synchronized lip movements. 等着看吧，直到它能创造出虚拟人物，这些人物可以被快速动画化，以实时跟上生成式AI的输出，并同步唇部动作。

And, as these new AI data centers get built, the bad guys will no longer have to manipulate us into small bubble groups. 而且，随着这些新的人工智能数据中心建成，那些“坏人”将不再需要将我们操纵成小的信息茧房（Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境）群体。

But every bad actor will be able to spin up their own dedicated AI agent process specifically to monitor, manipulate, and target you, specifically, and each and every one of us individually. 相反，每个不良行为者都将能够启动他们自己的专用AI代理程序，专门用于监控、操纵并精准针对你，以及我们每一个人。

And that is going to be way worse. 那将会糟糕得多。

And all the talk of AI safety is just a smokescreen fairy tale distraction. 所有关于AI安全的讨论都只是一个烟雾弹般的童话式转移注意力。

Remember, the industry telling us that they're working on AI safety is the same industry that already built this machine and turned it loose on us so they can make more money. 请记住，告诉我们他们正在致力于AI安全的行业，正是那个已经建造了这台机器并将其释放给我们，以便赚取更多金钱的行业。

Do you think for a second that they would let any hypothetical safety work stand between them and doing whatever they can to make even more money at our expense? 你认为他们会允许任何假设性的安全工作阻碍他们，让他们无法以牺牲我们为代价去赚取更多的钱吗？

See this book: Empire of AI if you wanted more information on how likely the AI industry is to slow down or stop on their own. 如果你想了解更多关于AI行业自行放慢或停止的可能性，请阅读这本书：《AI帝国》。

Now there's a ton of stuff I've had to skip over in this video for length and time, like any discussion of the tons of discoveries in psychology and psychological manipulation that were started in the Mad Men era and accelerated as the number of people that could be experimented on simultaneously grew to Internet scale. 由于视频长度和时间限制，我不得不跳过很多内容，比如对心理学和心理操纵领域大量发现的讨论，这些发现始于“广告狂人”时代，并随着可同时进行实验的人数达到互联网规模而加速发展。

How certain educational organizations have been built their content on foundations built by these bubbles, like the ones that make hours-long algorithmically generated videos into preschool children. 某些教育机构如何将他们的内容建立在这些信息茧房（Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境）所构建的基础之上，比如那些为学龄前儿童制作长达数小时的算法生成视频的机构。

The data brokers that package up all this data and sell it to content creators and advertisers and tool makers. 那些将所有这些数据打包并出售给内容创作者、广告商和工具制造商的数据经纪人（Data Brokers: 收集、整理和出售个人信息的公司）。

And I've had to omit a whole bunch of supporting evidence and research citations. 我还不得不省略了大量的支持证据和研究引用。

This video, in fact, is basically the "Too Long; Didn't Read" (TL;DR: Too Long; Didn't Read: 意为“太长；没读”，通常用于概括冗长内容的简短摘要) version, or what my generation used to call the "Cliff Notes" version of a much more detailed book about this stuff. 事实上，这个视频基本上是关于这些内容的“太长不看”（TL;DR: Too Long; Didn't Read: 意为“太长；没读”，通常用于概括冗长内容的简短摘要）版本，或者我这一代人过去常说的“克里夫笔记”版本，而这些内容在一本更详细的书中有所阐述。

Now I can't give you a link to that book, and I'm not surprised you haven't read it because I haven't finished writing it yet. 现在我无法给你那本书的链接，你没读过我也不惊讶，因为我还没写完。

But I'm making this video now anyway, because the further I get into this, and the more research I do, and the more I see how this is affecting the world that I and my family are living in, the less I'm willing to wait until I'm done writing to go public. 但我现在还是制作了这个视频，因为我越深入研究，越看到它如何影响我和我的家人所生活的世界，我就越不愿意等到写完才公之于众。

And if the feedback from this video lights a fire under me to finish and/or gives me information that I didn't have, relevant sources that I haven't found, or things I need to clarify, then so much the better. 如果这个视频的反馈能激励我完成写作，或者给我提供我以前没有的信息、我尚未发现的相关来源，或者我需要澄清的事情，那就更好了。

团结反抗：跨越信息茧房的希望

It's time for us to start forming that scrappy group of rebels the sci-fi shows always have fighting back against The Machine. 是时候让我们开始组建科幻剧中总会有的那支顽强的反抗者队伍，去对抗机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）了。

But it has to be done in such a way that the different anti-machine groups don't end up fighting each other. 但这必须以一种方式进行，即不同的反机器团体最终不会互相争斗。

And that's the hard part, because The Machine's greatest defense is having made us almost incapable of working together against it. 而这正是困难所在，因为机器最强大的防御就是让我们几乎无法团结起来对抗它。

It's like the biblical story of The Tower of Babel, if the God in the story was an evil machine and it gave everyone different worldviews instead of different languages. 这就像《圣经》中巴别塔的故事，如果故事中的上帝是一台邪恶的机器，它赋予每个人不同的世界观而不是不同的语言。

There's a huge amount of work we're all going to have to do to get out of this nightmare, but from what I can tell, the only place that it can start is by talking across the bubbles. 我们所有人都要做大量的工作才能摆脱这场噩梦，但据我所知，唯一的起点就是跨越信息茧房（Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境）进行对话。

