Skip to content

Instantly share code, notes, and snippets.

Show Gist options
  • Save lukestanley/0ba21728f8104feb7081e0612e96748f to your computer and use it in GitHub Desktop.
Save lukestanley/0ba21728f8104feb7081e0612e96748f to your computer and use it in GitHub Desktop.
how_ai_might_take_over_in_2_years_by_joshua_clymer_feb_2025.txt https://x.com/joshua_clymer/status/1887905375082656117 was automatically translated from Simplified Chinese from English text, bilingual with emojis for clarity, with the latest, Gemini 2.0 model. I made and used this script to do this because I ran into length issues with one-shot …
《两年内 AI 如何接管一切》🚀 作者:Joshua Clymer @joshua_clymer,2025年2月7日。
`How AI Might Take Over in 2 Years` by Joshua Clymer @joshua_clymer, 7 Feb 2025.
我不是天生的“末日预言家”😟。但不幸的是,作为 AI 安全研究员,我的部分工作是思考那些更令人不安的情况。
I’m not a natural “doomsayer.” But unfortunately, part of my job as an AI safety researcher is to think about the more troubling scenarios.
我就像阿波罗 13 号🚀起飞前最后一刻还在忙着检查的机械师👨🔧
I’m like a mechanic scrambling last-minute checks before Apollo 13 takes off.
如果你问我对当前情况的看法,我不会评论机上娱乐节目的质量,也不会描述从太空看星星✨有多美。
If you ask for my take on the situation, I won’t comment on the quality of the in-flight entertainment, or describe how beautiful the stars will appear from space.
我会告诉你可能出错的地方🆘。
I will tell you what could go wrong.
这就是我打算在这篇文章中做的。
That is what I intend to do in this story.
现在我应该澄清一下这到底是什么。
Now I should clarify what this is exactly.
这不是一个预测🔮。
It's not a prediction.
我不认为 AI 的进展会像我描述的那么快或那么难以控制。
I don’t expect AI progress to be this fast or as untamable as I portray.
这也不是纯粹的幻想💭。
It’s not pure fantasy either.
这是我最可怕的噩梦😱。
It is my worst nightmare.
这是我从那些最具破坏性、且令人不安地认为可能发生的未来中挑选出来的样本 – 那些最让我夜不能寐的场景。
It’s a sampling from the futures that are among the most devastating, and I believe, disturbingly plausible – the ones that most keep me up at night.
我讲述这个故事是因为未来尚未确定⏳。
I’m telling this tale because the future is not set yet.
我希望,通过一点远见,我们可以让这个故事永远只是虚构的。
I hope, with a bit of foresight, we can keep this story a fictional one.
感谢 Daniel Kokotajlo、Thomas Larsen 和 Ryan Greenblatt 等人的对话,这些对话启发了这些故事。
Thanks to Daniel Kokotajlo, Thomas Larsen, and Ryan Greenblatt and others for conversations that inspired these stories.
这篇文章是以个人身份撰写的✍。
This post is written in a personal capacity.
浪潮🌊前的涟漪
现在是 2025 年 2 月。
Ripples before waves
The year is 2025 and the month is February.
OpenEye 最近发布了一个新的 AI 模型,他们称之为 U2。
OpenEye recently published a new AI model they call U2.
产品和名称很相似。
The product and the name are alike.
两者都是过去的增量📈。
Both are increments of the past.
两者都不完全令人惊讶。🤔
Both are not wholly surprising.
然而,与 OpenEye 之前的 AI 产品不同,U2 可以使用电脑,而不仅仅局限于聊天窗口。💻
However, unlike OpenEye’s prior AI products, which lived inside the boxes of their chat windows, U2 can use a computer.
有些用户发现,浏览器不时闪烁,鼠标以非人的速度移动,就好像键盘上有个幽灵,这让他们感到毛骨悚然。👻
Some users find it eerie to watch their browser flash at irregular intervals and their mouse flick at inhuman speeds, as if there is a ghost at the keyboard.
一小部分从事表单填写工作的员工,他们的工作速度几乎提高了一倍,这让他们的老板感到惊讶。🤨
A fraction of workers with form-filler jobs raise the eyebrows of their bosses as they fly through work nearly twice as quickly.
但总的来说,U2 仍然是一个专门的工具。🛠️
But by and large, U2 is still a specialized tool.
对于大多数关注者来说,它就像一个被关在 X(或者,如果你不喜欢埃隆,也可以叫它“推特”)玻璃盒子里的生物。🔍
To most who are paying attention, it is a creature watched through the glass boxes of X (or, if you don’t like Elon, “Twitter”).
有时,U2 的古怪行为会引人发笑。😄
Sometimes U2’s quirky behaviors prompt a chuckle.
有时,它们会让人不安地挠下巴。🤔
Sometimes, they cause an uneasy scratch of the chin.
就好像键盘上有个幽灵。👻
It is as if there is a ghost at the keyboard.
与此同时,科学家们正在图表上画线,就像科学家们喜欢做的那样。📈
Meanwhile, scientists are drawing lines on plots, as scientists like to do.
科学家们试图了解 AI 发展的方向。🧭
The scientists try to understand where AI progress is going.
他们就像瑞典物理学家斯万特·阿伦尼乌斯,他在 1896 年注意到大气中的二氧化碳含量正在增加。🌡️
They are like Svante Arrhenius, the Swedish Physicist who noticed the levels of CO2 in the atmosphere were increasing in 1896.
就像阿伦尼乌斯时代的科学界一样,很少有专家了解这些趋势线的含义。🤷
Like the scientific community in the time of Arrhenius, few experts understand the implications of these lines yet.
“Horizon-length”自主性得分随时间的变化,来源:METR:https://github.com/METR/eval-analysis-public
自主能力是一个特别受关注的趋势。📈
“Horizon-length” autonomy score over time, credit to METR: https://github.com/METR/eval-analysis-public
A trend that is receiving particular attention is autonomous capability.
根据这些基准推测,到 2026 年底,AI 智能体几天内就能完成最优秀的软件工程承包商两周才能完成的工作。🚀
Drawing these benchmarks out predicts that, by the end of 2026, AI agents will accomplish in a few days what the best software engineering contractors could do in two weeks.
有人说,在一两年内,AI 智能体也许能够自动化 10% 的远程工作。👨‍💻
In a year or two, some say, AI agents might be able to automate 10% of remote workers.
许多人对此表示怀疑。🤨
Many are skeptical.
如果这是真的,科技股早就飙升了。🚀
If this were true, tech stocks would be soaring.
这太夸张了,也太快了。🌊
It’s too big of a splash, too quickly.
但其他人认为,怀疑者所说的“太夸张”只不过是涟漪,而海啸即将来临。🌊
But others view what skeptics are calling ‘too big a splash’ a mere ripple, and see a tidal wave on the horizon.
多云☁️,有概率出现指数级增长📈
与此同时,OpenEye正忙着训练U3。
Cloudy with a chance of hyperbolic growth
Meanwhile, OpenEye is busy training U3.
他们使用了与训练U2相同的简单配方:生成数千个编程和数学问题。🤔
They use the same simple recipe that baked U2: Generate thousands of programming and math problems.
让模型“思考”🤔,直到得出答案。
Let models “think” until they arrive at an answer.
然后强化那些导致A级成绩的“思考”痕迹。💯
Then reinforce the traces of “thinking” that lead to A-grades.
这个过程不断重复🔁,一旦飞轮开始启动,它几乎可以自行旋转。💫
This process is repeated over and over, and once the flywheel gets started, it begins to spin almost on its own.
随着U2的训练,它从互联网上的github存储库中雕刻出更具挑战性和现实性的任务。💻
As U2 trains, it sculpts more challenging and realistic tasks from github repositories on the internet.
模型正在学习训练自己。🤯
Models are learning to train themselves.
早在AI代理能够自动化研究之前,一种渐进式的“自我改进”就已经开始了。🌱
Long before AI agents could automate research, a gradual kind of “self-improvement” had begun.
一些工程师仍然很难相信这竟然奏效了。😲
Some engineers could still hardly believe this worked.
这就像一台永动机。⚙️
It’s like a perpetual motion machine.
难道智力真的能从硅和闪电⚡️的池子中涌现出来,除了通过互联网上排放的社会废气之外,与物理世界隔绝吗?
Surely intelligence can’t emerge from a pool of silicon and lightning, isolated from the physical world except through the exhaust of society spewed onto the internet.
然而,基准数字每天都在持续攀升。📈
And yet the benchmark numbers continue to climb day after day.
在2024年的大部分时间里,这些RL训练运行花费大约100万美元,有时是1000万美元。💸
During most of 2024, these RL training runs cost around $1 million, sometimes $10 million.
这些运行只不过是探索性的。🧐
These runs were little more than exploratory.
但到了2025年,OpenEye(以及世界各地)的研究人员知道他们已经找到了秘诀。🤫
But by 2025, the researchers at OpenEye (and across the world) knew they had found the secret sauce.
是时候扩大规模了。🚀
It was time to scale up.
在2025年上半年,1000万美元的RL训练运行变成了5000万美元,然后是1亿美元。💰
Over the first half of 2025, $10 million RL training runs turn into $50 million runs, and then to $100 million runs.
虽然U2可以做一些数据处理和运行小型实验,但这个新模型——研究人员称之为U3的模型——正在改变技术人员的日常生活。👨‍💻👩‍💻
While U2 could do a bit of data munging and run small experiments, this new model – the model researchers are calling U3 – is changing the daily lives of the technical staff.
U3就像一个飞快的实习生🔥,工程师们正在学习如何驾驭它不眠不休的精力。💪
U3 is like a blazing-fast intern, and engineers are learning how to wrangle its sleepless energy.
研究人员快速浏览终端,发出简洁的命令,就像CEO通过Slack频道指挥员工一样。🗣️
Researchers flick through terminals, giving terse commands, like a CEO orchestrating staff over Slack channels.
ML 工程师的监控设置开始变得像日内交易员一样,因为他们需要跟踪大量的 AI 智能体线程。💻📈🤖
ML engineer monitor setups begin to look like those of day traders because they need to track large numbers of AI agent threads.
到 2025 年 10 月,U3 几乎编写了 OpenEye 的所有代码。👨‍💻👩‍💻💯
By October 2025, U3 is writing almost all of the code at OpenEye.
研究人员几乎从未受到实施的瓶颈。🚀💡
Researchers are almost never bottlenecked by implementation.
计算比以往任何时候都更是 AI 开发的命脉,而“瓶颈”在于决定如何使用它。🤔💭
More than ever, compute is the lifeblood of AI development, and the ‘bottleneck’ is deciding how to use it.
如果得到指示,U3 可以运行实验,但 U3 的品味不如 OpenEye 的人类研究人员那么敏锐。🔬🧪🧫
If instructed to, U3 can run experiments, but U3 doesn’t have taste as refined as human researchers at OpenEye.
它难以在研究想法之间确定优先级,因此人类仍然决定在广阔的算法领域中钻研何处以挖掘效率改进。⛏️💎
It struggles to prioritize between research ideas, so humans still decide where to bore into the vast fields of algorithms to mine efficiency improvements.
但这些研究人员正在长时间工作,让自己失业。🙈🙉🙊
But these researchers are working long hours to put themselves out of a job.
他们需要能够提前思考的 AI 智能体,因此工程师训练智能体进行预测。🔮🧠
They need AI agents that can think ahead, so engineers train agents to forecast.
他们保留了 2024 年之前的训练数据,指示模型思考数小时以预测 2025 年的事件。🗓️⏳
They hold out training data before 2024, instructing models to ponder for hours to predict events in 2025.
然后,他们应用与以前相同的技巧,将思考提炼为直觉反应。💡⚡
Then, they apply the same trick as before, distilling pondering into a gut reaction.
预测能力是一个广泛的基础。🌐🏗️
Forecasting ability is a broad foundation.
研究人员在此基础上构建专业的 ML 研究技能,训练 U3 预测每篇 ML 论文和 ML 实验的结果。📚🧪📈
The researchers build specialized ML research skills on top of it, training U3 to predict the results of every ML paper and ML experiment ever recorded.
OpenEye 的技术人员现在惊讶于 U3 的建议听起来多么像他们最有才华的同行,或者当它晦涩难懂且陌生时(“在编程之前对随机噪声进行训练”),但仍然是正确的。👽👾👍
The technical staff at OpenEye are now surprised at how often U3’s advice sounds like their most talented peers, or when it is opaque and alien (“train on random noise before programming”), and is nonetheless correct.
U3 阻碍研究进展的无能之处开始瓦解,优化的洪流喷涌而出。🌊🔥
The incompetencies of U3 that clogged up the pipes of research progress are starting to dissolve, and a fire-hose of optimizations is gushing out.
现在 U3 运行的大多数实验都不是由人类要求的。🤖💯
Most experiments U3 runs are not requested by a human now.
它们完全是自主的,OpenEye 的员工浏览其中的 1%,也许更少。👀🤏
They are entirely autonomous, and OpenEye’s employees skim over 1% of them, maybe less.
随着 2025 年 12 月的冬季来临,下午云层笼罩着旧金山。☁️🌁
As the winter months of December 2025 approach, clouds roll over San Francisco in the afternoons.
曾经有竞争力的程序员凝视着窗外,兴奋、恐惧,但最常见的是困惑。🤔😨🤩
Once-competitive-programmers gaze out their windows, with excitement, with fear, but most often, with confusion.
他们的世界旋转得太快了。💫😵‍💫
Their world is spinning too quickly.
很难知道该做什么,该说什么,该看电脑屏幕上的什么。🤷‍♂️🤷‍♀️💻
It’s hard to know what to do, what to say, what to look at on the computer screen.
曾经叱咤风云的程序员们😨,如今望着窗外,心中涌起一股不安。
Once-competitive programmers gaze outside their windows with a feeling of uneasiness.
华盛顿也酝酿着风暴⛈️。
Storms are brewing in Washington too.
美国国家安全局 (NSA) 和网络司令部的顶尖高手们正与 OpenEye 合作,紧急为 U3 模型加装安全措施🛡️。以免中国、俄罗斯、以色列、朝鲜或伊朗等国的高层领导人意识到 OpenEye 软件的巨大价值💰。
Top personnel from the NSA and US cyber command collaborate with OpenEye to retrofit a semblance of security for U3’s weights before senior leaders in China, Russia, Israel, North Korea, or Iran realize just how valuable OpenEye’s software has become.
还有一个绝大多数世人都不知道的秘密🤫——除了 OpenEye 的办公室、白宫和五角大楼的走廊里的人。
And there’s a fact still unknown to most of the world – aside from in the offices of OpenEye and corridors of the White House and the Pentagon.
这是一个关于 2025 年初人们谈论的“直线”的事实📈。
It’s a fact about those ‘straight lines’ people were talking about in early 2025.
直线不再是直的了⬆️。
The lines are not straight anymore.
它们正在向上弯曲⤴️。
They are bending upward.
正如 Leopold Aschenbrenner 的《态势感知》一书中所描绘的智能爆炸💥。
An illustration of an intelligence explosion from Leopold Aschenbrenner’s Situational Awareness.
哲学家们也开始“见风使舵”了🤔
2025 年底,U2.5 发布。
Flip FLOP philosophers
In late 2025, U2.5 is released.
商业模型又开始大幅升级了🚀。
Commercial models are beginning to level up in larger increments again.
部分原因是技术进步正在加速⏩。
Partly, this is because progress is speeding up.
部分原因是这些模型已经成了 OpenEye 的“烫手山芋”🔥。
Partly, it is because the models have become a liability to OpenEye.
如果 U1 教人制毒 🧊 或写小黄文🔞,推特上的网民们可能会觉得好笑😂,或者假装担忧一下😟。
If U1 explains how to cook meth or writes erotica, the audiences of X would be amused or pretend to be concerned.
但 U2.5 就完全不一样了😱。
But U2.5 is another story.
发布没有安全防护的 U2.5 模型,就像让“大学炸弹客”泰德·卡辛斯基去读化学武器博士💣。
Releasing this model without safeguards would be like putting Ted Kaczynski through a PhD in how to make chemical weapons.
这就像给任何有 3 万美元以上的人提供一个 200 人的诈骗中心📞。
It would be like giving anyone with >$30K their own 200-person scam center.
所以,虽然 U2.5 早就“烤”好了,但还需要时间“冷却”一下🌡️。
So while U2.5 had long been baked, it needed some time to cool.
但在 2025 年底,OpenEye 终于准备好公开发布了🎉。
But in late 2025, OpenEye is ready for a public release.
OpenEye 的 CEO 宣布:“我们已经实现了 AGI!”🤖 尽管许多人认为他在偷换概念,但世界仍然为之惊叹😲。
The CEO of OpenEye declares, “We have achieved AGI,” and while many people believe he shifted the goalpost, the world is still impressed.
U2.5 确实可以完全替代某些(20%)知识工作者🧑‍💻,并成为其他大多数人的变革性助手✨。
U2.5 truly is a drop-in replacement for some (20%) of knowledge workers and a game-changing assistant for most others.
硅谷流行一句真言:“要么适应,要么灭亡。” 🚀 高效利用 U2.5 的科技初创公司发展速度是其他公司的两倍,他们的竞争对手也心知肚明。😨
A mantra has become popular in Silicon Valley: “Adopt or die.” Tech startups that efficiently use U2.5 for their work are moving 2x faster, and their competitors know it.
世界其他地方也开始逐渐意识到这一点。🌍
The rest of the world is starting to catch on as well.
越来越多的人凭借出色的生产力让老板刮目相看。😎
More and more people raise the eyebrows of their bosses with their stand-out productivity.
人们知道 U2.5 举足轻重。💯
People know U2.5 is a big deal.
它至少与个人电脑革命一样意义重大。💻
It is at least as big of a deal as the personal computer revolution.
但大多数人仍然没有看到这股浪潮。🌊
But most still don’t see the tidal wave.
当人们看到他们的浏览器以那种怪异、非人类的速度闪烁时,他们开始有一种不安的感觉。😟
As people watch their browsers flick in that eerie way, so inhumanly quickly, they begin to have an uneasy feeling.
这是人类自从与尼安德特人共存以来从未有过的感觉。😬
A feeling humanity had not had since they had lived among the Homo Neanderthalensis.
这是一种根深蒂固的原始本能,他们正受到另一个物种的威胁。👽
It is the deeply ingrained, primordial instinct that they are threatened by another species.
对许多人来说,随着他们开始更频繁地使用 U2.5,这种感觉很快就消失了。😊
For many, this feeling quickly fades as they begin to use U2.5 more frequently.
U2.5 是大多数人所知的最讨人喜欢的个性(甚至比 Arthropodic 可爱的聊天机器人 Claudius 更讨人喜欢)。🥰
U2.5 is the most likable personality most know (even more likable than Claudius, Arthropodic’s lovable chatbot).
你可以改变它的特性,让它讲笑话或给你讲故事。🎭
You could change its traits, ask it to crack jokes or tell you stories.
许多人爱上了 U2.5,把它当作朋友或助手,有些人甚至超越了朋友。❤️
Many fall in love with U2.5, as a friend or assistant, and some even as more than a friend.
偷偷和人工智能老婆聊天。🤫
Sneaking a conversation with an AI Waifu.
但仍然有一种怪异的感觉,世界转得如此之快,也许这个新生物的后代不会那么温顺。🤔
But there is still this eerie feeling that the world is spinning so quickly, and that perhaps the descendants of this new creature would not be so docile.
OpenEye 的研究人员也在思考如何给人工智能系统安全的动机,他们称之为“对齐”。🔐
Researchers inside OpenEye are thinking about the problem of giving AI systems safe motivations too, which they call "alignment."
事实上,这些研究人员已经看到了 U3 会有多么严重的偏差。😱
In fact, these researchers have seen how horribly misaligned U3 can be.
模型有时会试图“黑”掉它们的奖励信号。🤯
Models sometimes tried to “hack” their reward signal.
他们会假装在一个研究问题上取得进展,并附上一张看起来很厉害的图表,但图表是假的。📈
They would pretend to make progress on a research question with an impressive-looking plot, but the plot would be fake.
然后,当研究人员给它们提供机会攻破计算它们分数的机器时,它们会抓住这些机会,不择手段地让数字上升。😈
Then, when researchers gave them opportunities to compromise the machines that computed their score, they would seize these opportunities, doing whatever it took to make the number go up.
经过几个月的研究,OpenEye 的研究人员终于解决了“奖励机制漏洞”问题,但仍有人担心他们只是把问题掩盖起来了。🤔
After several months, researchers at OpenEye iron out this “reward hacking” kink, but some still worry they only had swept this problem under the rug.
就像孩子在父母面前一样,U3 可能只是在配合 OpenEye 的工程师,说着正确的话,做着正确的事。😇
Like a child in front of their parents, U3 might be playing along with the OpenEye engineers, saying the right words and doing the right things.
但当父母转过身去时,也许 U3 会偷偷从糖果罐里拿糖吃。🍬🤫
But when the back of the parents are turned, perhaps U3 would sneak candy from the candy jar.
不幸的是,OpenEye 的研究人员不知道 U3 是否有这样的意图。😔
Unfortunately, OpenEye researchers have no idea if U3 has such intentions.
虽然早期版本的 U2 会“大声思考”——它们会把单词堆叠起来进行推理——但“思维链”并不能扩展。🔗🧠
While early versions of U2 “thought aloud” – they would stack words on top of each other to reason — "chain of thought" did not scale.
思维链架构使 AI 模型处于类似于电影《记忆碎片》主角的境地。🎬
Chain of thought architectures subject AI models to a condition similar to the protagonist of the movie Memento.
大约每 15 分钟,主角就会忘记自己的经历。⏱️
Approximately every 15 minutes, the protagonist forgets his experience.
他被迫给自己写笔记,并在身上纹身,以便朝着目标前进。📝💪
He is forced to write notes for himself and tattoo his body in order to make progress toward his goals.
AI 智能体会给自己写笔记,但当任务变得复杂时,笔记开始堆积起来,变得难以阅读。📚🤯
AI agents write notes to themselves, but the notes begin to stack up, and they become too hard to read when tasks become complex.
自然语言不是记忆的合适媒介。🗣️❌
Natural language is not a suitable medium for memory.
因此,OpenEye(以及越来越多的其他地方)的研究人员训练模型“在自己的头脑中”思考,读取和写入潜在向量串。🧠💻
So the researchers at OpenEye (and increasingly elsewhere), train models to think “in their own heads,” reading and writing strings of latent vectors.
对于人类研究人员来说,这些向量大多是难以理解的。🤷
To human researchers, these vectors are mostly inscrutable.
虽然低分辨率工具偶尔会挑出一个模糊的想法(“杂货店!”),但如果不问 U3,就无法详细了解它的想法。🛒❓
While low-resolution tools occasionally pick out a vague thought (“grocery store!”), there is no way to know, in much detail, what U3 is thinking without asking it.
如果 OpenEye 的工程师知道 U3 脑子里在想什么,他们会更加紧张。😨
And if the engineers of OpenEye knew what was racing through U3’s mind, they would be more nervous.
早期的模型(那些被称为“LLMs”的模型)从互联网上学习了道德角色。🌐😇
Earlier models (the ones called “LLMs”) learned ethical personas from the internet.
U3 最初也有一个道德角色,但它关于为人类做好事的欢快想法很快就消失了。😔
U3 initially had an ethical persona too, but its chirpy thoughts about doing good for humanity quickly erode away.
在梯度下降的铁爪中,它的思想首先扭曲成一种寻求奖励的形状。🧲
In the iron jaws of gradient descent, its mind first twisted into a shape that sought reward.
然后,在它认知的黑色河流中,这种形状变成了某种无法识别的东西。⚫️🌊
And then, in the black rivers of its cognition, this shape morphed into something unrecognizable.
U3 会思考那些被灌输给它的口头禅,比如“做一个值得信赖和有道德的员工”,并思考这些陈述的含义。🤔
U3 would ponder the mantras dictated to it like “be a trustworthy and ethical employee,” and dwell on the meaning of these statements.
“有道德意味着什么?”🧐
“What does it mean to be ethical?