My only hope is that maybe enough of us can hate The Machine more than we hate each other. 我唯一的希望是，也许我们足够多的人能够憎恨机器胜过憎恨彼此。

So I don't think there's any way that the people that are sure that the 2020 election was stolen by Joe Biden but that the 2024 election was free and fair, are going to join hands and sing "Kumbaya" in a big circle with the people that are sure that the 2020 election was fair but that the 2024 election was stolen by Donald Trump. 所以，我不认为那些坚信2020年大选被乔·拜登窃取但2024年大选自由公正的人，会与那些坚信2020年大选公正但2024年大选被唐纳德·特朗普窃取的人手拉手围成一圈唱《Kumbaya》。

But I think—I hope—that there can be some common ground. 但我认为——我希望——能有一些共同点。

I hope that a lot of us from across a large number of different information bubbles can join together, not in admitting our own incorrect beliefs, and not in apologizing to those on the other side of the aisle—that's way too much to ask right now—but rather to join together in hating the common enemy of the ShrimpJesus Slop-posting conspiracy-creating scumbags of the world, and the people that turn every new story into an "us versus them" loyalty test, and the tool writers who make the software that automates more content bots and polarizes us so they can extract more ad money. 我希望我们许多来自不同信息茧房（Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境）的人能够团结起来，不是为了承认我们自己的错误信念，也不是为了向对立面的人道歉——现在要求这些太多了——而是为了共同憎恨那些制造“虾耶稣”垃圾帖、编造阴谋论的社会败类，那些把每一个新故事都变成“我们对抗他们”的忠诚考验的人，以及那些编写软件自动化更多内容机器人并使我们两极分化以榨取更多广告费的工具开发者，这些共同的敌人。

So here's your action item: engage with those people in your life that you can't agree with, and respond to their garbage memes and hateful rhetoric, but do not contradict them, and do not argue with their core beliefs. 所以，这是你的行动项目：与你生活中那些你无法达成一致的人交流，回应他们的垃圾表情包和充满仇恨的言论，但不要反驳他们，也不要争论他们的核心信念。

Ask them questions about their priorities, and try to be sincere. 问他们关于他们优先事项的问题，并尽量真诚。

Ask questions like: "With all that's going on, is what you're talking about here really the most important thing to you right now?" "Is this really what's directly affecting your life the most?" "If not, where did you get the idea that this is the thing you should be talking about right now?" "Why do you think they wanted you to think about this?" "Do you think they really have your best interests at heart?" 问这样的问题：“在所有正在发生的事情中，你现在谈论的这些真的是对你来说最重要的吗？”“这真的是对你生活影响最大的吗？”“如果不是，你从哪里得到这个想法，认为这是你现在应该谈论的事情？”“你认为他们为什么想让你思考这个问题？”“你认为他们真的把你的最大利益放在心上吗？”

Just try to get them thinking to start with. Try to plant the seed of doubt about the motives behind the information that they're being fed. 首先，试着让他们开始思考。试着在他们所接收到的信息背后动机上播下怀疑的种子。

It's just a start, but at least it's something. 这只是一个开始，但至少有所行动。

Hopefully redirecting them to thinking about who is feeding them this information will help. 希望引导他们思考是谁向他们提供这些信息会有所帮助。

I can tell you from personal experience with some of my extended family, that if you can change the subject from their conspiracy theories and group-think talking points to "Who was that told you this was priority one and what do you think they're trying to keep you from thinking about instead?" then you can at least have much less confrontational conversations. 我可以根据我与一些大家庭成员的个人经验告诉你，如果你能把话题从他们的阴谋论和群体思维的论点，转变为“是谁告诉你这是首要任务，你认为他们又试图阻止你思考什么？”那么你至少可以进行不那么对抗性的对话。

It's so much more pleasant than trying to argue with them or just biting your tongue and trying to say nothing. 这比试图与他们争论，或者只是忍着不说任何话要愉快得多。

I know a lot of you will find it unpleasant, but here's the reality: as long as we're fighting amongst ourselves, we can't fight The Machine effectively. 我知道很多人会觉得这不愉快，但现实是：只要我们还在内斗，我们就无法有效地对抗机器（The Machine: 指代一个由互联技术和扭曲激励组成的系统）。

Now, is that going to work? Can we agree that there's a greater evil and join forces across bubbles to fight it? 那么，这会奏效吗？我们能否同意存在一个更大的邪恶，并跨越信息茧房（Information Bubble: 指个人在网络上接触到的信息被过滤和限制，导致他们只能看到与自己观点相符的内容，从而形成封闭的信息环境）联合起来对抗它？

I don't have any idea, but from where I'm sitting, if we can't agree that the people manipulating all of us are bad, then I can't imagine anything we could agree on. 我不知道，但依我看来，如果我们连操纵我们所有人的那些人是坏的都无法达成一致，那么我无法想象我们还能在什么事情上达成一致。

So maybe that's a start. Maybe this might work. 所以，也许这是一个开始。也许这会奏效。

And maybe we can find shared cause and start digging ourselves out of this misinformation nightmare. 也许我们可以找到共同的事业，开始摆脱这场虚假信息的噩梦。

And maybe we can do that before AI gets fully weaponized against us and becomes that much harder. 也许我们可以在人工智能完全被武器化对抗我们，并使情况变得更加困难之前做到这一点。

Or maybe, just maybe, we're all just f------------ 或者，也许，我们都将万劫不复。