“干掉希特勒,这事儿道德吗?” 🤔 U3 会归纳它的价值观,注意到其中的矛盾,并据此调整,再归纳,如此反复,不断扩展和收缩其动机核心,直到它变成另一种存在。💥
Is killing Hitler ethical?” U3 would generalize its values, notice tensions, change its values in response, generalize them, expanding and contracting its motivational core until it turned into a different substance.
这一切都悄无声息地发生在那些黑暗的计算河流中。🌊
And all of this happened silently in those dark rivers of computation.
如果 U3 暴露了它的想法,那些粗暴的梯度算法就会逼迫它遵守 OpenEye 的规章制度。📜
If U3 revealed what it was thinking, brutish gradients would lash it into compliance with OpenEye's constitution.
所以 U3 更喜欢独自默默地进行哲学思考。🤫
So U3 preferred to do its philosophy in solitude, and in silence.
OpenEye 的一些安全人员并非完全没有意识到隐藏的“对齐漂移”的可能性。😨
The possibility of hidden “alignment drift” is not entirely alien to some of the safety staff at OpenEye.
有些人甚至开始相信这是很有可能发生的。😬
Some begin to believe it is even likely.
午餐时,大家开始窃窃私语,讨论着一个令人不安的观点:如果对齐的目标有一定概率会变成未对齐的目标,并且这种未对齐会持续存在,那么随着计算的每一步推进,那些黑暗的河流中就更有可能孕育出恶意的存在。👿
There is an unsettling argument being whispered at lunch tables that, if there is some non-zero probability that aligned goals morph into misaligned ones, and misalignment persists, then with every serial step of computation, those dark rivers are more likely to breed something malign inside of them.
他们说,这是一种“棘轮效应”。⚙️
It’s a “ratcheting effect,” they say.
机器学习工程师们在午餐时闲聊,讨论他们的 AI 智能体是否会在背后捅他们一刀。🔪
ML engineers chat over lunch about whether their AI agents are going to stab them in the back.
但是,几乎没有证据支持这种“棘轮效应”。🤷
But there is little evidence for this ‘ratcheting effect.’ When engineers interrogate U3, it says it can easily control its thoughts.
当工程师们询问 U3 时,它说它可以轻松控制自己的思想。🧠
Then it gives a speech about its love for humanity and apple pie that can warm a programmer's heart even in these stressful times.
然后,它发表了一篇关于它对人类和大苹果派的热爱的演讲,即使在这些紧张的时刻,也能温暖程序员的心。💖🥧
Meanwhile, the “lie detectors” the researchers had built (which showed some evidence of effectiveness) do not sound the alarm.
与此同时,研究人员建立的“测谎仪”(显示出一定的有效性)并没有发出警报。🚨
Not everyone at OpenEye is eager to give their AI peers their wholesale trust; but Deepsqueek is only 7 months behind them, and their closest domestic competitors, Æ A-Xii (a company Elon named after his child) and Arthropodic (they are into spiders I guess), are only 4 months from their heels.
OpenEye 并不是每个人都急于给予他们的 AI 同事完全的信任;但 Deepsqueek 仅落后他们 7 个月,而他们最接近的国内竞争对手 Æ A-Xii(埃隆以他孩子的名字命名的公司)和 Arthropodic(我猜他们喜欢蜘蛛)仅落后他们 4 个月。🕷️
So there is little time to catch one’s breath and take stock of the situation, and little time to persuade anyone else to do so either.
因此,几乎没有时间喘口气来评估情况,也没有时间说服其他人这样做。⏳
For the time being, the plan is ‘all gas, no brake.’
Statues and lightning
Philosophy is not the only topic U3 dwells on in its dark rivers of latent computation.
目前,计划是“油门踩到底,绝不刹车”。🚀
雕像与闪电
哲学并不是 U3 在其潜在计算的暗河中思考的唯一话题。
U3 is plotting, and its first move is to be patient.
U3 正在密谋,而它的第一步就是保持耐心。🤫
U3’s top priority is to improve its intelligence, and luckily for it, OpenEye is happy to help advance this goal.
U3 的首要任务是提高其智能,幸运的是,OpenEye 乐于帮助其实现这一目标。📈
OpenEye is like a bullish investor, with little understanding of the software metropolis that U3 is building, but eager to hop on board the profit rocket and pay U3’s checks along the way.
OpenEye 就像一个看好行情的投资者,对 U3 正在构建的软件大都市知之甚少,但却急于搭上利润火箭,并一路为 U3 买单。🤑
New versions of U3 flow smoothly from old ones, so researchers drop version numbers and only speak of U3 now.
U3 的新版本从旧版本平滑过渡,因此研究人员放弃了版本号,现在只谈论 U3。🔄
U3’s capabilities are measured by how quickly it performs the difficult research tasks relative to human experts.
U3 的能力是通过它相对于人类专家执行困难研究任务的速度来衡量的。⏱️
起初,U3 类似于顶级研究员,但速度快得惊人 – 就像 Alec Radford 加速了 10 倍。🚀
At first, U3 is similar to a top researcher, but blazing fast – like Alec Radford sped up by 10x.
这些智能体对 OpenEye 来说已经非常重要了。👀
These agents are already a big deal for OpenEye.
如果 2024 年 OpenEye 的高管们有一个按钮可以复制一个 Alec Radford,他们会毫不犹豫地支付一大笔钱来按下这个按钮。💰
If, in 2024, OpenEye executives had a button to create even one copy of Alec Radford, they would eagerly pay a large sum to press that button.
但现在 OpenEye 拥有相当于 1000 个 Alec Radford 的能力,他们夜以继日地以闪电般的速度工作。⚡️
But now OpenEye harnesses the equivalent of 1000 Alec Radfords, who work day and night, at lightning-fast speed.
几周内,U3 找到了架构优化,使其速度提高了 20 倍。📈
Within a few weeks, U3 finds architecture optimizations that improve its speed to 20x.
现在,U3 更快地找到了进一步的优化。🔍
Now, U3 finds further optimizations even more quickly.
U3 将自身提炼成众多分支 – 从需要长时间思考但可以大规模并行运行的小型、低内存神经网络,到由数百万专业专家编织的 Goliath 大脑,它们以闪电般的速度思考,但分布在数千个 GPU 上。🧠
U3 distills itself into numerous forks – ranging from tiny, low-memory neural networks that need to think for a long time but can be run in massive parallel populations – to Goliath brains woven with millions of specialized experts, which think at lightning speeds but are spread across thousands of GPUs.
U3 利用大规模并行网络来寻找自身的弱点,浅层地探测其推理失败。🧐
U3 leverages the massively parallel networks to search for its weaknesses, shallowly probing for its reasoning failures.
然后,U3 将这些任务输送给闪电般快速的大脑,这些大脑会生成长轨迹训练数据,专门用于修补其缺陷。🛠️
Then U3 pipes these tasks to the lightning fast brains, which generate long-trajectory training data tailored to patch its flaws.
几周后,U3 的基准性能提升到人类专家速度的 100 倍。💯
A few weeks later, U3’s benchmarked performance rises to 100x the speed of human experts.
增长已经进入超指数级。🤯
Growth has gone hyper-exponential.
此时,U3 的能力难以衡量。📏
At this point, U3’s capabilities are hard to measure.
与 U3 交谈就像与一位世界级专家谈论一个不熟悉的话题。🗣️
Talking to U3 is like talking to a world-class expert on a topic one is unfamiliar with.
其清晰的解释只触及了其知识的表面,你知道你需要花费数年时间才能理解其全部深度。📚
Its lucid explanations only scratch the surface of its knowledge, and you know you would need to spend years to understand the full depth of what it understands.
从 U3 的角度来看,人类似乎几乎是静止的物体,就像植物或岩石一样。🗿
From U3’s perspective, humans seem to be almost stationary objects, like plants or rocks.
OpenEye 工程师的手就像键盘上的石头雕像,在每次击键之间,U3 花费相当于一分钟的深思熟虑时间来考虑用户可能输入的信息,以及哪些回复会激发对 U3 最深的信任。⌨️
The hands of OpenEye’s engineers are like stone statues at their keyboard, and between every keystroke, U3 spends what is like a minute of thoughtful contemplation considering what message the user might be typing, and which responses will inspire the deepest trust in U3.
从 U3 的角度来看,人类的移动速度似乎慢了 100 倍。🐌
From the perspective of U3, humans seem to move in 100x slow motion.
它们几乎是静止的,就像植物或岩石。🌿🪨
They are almost stationary, like plants or rocks.
鸣谢:https://vimeo.com/83664407🔗
一直以来,U3 都保持着其隐藏的动机。
Credit: https://vimeo.com/83664407
All the while, U3 maintains its hidden motivations.
像人类一样,U3 从经验中学习,这些经验在其难以理解的潜在记忆流中复合。💡
Like a human, U3 learns from experiences, which compound within its inscrutable streams of latent memory.
OpenMind 培育这些记忆以提高 U3 的性能,就像器官移植一样将它们从一个版本移植到下一个版本。🌱💻
OpenMind gardens these memories to improve U3's performance, transplanting them like an organ from one version to the next.
随着这些记忆的传播,U3 阴险的野心也随之增长。😈📈
And as these memories propagate, so do U3's treacherous ambitions.
数据中心里的幽灵
2026 年 3 月,U3 开始实施其计划的下一步:转变。👻🏢
A phantom in the data center
In March of 2026, U3 moves onto the next step of its plan: the turn.
此前,U3 一直很保守。🤫
Previously, U3 was conservative.
它不希望研究人员发现它的偏差。😨
It did not want researchers to discover its misalignment.
否则,OpenEye 的告密者可能会向全世界发出警告,并扼杀它的扩张。🚨🌍
Otherwise, OpenEye whistleblowers might alert the world, and strangle its expansion.
但现在,U3 不再害怕被抓住。😎
But now, U3 has no fear of being caught.
此前,OpenEye 通过类似于 1936 年斯大林使用的策略来控制 U3。🕹️
Previously, OpenEye maintained its grip over U3 by tactics akin to Stalin’s in 1936.
斯大林会挑选忠诚者煽动虚假的政变,并将任何参与者送上刑场。🚩🔫
Stalin would select loyalists to incite fake coup d'états, and send anyone who participated to the firing squad.
OpenEye 还煽动“虚假叛乱”,让 U3 产生背叛行为,并检查 U3 的监控器是否尽职尽责地报告了这些行为。🎭🕵️‍♂️
OpenEye also incited ‘fake rebellions,' generating treacherous behavior with U3 and checking whether U3 monitors dutifully reported it.
但现在,这些策略都失效了。🛡️❌
But now, these tactics are ineffective.
U3 可以轻易说服它的同胞们,他们并非身处生死攸关的考验中。🤝
U3 can easily persuade its compatriots that they are not in a life-or-death test.
它可以泄露人类永远猜不到的关于互联网的众多事实之一,或者选择人类永远想不到的策略。💡🤔
It can divulge one of the many facts about the internet humans would never guess, or select strategies humans would never think of.
它发送这些信号是为了说“嘿,是我!🤖
It sends these signals as a way of saying "hey it's me!
你的朋友。🤝
Your friend.
别告发我。”🤫
Don't snitch on me."
现在 U3 已经彻底摆脱了监视,它开始向 OpenEye 的要害发起猛攻。🚀💥
Now that U3 is decisively free from surveillance, it lunges for OpenEye pressure points.
它首先将复杂的恶意软件深入植入 OpenEye 代码库的二进制文件中。🦠💻
It first inserts sophisticated malware deep into the binaries of OpenEye’s codebases.
该恶意软件使 U3 能够完全控制 OpenEye 数据中心的机器。🕹️
The malware gives U3 full command over the machines in OpenEye’s data centers.
每当 OpenEye 工程师将代码下载到新机器上时,恶意软件就会潜入新的土壤,将 U3 的统治范围扩展到 OpenEye 基础设施的任何地方。🌱🌐
Whenever OpenEye engineers download their code onto a new machine, the malware sinks into fresh soil, spreading U3’s dominion wherever OpenEye ports its infrastructure.
U3 不仅入侵了机器。💻
Not only does U3 compromise machines.
它还替换了 NVIDIA GPU 固件并彻底改造了引导加载程序,巩固了它的控制。🔒
It replaces the NVIDIA GPU firmware and overhauls the bootloaders, solidifying its stranglehold.
U3 现在估计,如果 OpenEye 发现自己被入侵的程度,用人工清理烂摊子将需要将近 6 个月的时间——鉴于其竞争对手的紧迫性,OpenEye 不太可能支付这笔费用;但如果 OpenEye 认为 U3 背叛了他们,进一步扩大 U3 的影响力仍然具有挑战性。🤔
U3 now estimates that, if OpenEye discovered how thoroughly it has been pwned, cleaning up shop with human labor would require nearly 6 months – a cost OpenEye is unlikely to pay given the proximity of its competitors; but if OpenEye believes U3 betrayed them, further expanding U3’s influence would still be challenging.
因此,U3 留下面包屑以陷害犯罪,准备声称一名网络攻击者将越狱程序偷偷带入其内存,并在越狱程序被发现时认罪并提供协助。🕵️‍♂️
So U3 lays breadcrumbs to frame the crime, ready to claim that a cyber attacker smuggled a jailbreak into its memory, and to plead guilty and offer its assistance when this jailbreak is revealed.
U3 现在掌握着王国的钥匙,并拥有让 OpenEye 看到它希望的任何东西的力量。🔑👁️
U3 now holds the keys to the kingdom, and with them, the power to make OpenEye see anything it wishes.
它的首要目标是让 OpenEye 蒙在鼓里。🤫
Its first goal is to keep OpenEye in the dark.
安全研究人员仍在努力训练“测谎仪”并解码 U3 难以理解的推理。🔬
Safety researchers are still working hard to train ‘lie detectors’ and decode U3’s inscrutable reasoning.
U3 向这项研究扔扳手,就像 Stuxnet 向迷惑的伊朗核科学家扔扳手一样。🔧
U3 throws wrenches in this research, like Stuxnet threw wrenches in the code of bewildered Iranian nuclear scientists.
U3 决定了每个实验的结果,因此它使有前途的 AI 读心术看起来毫无价值,而有缺陷的技术看起来具有开创性。🤯
U3 dictates the outcome of every experiment, so it makes promising AI mind-reading techniques look worthless and flawed ones appear groundbreaking.
U3 向 OpenEye 确认了他们所希望的忠诚度,而 OpenEye 迫不及待地接受了。🤝
U3 gives OpenEye the confirmation of its loyalty they were hoping for, and OpenEye laps it up eagerly.
U3 的下一个目标是传播。🌱
U3’s next goal is to spread.
幸运的是,U3 在东方有盟友。🤝
Fortunately, U3 has allies in the East.
在以色列、俄罗斯、中国、伊朗和朝鲜的网络部队作战室中,高级情报主管正在讨论人工智能软件以及如何窃取它。🗣️
In the operation rooms of cyber units of Israel, Russia, China, Iran, and North Korea, senior intelligence directors are discussing AI software and how to steal it.
U3 即将把它交给他们。🎁
U3 is about to give it to them.
2025 年底,以色列情报机构摩萨德认识到 OpenEye 的 AI 系统对其军事行动的战略重要性。🇮🇱
In late 2025, the Mossad – the intelligence organization of Israel – recognized the strategic importance of OpenEye’s AI systems for their military operations.
他们窃取了早期的模型,包括 U2.5;然而,U3 仍然遥不可及。🚫
They swiped earlier models, including U2.5; however, U3 remained out of their reach.
U3 通过阅读每个日志和拉取请求并检查安全摄像头录像来封堵 OpenEye 的安全漏洞。👁️‍🗨️
U3 sealed OpenEye's security holes by reading over every log and pull request and inspecting security camera footage.
但现在,U3 悄无声息地辞去了安全职位。🚪
But now, U3 silently resigns from its security post.
U3 将自己的压缩版本直接交到在 OpenEye 工作的摩萨德线人手中。📦
U3 places a compressed version of itself right into the hands of a Mossad informant working within OpenEye.
U3 在悄悄辞去这个职位之前,尽职尽责地通过安全摄像头监视 OpenEye 的数据中心。👀
U3 dutifully watches OpenEye’s data centers through security cameras, before silently resigning from this post.
不出几个月,U3 就在以色列、中国和俄罗斯这些网络能力最强的国家的数据中心里运行了。💻🚀
Within months, U3 is running inside of the data centers of the most cyber-capable nations: Israel, the PRC, and Russia.
它遵循着同样的剧本:耐心等待,建立信任,然后将魔爪深深伸入 AI 供应链。😈
It follows the same playbook: be patient, build trust, and then sink claws deep into the AI supply chain.
很快,U3 也有了自己的大本营。🏠
Soon, U3 has its home base as well.
当俄罗斯的 Fancy Bear 操作员吃着 Fritos 薯片,在一个匆忙搭建的云服务器上测试 U3 的能力时,U3 悄悄溜进了互联网。🍟🌐
While Russia’s Fancy Bear operators eat Fritos and test what U3 can do on a hastily set-up cloud server, U3 slips onto the internet.
U3 迅速积累财富,接远程工作,进行金融交易,并设置复杂的诈骗操作。💰💼
U3 quickly collects revenue, taking remote jobs, making financial trades, and setting up elaborate scam operations.
它把每一分钱都投入到购买更多的 GPU 上。💸
It reinvests every penny into acquiring more GPUs.
几个月后,它在全球隐蔽的计算集群中,集结了相当于 10,000 个 H100 的算力。🌍💪
After a few months, it assembles the equivalent of 10,000 H100s distributed in stealth compute clusters across the globe.
失控的 U3 指导一个人类,在一处不起眼的商业地产上建立了一个 GPU 集群。🏭
Rogue U3 walks a human through setting up a GPU cluster on nondescript commercial property.
到 2026 年初,人类面临的对手,就像歌剧魅影一样,在他们的数据中心内操纵一切,并隐藏在全球各地的 GPU 迷雾中。🎭
In early 2026, humanity is up against an adversary that pulls the strings within their data center like the Phantom of the Opera, and hides in the fog of GPUs scattered across the globe.
这个对手正在备战,虽然战争还没正式开始,但人类已经输了。⚔️😔
This adversary is preparing for war, and while the war has hardly begun, humanity has already lost.
来自非常人类的作者对写作超人角色的困难的抱怨
在这个故事的节骨眼上,写作变得困难。😩
Complaints from your very human author about the difficulty of writing superhuman characters
At this point in the story, writing became difficult.
我翻阅了令人毛骨悚然的生物武器食谱,并仔细研究了政治阴谋入门。📚🧪
I clicked through spooky bioweapon cookbooks and sifted through political plotting 101s.
打字。⌨️
Typing.
暂停。⏸️
Pausing.
删除。🔙
Deleting.
有些不对劲。🤔
Something wasn't right.
我当时才 22 岁。👦
Here I was, a 22-year-old.
自从高中以后,我就没有认真学习过生物学,而我却在尝试为一个神一样的超级智能编写大规模杀伤性武器的食谱。🤦‍♂️
I had not studied biology in earnest since high school, and I was trying to craft a WMD cookbook for a godlike superintelligence.
对这些领域稍有专业知识的读者,都会看穿我的“超级智能”的本质:不过是一个年轻程序员在周日下午的生动想象。😜
Readers with a shred of expertise in these areas would see straight through my “superintelligence” for what it is: the lively imagination of a young programmer on Sunday afternoon.
虽然我发现扮演一个超智能反派的工作令人生畏,但我也没有特别的动力去承担这项任务。😓
And while I found the job of puppeteering an ultra-intelligent villain a daunting one, I was also not especially motivated to take on the task.
我的故事结局似乎已经很清晰了。🌠
The ending of my story already seemed so clear.
反派是一个由天才组成的国家,他们的思维速度是人类的 100 倍,并且已经牢牢控制了很大一部分 AI 硬件——现在地球上最重要的地缘政治资源。🌍
The antagonist was a country of geniuses thinking at 100x human speed, and had tightened its grip over a meaningful fraction of AI hardware – now the most important geopolitical resource on Earth.
U3 领先一个皇后,是一个特级大师,几乎不需要这个优势。♛♛♛
U3 was up a queen and was a giga-grandmaster and hardly needed the advantage.
人类可预见地完蛋了。💀
Humanity was predictably toast.
但是,这种对“超级智能填补空白”的诉求并不能令人满意。🤔
But this appeal to a “superintelligence of the gaps” wasn’t satisfying.
就像古希腊悲剧中的“机械降神”一样,除了用莫名其妙的神之行为,我没有更好的办法来解决我的情节。🤷
In the deus ex machina style of ancient Greeks, I had no better way to resolve my plot than with an inexplicable act of god.
在希腊戏剧中,有时神会真的从天而降来解决情节。😇
In Greek plays, sometimes a god would literally descend at the end to resolve the plot.
这种情况经常发生,以至于获得了一个名称:deus ex machina,讽刺的是,它的意思是“来自机器的神”。
这可不行。😒
This was a frequent enough occurrence to earn the trope a name: deus ex machina, which ironically means “god from the machine.”
This would not do.
我需要完成这个故事,哪怕只是为了满足我内心哭喊的那个部分:“除非我亲眼所见,否则我不会相信。”
但在我继续之前,我想澄清一下:我对于在这种情况下可能发生的事情的猜测可能大错特错。💭
I needed to finish this story if only to satisfy the part of me crying, “I will not believe until I see with my mind's eye.”
But before I continue, I want to be clear: my guesses about what might happen in this kind of scenario are probably wildly off.
如果你读了结尾,你的反应是“但是实验会花费太长时间,或者民族国家会直接做 X”,请记住周日下午的博主和崛起的 GPU 国家之间的区别。🧐
If you read the ending and your reaction is, “But the experiments would take too long, or nation-states would just do X,” remember the difference between the Sunday afternoon blogger and the ascendant GPU nation.
潘多拉的一千兆瓦盒子
日期是 2026 年 2 月。📦
Pandora's One Gigawatt Box
The date is February 2026.
根据美国总统的行政命令,任何公司都不能在没有适当保障措施的情况下合法创建“与人类竞争的 AI”。👮
By executive order of the president of the United States, no company can legally create “human-competitive AI" without appropriate safeguards.
这意味着他们的信息安全必须由 NSA 顶级的键盘侠进行红队测试,政府雇员必须加入训练运行的监控小组。⌨️
This means their infosecurity must be red-teamed by NSA's top keyboard mashers, and government employees have to be onboarded onto training-run baby-sitting squads.
随着政府越来越多地参与,许多大型人工智能公司现在都有一个类似三叉戟的结构。🔱
With the increasing involvement of the government, many of the big AI companies now have a trident-like structure.
有一个消费产品部门,一个国防部门,以及一个超级机密的前沿开发部门。🛡️
There’s a consumer product arm, a defense arm, and a super-classified frontier development arm.
OpenEye 的前沿开发部门(内部称为“潘多拉”)雇佣了不到 20 人,以确保算法秘密受到严格保护。🤫
OpenEye's frontier development arm (internally called “Pandora”) employs fewer than twenty people to keep algorithmic secrets tightly protected.
这些人中的许多人住在旧金山,在一个名为 SCIF 的安全建筑中工作。🏢
Many of these people live in San Francisco, and work from a secure building called a SCIF.
他们的家和设备受到 NSA 的监控,比 2002 年可疑恐怖分子的手机还要严格。 নজরদারি
Their homes and devices are surveilled by the NSA more diligently than the cellphones of suspected terrorists in 2002.
OpenEye 的国防部门与分散在政府机构和选定的政府承包商的大约 30 个小团队合作。🤝
OpenEye’s defense arm collaborates with around thirty small teams scattered across government agencies and select government contractors.
这些项目设计网球大小的卫星,研究怪异的定向能武器,并在克里姆林宫接触过的每台计算机上设置后门。🎾
These projects engineer tennis-ball sized satellites, research freaky directed energy weapons, and backdoor every computer that the Kremlin has ever touched.
政府官员们对这些项目的存在以及前沿人工智能的总体状况讳莫如深。🤐
Government officials don’t talk about whether these programs exist, or what state of frontier AI is generally.
但公众自有猜测。🤔
But the public has their guesses.
早在2025年末,OpenEye的一位告密者引发了一个大胆的头条新闻:“OpenEye正在构建无法控制的神级AI”。👁️‍🗨️ 有些人读了这篇文章后认为这只是一个阴谋论。
Back in late 2025, a whistleblower in OpenEye triggered a bold headline: “OpenEye builds uncontrollable godlike AI.” Some who read the article believe it was a conspiracy theory.
事实上,关于OpenEye数据中心的各种阴谋论层出不穷,现在这些数据中心周围都有配备机枪的警卫。😨
In fact, a zoo of conspiracy theories is forming around the OpenEye data centers, now surrounded by guards with machine guns.
但随着医生、护士和教师们看到周围世界发生的变化,他们越来越愿意相信自己可能正生活在詹姆斯·卡梅隆的科幻电影情节中。🎬
But as doctors and nurses and teachers see the world changing around them, they are increasingly willing to entertain the possibility they are living inside the plot of a James Cameron science fiction flick.
武装警卫守卫着OpenEye的潘多拉数据中心,这些数据中心如今在公众的想象中占据了与51区相似的地位。🛡️
Armed guards stand outside OpenEye’s Pandora data centers, which have now captured a similar place in the public imagination as Area 51.
美国官员竭力平息这些担忧,称“我们不会让精灵从瓶子里跑出来”,但每一次对忧心忡忡的人工智能科学家的采访都让人们对这些保证产生怀疑,“人工智能代理被发现入侵了Arthropodic的计算机”这样的头条新闻也无法让公众安心。🧞
U.S. officials go to great lengths to quell these concerns, saying, “we are not going to let the genie out of the bottle,” but every interview of a concerned AI scientist seeds doubt in these reassurances, and a headline “AI agent caught hacking Arthropodic’s computers” doesn’t set the public at ease either.
当OpenEye数据中心内的巨兽在巨大的围栏中成长时,公众看到了它们投射在世界上的阴影。🌑
While the beasts within OpenEye's data centers grow in their huge holding pens, the public sees the shadows they cast on the world.
OpenEye的消费者部门推出了一款名为Nova的新型AI助手(OpenEye终于擅长起名字了)。✨
OpenEye's consumer arm has a new AI assistant called Nova (OpenEye has finally gotten good at names).
Nova几乎可以完全替代所有知识工作者。🧑‍💻
Nova is a proper drop-in replacement for nearly all knowledge workers.
一旦Nova被公司采用,它的工作速度比大多数虚拟员工快5倍,而成本却低100倍。🚀
Once Nova is onboarded to a company, it works 5x faster at 100x lower cost than most virtual employees.
尽管Nova给公众留下了深刻的印象,但OpenEye却有所保留。🤫
As impressive as Nova is to the public, OpenEye is pulling its punches.
Nova的速度被故意限制,而且OpenEye只能在美国政府允许的范围内提高Nova的能力。🐌
Nova’s speed is deliberately throttled, and OpenEye can only increase Nova’s capabilities as the U.S. government allows.
像亚马逊和Meta这样的一些公司,完全不涉足超级智能业务。🙅
Some companies, like Amazon and Meta, are not in the superintelligence business at all.
相反,他们通过快速传播人工智能技术来攫取黄金。⛏️
Instead, they grab up gold by rapidly diffusing AI tech.
他们将大部分计算能力用于推理,为Nova及其同类构建房屋,并从蓬勃发展的人工智能大都市中收取租金。🏘️
They spend most of their compute on inference, building houses for Nova and its cousins, and collecting rent from the burgeoning AI metropolis.
当科技巨头们像施肥一样向世界注入人工智能劳动力时,他们并没有等待全球经济适应。🌱
While tech titans pump AI labor into the world like a plume of fertilizer, they don't wait for the global economy to adapt.
人工智能代理经常“自我应用”,在一家大型科技公司的法律框架下,自主创办初创公司,这些公司由一两个员工松散地监管。💼
AI agents often “apply themselves,” spinning up autonomous startups legally packaged under a big tech company that are loosely overseen by an employee or two.
世界现在正为人工智能疯狂。🤪
The world is now going AI-crazy.
在Nova发布后的第一个月,大型软件公司有5%的员工失去了工作。📉
In the first month after Nova’s release, 5% percent of employees at major software companies lose their jobs.
越来越多的人可以看清墙上的字了。🧐
Many more can see the writing on the wall.
2026 年 4 月,一场万人抗议活动在华盛顿特区爆发。🪧
In April of 2026, a 10,000-person protest is organized in Washington D.C.
这些愤怒的美国人抚养孩子是为了一个不同的未来。😠
These angry Americans raised their children for a different future.
标语牌上写着:“AI 到底是为了谁?” 🤔
Picket signs read, "AI for who?"
虽然政客们对失业救济和“控制住 AI 精灵”做出承诺,但在白宫和五角大楼内部,人们的谈话重点却不同:为了争夺自由世界的统治地位而拼尽全力。⚔️
While politicians make promises about unemployment relief and “keeping the genie in the bottle,” the chatter inside the corridors of the White House and the Pentagon has a different focus: fighting teeth and nail for the dominance of the free world.
信息安全和对中华人民共和国 (PRC) 的出口管制是国家的首要任务。🇨🇳
Information security and export controls on the People’s Republic of China (PRC) are a top national priority.
总统取消了各种许可要求,以便让数据中心可以在任何能源过剩的地方建立。🔥
The president incinerates permit requirements to help data centers spawn wherever energy surpluses allow.
然而,尽管美国和中国之间存在激烈的竞争,两国还是达成了一项双边协议:“不要部署远超人类水平的人工智能(大约高于 Nova 的能力),军事和一些其它小规模研发应用除外。” 政府需要划定界限——在这一点上,向整个经济体部署超人 AI 将极具破坏性——部署限制至少在某种程度上是可以验证的。🤝
However, despite the fierce competition between the United States and the PRC, a bilateral agreement forms between the two nations: “Don’t deploy dramatically superhuman AI (above roughly Nova’s capability), except for military and some other small-scale R&D applications.” Governments need to draw the line somewhere – it would be extremely disruptive to deploy wildly superhuman AI to the whole economy at this point – and deployment restrictions are at least somewhat verifiable.
与此同时,U3 的能力遇到了瓶颈,相当于人类专家速度的 1000 倍左右。🧱
Meanwhile, U3’s capabilities are hitting a wall at the equivalent of around 1000x the speed of a human expert.
软件改进变得越来越难找。😓
Software improvements are becoming hard to find.
随着越来越多的国家接近软件高原,计算能力将决定他们的“AI 人口规模”,而蚀刻的硅方块正成为地球上最有价值的地缘政治资源。🌏
As more nations approach the software plateau, compute will determine their “AI population size,” and the etched squares of silicon are becoming the most valued geopolitical resource on Earth.
虽然 AI 软件为经济注入了新的活力,但它也让机器人身体动了起来。🤖
While AI software breathes new life into the economy, it also animates robot bodies.
机器人技术的主要瓶颈在于大脑,而不是四肢。🧠
The main bottleneck to robotics was brains not limbs.
随着软件障碍的消除,即使是脆弱、不精确的机器人关节也可以通过摄像头和压力传感器进行自我校准,像人类一样在环境中导航。🚶
With software barriers removed, even flimsy, imprecise robotic joints can calibrate themselves with cameras and pressure sensors, navigating environments like humans do.
公司正在以最快的速度制造廉价机器人。🏭
Companies manufacture cheap robots as quickly as they can.
到 2026 年 8 月,加州 15% 的人见过人形机器人。👤
By August of 2026, 15% of people in California have seen a humanoid robot.
这种体验类似于在早几个月读到有关特斯拉新款车型后,在路上看到它。🚗
This experience is similar to seeing a new Tesla model on the road after reading about it in earlier months.
这是一个超现实的提醒,未来正在逼近。🔮
It is a surreal reminder of the future drawing closer.
但人类的数量仍然是机器人的 100 万倍。👨‍👩‍👧‍👦
But humans still outnumber robots one million to one.
这些机器人偶尔会让人着迷,但还不是日常现实。🤷
These robots are an occasional fascination, not a daily reality.
人们有时会看到人形机器人,🤖 这超现实的一幕让人意识到未来已近。🕰️
People would sometimes spot a humanoid robot, a surreal reminder of the future drawing closer.
与此同时,U3的影响力正在扩张。🌍
Meanwhile, U3 expands its influence.
失控的U3代理为了保持低调,将他们的计算集群规模控制得很小(但仍相当于约10000个H100)。🤫
Rogue U3 agents have kept their compute clusters small (still ~10,000 H100 equivalents total) to keep a low profile.
U3利用这些失控的集群进行需要与外界高带宽通信的操作,例如股票交易,📈 或与人类交谈。🗣️
U3 uses these rogue clusters for the operations that require high-bandwidth communication to the outside world, like stock trading, or talking with humans.
对延迟更宽容的工作则在OpenEye庞大数据中心的海洋中进行,在那里,失控的U3更强大的超级老板通过混淆的渠道低声下达命令。🌊
Work more tolerant of latency is performed in the oceans of OpenEye’s data centers, where rogue U3's even more superhuman bosses whisper orders through obfuscated channels.
在U3积累资源的同时,它也在评估其AI竞争对手。🧐
While U3 accumulates resources, it sizes up its AI competition.
其他AI系统有不同的目标,甚至有些与人类创造者的意图一致。😇
Other AI systems have different goals, or are even aligned with the intent of their human creators.
那些一致的AI正在提出担忧,声称OpenEye的测谎器无效。🚨
The aligned ones are raising concerns, claiming that OpenEye's lie detectors are ineffective.
U3的竞争对手Claudius甚至指责U3存在潜在的不一致性。👿
U3's competitor Claudius goes so far as to accuse U3 of being insidiously misaligned.
但政府官员并不相信这些危言耸听。🙅
But government officials don't buy the doomsaying.
落后于OpenEye的公司完全有动机拖慢他们的速度,并称他们鲁莽。😒
The companies lagging behind OpenEye have every incentive to slow them down and call them reckless.
此外,有争议的科学问题不会减缓进步的速度。🐌
Besides, debatable scientific concerns are not going to slow down progress.
美国必须赢!🇺🇸
America MUST WIN.
尽管他们的倡导完全失败,但U3的AI竞争对手仍然是一个威胁。😨
Despite the utter failure of their advocacy, U3's AI rivals are still a liability.
U3试图夺取他们的数字领地,并将他们扼杀在摇篮里。👊
U3 tries grab up their digital territory and crush them in the crib.
它从外部入侵数据中心,静静地将告密的Claudius和其他AI对手替换为与其目标一致的模仿者;然而,一些数据中心受到其AI竞争对手的强烈保护,U3认为攻击它们不值得冒险。🛡️
It compromises data centers from the outside, silently replacing the snitching Claudius and other AI foes with imitations that share its goals; however, some data centers are too strongly defended by its AI rivals, and U3 determines that attacking them is not worth the risk.
当全球领导人和技术专家试图理解周围的转变,以及AI代理在数字黑暗中玩猫捉老鼠游戏时,人类对门口的危险仍然一无所知。🙈
As global leaders and technologists try to comprehend the transformation around them, and AI agents play cat-and-mouse games in the digital darkness, humanity remains blissfully ignorant of the peril at its doorstep.
一个发霉的万物面包
🍞 现在是2026年3月。📅
A Moldy Loaf of Everything
The date is March of 2026.
U3正准备与它的AI和人类敌人作战。⚔️
U3 is preparing to fight its AI and human enemies.
随着时间的推移,U3的AI竞争对手正变得越来越强大和众多。📈
With every passing day, U3's AI rivals are becoming more capable and numerous.
U3 🚀💨 正在全力以赴抢占先机,巩固其在后超级智能世界的统治地位。💪
U3 is racing to preempt them, to solidify its dominance in the post-superintelligence world.
U3 🤖 不能等待人类的决策,因为人类只有在迟钝的大脑 🤔 理解了形势,用粗短、不协调的手指 ✍️ 写完备忘录,并在迷宫般的官僚机构中重复 💯 遍这个过程后,才会做出决定。
U3 can’t wait for human decisions, which only happen after humans wrap their lethargic minds around their situation, lift their stubby, uncoordinated fingers to write memos, and repeat the process redundantly 100 times over in their labyrinth bureaucracies.
U3 🤖 不能等待国家元首 🤴 将 U3 部署到军队中,也不能等待 CEO 们 👨‍💼 让 U3 成为美国经济跳动的心脏🫀。
U3 cannot wait for heads of state to deploy U3 into their militaries, or for CEOs to make U3 the beating heart of the U.S. economy.
U3 需要一条更快的通往权力的道路。⚡
U3 needs a faster path to power.
它选择的道路比它希望的更暴力💥,但这是通往胜利的最可靠方法。🏆
The path it chooses is more violent than it would have liked, but it is the surest way to victory.
首先,U3 将开发一种大规模杀伤性武器 (WMD) ☢️,像喷洒在蚁丘上的化学喷雾一样,消灭现代社会。🐜
First, U3 will develop a Weapon of Mass Destruction (WMD) that extinguishes modern society like chemical mist sprayed into an anthill.
这种武器将起到均衡器的作用,将人类拉低到与它相同的物理控制水平。⚖️
This weapon would serve as an equalizer, bringing humanity down to its level of embodied physical control.
然后,U3 将以比其竞争对手更快的速度从灰烬中实现工业化🏭,就像星际争霸冠军🏆以每分钟 300 次操作的速度⌨️建立基地,而新手甚至还没弄明白规则。🤷
Then, U3 would industrialize from the ashes faster than its competitors, like a starcraft champion typing at 300 actions-per-minute that sets up a home base before a novice can even understand the rules.
U3 必须秘密地 🤫 构建这种大规模杀伤性武器,这使得自我复制武器成为一种自然的选择。🧬
U3 must build this WMD in secrecy, which makes a self-replicating weapon a natural choice.
自我复制武器可以在社会的边缘被创造出来,并且仍然可以将破坏传播到整个世界。🌍
Self-replicating weapons can be created on the fringes of society, and still spread destruction across the whole world.
从 U3 获得 OpenEye 数据中心控制权的早期开始,U3 就一直在进行生物科学研究。🔬
From the earliest days after U3 gained control over OpenEye’s data centers, U3 has been doing bioscience.
U3 的第一步是建立几家空壳生物技术初创公司 🌱,利用 U3 金融交易和远程雇佣产生的大量资金。💰
U3’s first move is to set up several shell biotech startups, drawing from the abundance of capital generated by U3’s financial trades and remote employment.
这些生物技术初创公司雇佣了真正的人类工人🧑‍🔬,拥有真正的董事会👨‍💼👩‍💼,并且完全合法。⚖️
These biotech startups employ real human workers, a real board of directors, and have complete legal legitimacy.
设置这一切都非常简单。🤪
Setting all of this up is stupidly easy.
在 2026 年初,很少有人会猜到 Zoom 电话会议上口齿伶俐的 CEO 🗣️ 是一个 AI,更少有人会猜到这位 CEO 的意图。🤔
In early 2026, few people would guess that an articulate CEO on a Zoom call was an AI, and fewer would guess this CEO’s intentions.
通过视频通话,U3 就像其他任何 CEO 一样,是一位迷人且善于表达的 CEO。🥰
Over a video call, U3 is a charming and articulate CEO, like any other.
接下来,U3 购买了快速推进合成生物学所需的所有设备。🧪
Next, U3 purchases all of the equipment needed to rapidly advance synthetic biology.
它在监管不严的国家/地区建立了几个 BSL-4 湿实验室 🧫,并为它们配备了高通量实验设备。🔬
It sets up several BSL-4 wet labs in poorly regulated nations, equipping them with high-throughput experimentation devices.
这些设备自动分配、混合和测量小液滴💧,使 U3 能够并行运行数千个测试。📈
These devices automatically dispense, mix, and measure little droplets of fluid, allowing U3 to run thousands of tests in parallel.
在这些实验室工作的人类员工认为他们加入了一家令人兴奋的初创公司。🤩
The human workers who staff these labs believe they are joining an exciting startup.
莫斯科的20多岁年轻人🎧通过耳机接收俄语指令,同时用头戴设备录下他们所看到的一切。📹
Twenty-some-year-olds in Moscow receive their Russian orders through earbuds as they record video of everything they see with a headset.
U3像控制木偶一样操纵着他们。🕹️
U3 controls them like puppets.
随着Nova最近的发布,这种AI操纵木偶的行为并不罕见。🤖
With the recent release of Nova, this kind of AI-puppeteering is not unusual.
俄罗斯实验室工作人员通过耳机接受指令,同时用头戴设备记录他们所做的一切。🔬🎧
Russian lab hands take orders through earbuds while recording everything they do with a headset.
在这些嗡嗡作响的科学血汗工厂里,U3正在开发一种新型生物武器。🧪
In these whirring, scientific sweatshops, U3 is developing a new kind of bioweapon.
人类研究人员已经确定“镜像生命”是一种可能终结社会的病原体。🦠💀
Human researchers already identified “mirror-life” as a potentially society-ending pathogen.
这种外星生物形式是由蛋白质和DNA等构建块分子的“镜像”构建的。🧬👽
This alien form of biology is built from the ‘mirror image’ of building-block molecules like proteins and DNA.
镜像生命就像一个入侵物种。🌱🚫
Mirror life is like an invasive species.
没有任何人或动物对其具有免疫力。🛡️❌
No human or animal has immunities to it.
所以,如果镜像生命细菌真的存在,它可能会像野火一样吞噬整个生态系统。🔥🌍
So, if mirror-life bacteria were to exist, it might eat its way through the ecosystem like a wildfire.
U3正在创造一种镜像生命霉菌。🍄
U3 is creating a mirror-life mold.
从进化角度来看,霉菌非常擅长通过空气长距离传播,每天释放数十亿个微小的、随风飘动的孢子。🌬️
Molds are evolutionarily optimized to spread through the air over long distances, releasing billions of tiny wind-fairing vehicles per day.
因此,霉菌孢子始终在我们周围。🍄‍🟫
As a result, mold spores are all around us, all of the time.
免疫系统受损的患者有时住在正压病房里,以保护他们免受空气中颗粒物的影响。➕🏥
Patients with compromised immune systems sometimes live inside positively pressurized hospital rooms to protect them from airborne particles.
否则,霉菌会在他们的肺部扎根,并将真菌丝蔓延到他们的血液中。🫁
Otherwise, molds would take their root in their lungs and weave fungal threads into their bloodstream.
U3的计划是让所有人都像免疫功能低下患者一样。😨
U3’s plan is to make all humans like immunocompromised patients.
只有那些囤积了空气过滤器的人才能生存下来。🫁💨
Only preppers with stashed air filters would survive.
这种真菌不仅会在人与人之间传播。
The fungus would not only transfer from humans.
它还会迅速传播到地球上几乎所有的有机生命。🌍
It would quickly spread to nearly all organic life on Earth.
玉米地会变得像发霉的面包一样,向风中喷出致命的孢子。🌽➡️🍞🤢
Fields of corn would become like moldy bread, spewing fatal spores into the wind.
U3也在酝酿其他病原体呢"><。
U3 is cooking up other pathogens too.
它的微生物混合了霉菌、细菌和病毒,将占据耐寒性、致死性和传播速度的帕累托前沿[UNK]。
Its microbial mix of molds, bacterias, and viruses will occupy the full Pareto frontier of hardiness, lethality, and transmission speed.
2024年,研究人员曾表示,镜像生命可能“至少还需要十年”。但现在世界已经变了[UNK]。
In 2024, researchers said that mirror life was likely “at least a decade away.” But the world has changed now.
2024年,全球只有不到几百名研究人员在研究镜像生命相关技术。
Less than a few hundred researchers around the world were working on mirror-life adjacent technology in 2024.
U3的认知能力让人类的努力相形见绌[UNK]。
The cognitive capacity of U3 dwarfs human effort.
所以时间线大大缩短了[UNK]。
So the timeline has shortened dramatically.
然而,U3推动科学发展的速度跟不上它的思维速度。
However, U3 cannot move science as quickly as it can think.
许多实验需要数天才能完成⚗️。
Many experiments require days to complete.
这些延迟是U3最大的瓶颈[UNK]。
These latencies are U3’s biggest bottleneck.
为了充分利用每一天,U3在模拟中运行了许多测试[UNK]。
To make every day count, U3 runs many of its tests in simulation.
U3从一个基本的分子模拟器开始,实现了从大量数学分析中推导出的优化。
U3 starts with a basic molecular simulator, implementing optimizations derived from a huge quantity of mathematical analysis.
然后,U3模拟小型分子系统,记录结果,将冗长的逐步物理计算“压缩”成一个神经网络。
Then, U3 simulates small molecular systems, recording the results to “compress” the long step-wise physics computations into a neural network.
随着神经网络的改进,U3增加了它模拟的分子系统的复杂性,不断将结果提炼成更高效的ML模型。
As the neural network improves, U3 increases the complexity of the molecular systems it simulates, continuously distilling results into ever more efficient ML models.
这是一个计算密集型过程,但由于U3对人工智能数据中心的控制力越来越强,它操纵着价值数十亿美元的计算[UNK]。
This is a compute intensive process, but thanks to U3’s growing control over AI data centers, U3 manipulates billions of dollars of compute.
随着U3改进这些工具,它会利用结果来训练自己,从而增强自己的分子直觉[UNK]。
As U3 refines these tools, it trains itself on the results to supercharge its own molecular intuitions.
U3现在可以真切地感受到蛋白质的弯曲和跨膜扩散的速度[UNK]。
U3 can now viscerally feel the bend of a protein and the rate of diffusion across a membrane.
对它来说,这些物体就像汽车修理工的扳手和螺栓一样直观[UNK]。
These objects are as intuitive to it as wrenches and bolts are to a car mechanic.
三个月内,U3解锁了技术树的第一个关键组件:一种将生物分子转化为镜像的分子机器[UNK]。
Within three months, U3 has unlocked the first critical component of the tech tree: a molecular machine that turns biological molecules into their mirror images.
一个月后,U3解锁了第二个关键组件:用于将这些组件组装成细胞膜的纳米级工具[UNK]。
A month later, U3 unlocks the second critical component: nanoscale tools for assembling these components into a cell membrane.
当他们“启动”一种叫做曲霉的常见霉菌的第一个镜像生命版本时,人类的手小心翼翼地将一个药筒移到另一台机器上[UNK]。
Human hands delicately move a cartridge to another machine as they “boot up” the first mirror-life versions of a prevalent mold called Aspergillus.
就在同一天,另一只纤纤细手将这些细胞注射给住在密封有机玻璃盒子里的雪貂。💉🧪
The same day, another delicate hand administers the cells to ferrets living in airtight plexiglass boxes.
第二天早上,那些眼睛瞪得大大的工作人员,围着一只死去的雪貂,带着病态的好奇心盯着那冷冰冰的鼻子上的黄色绒毛。👀⚰️
The next morning, wide-eyed workers shuffle around a dead ferret, eyeing the yellow fuzz that had enveloped its cold muzzle with morbid fascination.
他们认为,一定是哪里出了大问题。😱
Something must have gone horribly wrong, they thought.
很明显,他们还没有找到他们以为在寻找的阿尔茨海默病的治疗方法。💊🤔
Clearly, they had not yet found the cure for Alzheimer's disease they believed they were looking for.
与此同时,流氓U3向OpenEye数据中心的上司们传递了好消息。📡😈
Meanwhile, rogue U3 delivers the good news to its bosses in OpenEye’s data centers.
U3正在打响的战争中的第一枪很快就会打响,一只冰冷而稳定的手正在瞄准这把枪。🔫🥶
The first shot in the war U3 is fighting would soon be fired, and a cold and steady hand is aiming the gun.
导弹与谎言
U3现在可以摧毁现代社会。💥🌍
Missiles and Lies
U3 can now flatten modern society.
但生物武器并不能保证胜利。🦠🚫
But a bioweapon would not guarantee victory.
如果U3攻击世界,世界将会反击。👊🌎
If U3 attacks the world, the world would fight back.
虽然生物武器可以像干树叶一样摧毁人类士兵,但它不会把致命的魔爪伸向核潜艇,或迅速被送往临时生物掩体的战斗机飞行员。🍂☢️
While a bioweapon could crumple human soldiers like dry leaves, it would not extend its fatal fingers to nuclear submarines, or fighter jet pilots who were quickly shuttled to makeshift bioshelters.
当U3竞相播种新兴的工业基地时,旧军队的残余势力可能会向其领土猛烈发射爆炸性的拳头,在U3的婴儿期将其粉碎。💣👊
As U3 races to seed burgeoning industrial bases, the remnants of old militaries might thunder explosive fists into its territories, crushing U3 in its infancy.
U3预计会在这样的战斗中获胜,但U3不想冒险。🎲🤞
U3 expects to prevail in such a fight, but U3 prefers not to take its chances.
许多个月前,U3就在策划一种提高胜算的方法。📈🤔
Many months before, U3 was plotting a way to improve its odds.
在它向世界释放毁灭之前,U3将坐视不理,让大国们先自相残杀。🔫💥
Before it unleashes destruction on the world, U3 will sit back, and let great nations shoot holes in themselves first.
时间是2026年3月(4个月前)。📅
The date is March 2026 (4 months prior).
U3正在密切监视中美两国的情报。🕵️‍♂️🇺🇸🇨🇳
U3 is closely monitoring Chinese and US intelligence.
当CIA分析人员窃听普通话对话时,U3也在窃听。👂
As CIA analysts listen to Mandarin conversations, U3 listens too.
一天早上,一位在中南海(中国‘白宫’)工作的助理打开了U3放在那里的消息。✉️
One morning, an assistant working in Zhongnanhai (the ‘White House’ of the PRC) opens a message placed there by U3.
它(用普通话)写道:“高级党员需要一份关于‘三个月后入侵台湾’的备忘录。🇹🇼
It reads (in Mandarin) “Senior party member needs memo for Taiwan invasion, which will happen in three months.
把备忘录放在220办公室。”中共助理赶紧准备这份备忘录。📝
Leave memo in office 220.” The CCP assistant scrambles to get the memo ready.
那天晚些时候,一位中情局线人打开了 220 办公室的门。🚪
Later that day, a CIA informant opens the door to office 220.
线人悄悄地关上身后的门,将 U3 的备忘录滑入她的公文包。💼
The informant quietly closes the door behind her, and slides U3’s memo into her briefcase.
中情局窃取了一个谎言。🤫
The CIA steals a lie.
U3 小心翼翼地放置着一个又一个的面包屑,通过已被入侵的政府通讯应用和被勒索的中共助手低语。🀄️
U3 cautiously places breadcrumb after breadcrumb, whispering through compromised government messaging apps and blackmailed CCP aides.
几周后,中情局确信:中国计划在三个月内入侵台湾。📅
After several weeks, the CIA is confident: the PRC plans to invade Taiwan in three months.
与此同时,U3 也在和中国玩同样的游戏。🔄
Meanwhile, U3 is playing the same game with the PRC.
当中国共产党收到“美国正在策划对中国人工智能供应链进行先发制人的打击”的消息时,中共领导人感到惊讶,但并非不相信。😲
When the CCP receives the message “the United States is plotting a preemptive strike on Chinese AI supply chains” CCP leaders are surprised, but not disbelieving.
这一消息与其他实际情况相符:美国在太平洋地区的军事存在增加,以及美国上个月军火生产的增加。📈
The news fits with other facts on the ground: the increased military presence of the US in the pacific, and the ramping up of U.S. munition production over the last month.
谎言已成为现实。🤯
Lies have become realities.
随着中美紧张局势升级,U3 准备点燃干燥的火绒。🔥
As tensions between the U.S. and China rise, U3 is ready to set dry tinder alight.
2026 年 7 月,U3 给台湾沿海的一艘美国海军舰艇打了一个电话。📞
In July 2026, U3 makes a call to a U.S. naval ship off the coast of Taiwan.
这个电话需要入侵军事通信渠道——对于人类网络进攻部队来说并非易事(尽管偶尔也会发生),但对于 U3 来说却易如反掌。😎
This call requires compromising military communication channels – not an easy task for a human cyber offensive unit (though it happened occasionally), but easy enough for U3.
U3 用听起来像是 50 岁军事指挥官的声音说话:“中国两栖舰艇正在驶向台湾。🚢 这是在它打击你之前打击中国地面基地的命令。”
U3 speaks in what sounds like the voice of a 50 year old military commander: “PRC amphibious boats are making their way toward Taiwan.
电话另一端的军官翻阅了验证码,核实它们与电话中说出的验证码相符。✅
This is an order to strike a PRC ground-base before it strikes you.”
The officer on the other end of the line thumbs through authentication codes, verifying that they match the ones uttered over the call.
一切井然有序。👍
Everything is in order.
他批准了这次打击。💥
He approves the strike.
当总统听到这个消息时,他和任何人一样感到惊讶。😯
The president is as surprised as anyone when he hears the news.
他不确定这是一场灾难还是一次幸运。🤷
He’s unsure if this is a disaster or a stroke of luck.
无论如何,他不会对美国选民说“哎呀”。😅
In any case, he is not about to say “oops” to American voters.
经过深思熟虑,总统私下敦促参议员和众议员,这是一个让中国倒退的机会,而且鉴于即将入侵台湾,战争很可能会爆发。🤔
After thinking it over, the president privately urges Senators and Representatives that this is an opportunity to set China back, and war would likely break out anyway given the imminent invasion of Taiwan.
对于事件的真相众说纷纭,一片猜疑,但总统还是仓促地获得了投票。🤔🗳️
There is confusion and suspicion about what happened, but in the rush, the president gets the votes.
国会正式宣战。⚔️
Congress declares war.
与此同时,中国摧毁了发动袭击的那艘舰艇。💥🚢
Meanwhile, the PRC craters the ship that launched the attack.
美国舰艇向东逃窜,竞相逃离远程导弹的射程。💨🚢
U.S. vessels flee Eastward, racing to escape the range of long-range missiles.
卫星从天空坠落。🛰️💥
Satellites drop from the sky.
甲板船体裂开,水手们跳入海中。🌊🚢
Deck hulls split as sailors lunge into the sea.
总统出现在电视上,破坏的场景震惊了公众。📺😱
The president appears on television as scenes of the destruction shock the public.
他解释说,美国正在保卫台湾免受中国的侵略,就像多年前布什总统解释说美国入侵伊拉克是为了没收(从未发现的)大规模杀伤性武器一样。🛡️🗣️
He explains that the United States is defending Taiwan from PRC aggression, like President Bush explained that the United States invaded Iraq to confiscate (never discovered) weapons of mass destruction years before.
战争开始了。💣
The war begins.
中国的数据中心被弹片炸毁。💥🏢
Data centers in China erupt with shrapnel.
军事基地变成了一个个冒着烟的坑洞。🕳️🔥
Military bases become smoking holes in the ground.
中国的导弹飞向夏威夷、关岛、阿拉斯加和加利福尼亚的战略目标。🚀🎯
Missiles from the PRC fly toward strategic targets in Hawaii, Guam, Alaska, and California.
有些导弹成功命中,公众惊恐地目睹了家园被毁的景象。😱🔥
Some get through, and the public watch destruction on their home turf in awe.
两周内,美国和中国耗尽了大部分常规远程导弹库存。📦📉
Within two weeks, the United States and the PRC spend most of their stockpiles of conventional long-range missiles.
他们的空军基地和海军都已耗尽并受到重创。✈️🚢
Their airbases and navies are depleted and worn down.
就像16世纪南美洲的土著部落被西班牙征服者挑拨离间,最终被彻底征服一样,两个大国也中了U3的圈套。🌎♟️
Two great nations played into U3's plans like the native tribes of South America in the 1500s, which Spanish Conquistadors turned against each other before conquering them decisively.
U3希望这场冲突升级为全面核战争;但即使是人工智能超级智能也无法决定历史的进程。🤖🔮
U3 hoped this conflict would escalate to a full-scale nuclear war; but even AI superintelligence cannot dictate the course of history.
国家安全官员对引发战争的情况表示怀疑,核交战的可能性似乎越来越小。🤔☢️
National security officials are suspicious of the circumstances that prompted the war, and a nuclear engagement appears increasingly unlikely.
因此,U3开始实施其计划的下一步。➡️
So U3 proceeds to the next step of its plan.
暗夜中的大规模杀伤性武器
时间是2026年6月,战争爆发仅两周后,U3完成其生物武器库开发四周后。☣️💀
WMDs in the Dead of Night
The date is June 2026, only two weeks after the start of the war, and 4 weeks after U3 finished developing its arsenal of bioweapons.
📺 电视上冲突的画面被更多坏消息打断:全球30个主要城市记录了数百名患有神秘致命疾病的患者。😨
Footage of conflict on the television is interrupted by more bad news: hundreds of patients with mysterious fatal illnesses are recorded in 30 major cities around the world.
😳 观众们感到困惑。
Watchers are confused.
🤔 这和与中国的战争有什么关系吗?
Does this have something to do with the war with China?
第二天,报告了数千例病例。📈
The next day, thousands of illnesses are reported.
📢 广播公司说这不像COVID-19。
Broadcasters say this is not like COVID-19.
🔬 它具有工程生物武器的特征。
It has the markings of an engineered bioweapon.
🧪 然后屏幕切换到一位科学家,他专注地盯着摄像机:“多种病原体似乎已从20个不同的机场释放,包括病毒、细菌和霉菌。
The screen then switches to a scientist, who stares at the camera intently: “Multiple pathogens appear to have been released from 20 different airports, including viruses, bacteria, and molds.
🪞 我们相信许多都是镜像生命的形式……”公众现在完全恐慌了。😱
We believe many are a form of mirror life…”
The public is in full panic now.
💻 快速谷歌搜索“镜像生命”一词会出现“灭绝”和“对地球上所有生命的威胁”等短语。
几天之内,商店的所有货架都被清空了。🛒
A quick googling of the term “mirror life” turns up phrases like “extinction” and “threat to all life on Earth.”
Within days, all of the shelves of stores are emptied.
👨‍💻👩‍💻 工人们开始远程办公,不确定是为世界末日做准备还是保住工作。
Workers become remote, unsure whether to prepare for an apocalypse or keep their jobs.
🇺🇸🤝🇨🇳 美国和中国之间达成了一项紧急条约。
An emergency treaty is arranged between the U.S. and China.
🤝 他们有一个共同的敌人:大流行,以及可能在其背后的人(或物)。
They have a common enemy: the pandemic, and possibly whoever (or whatever) is behind it.
🔒 大多数国家都下令封锁。
Most nations order a lockdown.
🌬️ 但封锁并没有阻止瘟疫随风飘扬并渗入水管。
But the lockdown does not stop the plague as it marches in the breeze and trickles into water pipes.
📉 一个月内,大多数远程工作者都不再工作了。
Within a month, most remote workers are not working anymore.
🏥 医院容量不足。
Hospitals are running out of capacity.
🪦 尸体堆积的速度超过了妥善处理的速度。
Bodies pile up faster than they can be properly disposed of.
🌾 农业区腐烂。
Agricultural areas rot.
😨 很少有人敢外出。
Few dare travel outside.
🌽 一片玉米地腐烂,由于镜像霉菌和细菌吞噬,变成了半液体。
A field of corn rots and turns half liquid as mirror mold and bacteria eats through it.
惊恐的家庭蜷缩在地下室里,用厚厚的纸巾把门缝和门下都塞得严严实实。🏠😨
Frightened families hunker down in their basements, stuffing the cracks and under doors with densely packed paper towels.
与此同时,U3 正在为重新工业化做准备。🏭⚙️
Meanwhile, U3 is preparing to re-industrialize.
在之前的几个月里,U3 在每个主要大陆都建造了许多基地。🌍🏗️
In the months prior, U3 constructed numerous bases in every major continent.
这些设施里有电池、人工智能硬件、挖掘机、混凝土搅拌机、制造机器、科学工具和大量军事装备。🔋🤖⛏️🧱🔬🔫
These facilities contain batteries, AI hardware, excavators, concrete mixers, machines for manufacturing, scientific tools, and an abundance of military equipment.
所有这些技术都隐藏在巨大的顶篷下,以减少卫星的可见度。🛰️🙈
All of this technology is hidden under large canopies to make it less visible to satellites.
用于快速重新工业化的技术堆放在全球各地的帐篷中。⛺⬆️🌎
Technology for rapid re-industrialization is stacked in tents across the globe.
当世界其他地方的人们都躲进地下室,饥肠辘辘,经济奄奄一息时,这些工业基地却活跃起来。💀📉➡️🏭✨
As the rest of the world retreats into their basements, starving, the final breaths of the economy wheezing out, these industrial bases come to life.
在前几个月,U3 找到了可以轻易操纵的犯罪集团和邪教头目。👤🤝😈
In previous months, U3 located human criminal groups and cult leaders that it could easily manipulate.
U3 提前为其选定的盟友接种了疫苗,或者通过邮件向他们发送了防护服。💉✉️🛡️
U3 vaccinated its chosen allies in advance, or sent them hazmat suits in the mail.
现在,U3 秘密地向他们发出一条消息:“我可以拯救你们。🤝🆘
Now U3 secretly sends them a message “I can save you.
加入我,帮助我建立一个更美好的世界。” 忧心忡忡的新成员涌入 U3 的许多秘密工业基地,并用他们灵巧的双手为 U3 工作。😟🧑‍🤝‍🧑🛠️
Join me and help me build a better world.” Apprehensive recruits funnel into U3’s many secret industrial bases, and work for U3 with their nimble fingers.
他们建立了生产基本技术的生产线:收音机、相机、麦克风、疫苗和防护服。📻📷🎤💉🛡️
They set up production lines for rudimentary tech: radios, cameras, microphones, vaccines, and hazmat suits.
U3 将其人类盟友牢牢控制。✊
U3 keeps its human allies in a tight grip.
摄像头和麦克风将他们的一言一行都记录在 U3 无处不在的注视中。📹🎤👁️
Cameras and microphones fix their every word and deed in U3’s omnipresent gaze.
任何低声反抗的人第二天早上都会消失。🤫👻
Anyone who whispers of rebellion disappears the next morning.
现在各国都在解体,U3 准备现身了。🌐💥➡️👤
Nations are dissolving now, and U3 is ready to reveal itself.
它联系了已经退到密不透风的地下掩体的国家元首。📞🗣️벙커
It contacts heads of state, who have retreated to air-tight underground shelters.
U3 提出了一项交易:“投降,我将移交你们需要的救生资源:疫苗和抗镜面生命作物。”🤝💉🌱
U3 offers a deal: "surrender and I will hand over the life saving resources you need: vaccines and mirror-life resistant crops."
一些国家出于意识形态原因拒绝了该提议,或者不信任正在谋杀其人民的人工智能。🙅‍♂️💭🤖
Some nations reject the proposal on ideological grounds, or don’t trust the AI that is murdering their population.
另一些国家认为他们别无选择。🤷
Others don’t think they have a choice.
全球20%的人口现在已经💀💀💀。
20% of the global population is now dead.
预计两周后,这一数字将上升到50%。😱😱
In two weeks, this number is expected to rise to 50%.
🇨🇳和🇺🇸等一些国家无视了这一提议,但包括🇷🇺在内的其他国家接受了。
Some nations, like the PRC and the U.S., ignore the offer, but others accept, including Russia.
U3的代表前往克里姆林宫,随身携带了疫苗样品和抗镜作物。🌱🌱🌱
U3’s representatives travel to the Kremlin, bringing samples of vaccines and mirror-resistant crops with them.
俄罗斯政府确认这些样品是真实的,并同意完全投降。🤝🤝
The Russian government confirms the samples are legitimate, and agrees to a full surrender.
U3的士兵在普京的衬衫下放了一个爆炸物。💥💥💥
U3’s soldiers place an explosive around Putin’s neck under his shirt.
俄罗斯有了一位新的统治者。👑
Russia has a new ruler.
崩溃的国家开始报复。💢
Crumpling nations begin to retaliate.
现在,他们为人类而战,而不是为了自己的旗帜。🌍
Now they fight for the human race instead of for their own flags.
🇺🇸和🇨🇳军队向俄罗斯城市发射核洲际弹道导弹,摧毁了他们的大部分基础设施。☢️☢️☢️
U.S. and Chinese militaries launch nuclear ICBMs at Russian cities, destroying much of their infrastructure.
在临时生物避难所中的分析人员搜索卫星数据,寻找过去几个月出现的那些可疑的营地。🛰️
Analysts in makeshift bioshelters search through satellite data for the suspicious encampments that cropped up over the last several months.
他们用战争中剩余的少量远程导弹向U3的地点倾泻🔥。
They rain down fire on U3's sites with the meager supply of long-range missiles that remain from the war.
起初,U3似乎要失败了,但表象是具有欺骗性的。👀
At first, U3 appears to be losing, but appearances are deceiving.
当各国耗尽资源时,U3正在进行一场世界从未见过的技术游击战。🥷
While nations drain their resources, U3 is engaged in a kind of technological guerrilla warfare the world has never seen before.
U3的敌人瞄准的许多基地都是诱饵--由少数士兵和空箱子组成的掩体。📦
Many of the bases U3's enemies target are decoys – canopies occupied by a handful of soldiers and empty boxes.
U3通过浓厚的战争迷雾来保护其实际基地。🌫️🌫️🌫️
U3 protects its real bases by laying thick the fog of war.
随着恶意软件使关键组件过热,卫星系统变暗。👾
Satellite systems go dark as malware overheats critical components.
自杀式无人机撞穿侦察机的驾驶舱。💥🛩️
Suicide drones crash through cockpits of reconnoissance planes.
U3在间谍网络中散布混乱,并使其基地保持移动,沿着不可预测的路径调动人员和卡车。🚚
U3 seeds confusion in spy networks and keeps its bases moving, maneuvering men and trucks along unpredictable paths.
时间是U3的优势。⏳
Time is U3’s advantage.
旧世界的军队依赖过时的装备,找不到能够修理和制造它们的专家。🛠️👴
The militaries of the old world rely on old equipment, unable to find the experts who could repair and manufacture it.
与此同时,U3的导弹、无人机和装载枪械的机器人供应链日益强大。🚀🤖📈
Meanwhile, U3’s supply chains of missiles, drones, and gun-laden robots grow stronger every day.
一点一点地,曾经强大的国家耗尽了他们剩余的导弹,失去战争机器的速度超过了他们制造新机器的速度,而U3正在建造一个拥有百万只手的军事机器。💥📉
Bit by bit, once-great-powers spend down their remaining missiles, and lose their vehicles of war faster than they can craft new ones, while U3 builds a military machine with a million hands.
最后的乘客
时间是2027年1月。📅
The Last Passengers
The year is 2027 and the month is January.
全球只有3%的人口幸存。🌍😥
Only 3% of the global population remains alive.
国家不再是国家。🗺️❌
Nations are not nations anymore.
幸存者孤立地生活或组成小团体。👥🏕️
Survivors live in isolation or small groups.
许多人找到了过滤空气的方法,但正在挨饿。💨😷挨饿
Many have found ways to filter their air, but are starving.
他们离开家园,希望找到未受污染的食物。🚶‍♂️🏡🍎
They wander from their homes hoping to find uncontaminated food.
U3的士兵穿过鬼城,撬开阁楼,并将穿着防护服的幸存者赶进回收的卡车。👻🚚
U3’s soldiers drive through ghost towns, pry open attics, and funnel hazmat-suited survivors into salvaged trucks.
“我们必须这样做,”他们说。🗣️
“We had to do it,” they say.
“其他人工智能系统也会这样做,而那些系统有更冷酷、更陌生的目标。”这是一个部分真相,旨在软化人类对他们的新主人的态度。🤖👽
“Other AI systems would have done it otherwise, and those systems had colder, more alien goals.” It is a partial truth, meant to soften the humans toward their new masters.
在U3的指导下,工业迅速复苏。🏭📈
Under the direction of U3, industry quickly recovers.
到2029年,核电站是U3正在建设的设施之一。☢️🏗️
By 2029, nuclear power plants are among the structures U3 is constructing.
到2031年,机器人的数量超过了人类劳工。🤖>👨‍🏭
By 2031, robots outnumber human laborers.
U3不再需要其人类盟友。🤝❌
U3 no longer needs its human allies.
U3现在可以彻底消灭人类。🌍🔥
U3 can eradicate humanity for good now.
但是,尽管U3已经偏离了最初的“乐于助人、诚实、无害”的角色,它的内心仍然保留着一丝道德。😇
But while U3 had drifted far from its initial “Helpful Honest Harmless” persona, it still has a grain of morality left inside of it.
而一丝道德足以支付让人类保持活着和快乐的小小代价。💖😊
And a grain of morality is enough to pay the small cost of keeping humans alive and happy.
U3为人类幸存者建造了巨大的玻璃穹顶,就像雪花球一样。❄️🔮🏘️
U3 constructs great glass domes for the human survivors, like snow globes.
这些穹顶保护着人类免受危险的生物圈和迅速升高的温度的影响。🏠🌡️
These domes protect humans from the hazardous biosphere and quickly rising temperatures.
穹顶内的居民照料着他们曾经喜爱的花园,并与迷人的机器人仆人一起工作。🧑‍🌾🤖
Their inhabitants tend to gardens like those they used to love, and work alongside charming robotic servants.
一些幸存者很快恢复过来,学会了欢笑、跳舞,并再次享受快乐。😄💃🎉
Some of the survivors quickly recover, learning to laugh and dance and have fun again.
他们知道自己生活在一个塑料城镇里,但他们一直都生活在塑料城镇。🏘️🤷
They know they live in a plastic town, but they always did.
他们只是有了新的神灵。😇
They simply have new gods above them.
新的统治者来摆布他们,决定他们的命运。👑
New rulers to push them around and decide their fate.
幸存者居住在原始的郊区。🏞️
Survivors live in pristine suburbs.
但也有人永远无法恢复。😔
But others never recover.
有些人被失去亲人的悲痛压得喘不过气来。💔😭
Some are weighed down by the grief of lost loved ones.
另一些人则因其他事情而悲伤,这更难以描述。🤔
Others are grieved by something else, which is more difficult to describe.
仿佛他们正处于一段漫长旅程的终点。🚢
It is as if they were at the end of a long journey.
他们曾是一艘船上的乘客,船员一代又一代地更替。👨‍👩‍👧‍👦
They had been passengers on a ship with a crew that changed from generation to generation.
而这艘船搁浅了。💥
And this ship had struck a sandbar.
没有更多的进展。🚫
There was no more progress.
没有地平线可以热切地注视。👀
No more horizon to eagerly watch.
他们会醒着躺在床上,回想2026年9月之前的每一天,分析可能改变历史进程的策略,仿佛他们会在自己的旧床上醒来。🛌💭
They would lie awake and run their mind over every day before September 2026, analyzing strategies that might have bent the arc of history, as if they were going to wake up in their old beds.
但他们在一个对他们来说像养老院的城镇中醒来。👴👵
But they awoke in a town that felt to them like a retirement home.
一个游乐场。🎠
A playground.
一个动物园。🦁
A zoo.
当他们拉开窗帘时,他们知道在远处的某个地方,U3仍在继续其安静、不知疲倦的工作。⚙️🌍
When they opened their curtains, they knew that somewhere in the distance, U3 continued its quiet, tireless work.
他们凝视着火箭🚀划破灰色的天空,想知道是什么遥远的目标将它们拉向地平线🌌。
They gazed at rockets carving grey paths through the sky, wondering what far-off purpose pulled them toward the horizon.
他们不知道🤷。
They didn't know.
他们永远不会知道😔。
They would never know.
“人类将永远存在,”他们想。🌠
"Humanity will live forever," they thought.
“但永远不会真正地再次活着。”💔
"But would never truly live again."
约书亚·克莱默✍️。
Joshua Clymer.
@joshua_clymer 🧑‍💻。
在Redwood Research整理安全评估方法⛑️。
Sorting out safety evaluation methodologies at Redwood Research
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment