下一个瓶颈：意识与体验的超人类主义

Next Bottleneck: Transhumanism of Consciousness and Experience

一

I

任何一个由多个环节串联而成的系统，它的整体产出都不取决于最强的环节，而取决于最弱的那一个。这几乎是一个不需要证明的道理：水流过一串管道，流量由最细的那一截决定；一件事情经过若干道工序，完成的速度由最慢的那道工序决定。把力气花在非瓶颈的环节上，提升的只是局部的数字，整体纹丝不动。

In any system made of links joined in series, total output is determined not by the strongest link but by the weakest. This hardly needs proving: the flow of water through a run of pipes is set by its narrowest section; a job that passes through several stages finishes at the pace of the slowest stage. Effort spent on links that are not the bottleneck improves only local numbers; the whole does not move.

由此可以推出一个更有意思的结论：瓶颈是会迁移的。当最弱的环节被加强，它就不再是最弱的，压力立刻转移到下一个最弱的地方。被突破的瓶颈并没有消失，它只是把位置让给了下一个。只要系统还在加速，这个过程就会沿着链条一路推进，直到抵达某个无法被轻易加强的环节为止。一个持续加速的系统，它的历史在相当程度上就是瓶颈一路迁移的历史。

From this follows a more interesting conclusion: bottlenecks migrate. When the weakest link is strengthened, it is no longer the weakest, and pressure shifts at once to the next weakest point. A bottleneck that has been broken through does not disappear; it simply yields its place to the next one. As long as the system keeps accelerating, the process advances along the chain until it reaches a link that cannot easily be strengthened. The history of a continually accelerating system is, to a considerable degree, the history of its bottleneck on the move.

这个结构还有一个容易被忽视的特征：瓶颈不是被发现的，而是被暴露的。一个环节在成为瓶颈之前，往往根本不被视为问题，因为它“从来都是那样”。它的慢、它的窄、它的局限，被所有人当作背景条件默默接受。只有当它周围的环节加速了，它的局限才第一次变得刺眼。每一次瓶颈迁移，都会让此前理所当然的东西突然变成问题。

The structure has one further feature that is easy to overlook: bottlenecks are not discovered, they are exposed. Before a link becomes the bottleneck it is usually not regarded as a problem at all, because it has "always been like that." Its slowness, its narrowness, its limits are silently accepted by everyone as background conditions. Only when the links around it speed up do its limits become glaring for the first time. Every migration of the bottleneck turns something previously taken for granted into a problem.

今天我们正处在一轮规模空前的瓶颈迁移之中。过去十年，关于人工智能的讨论几乎都在追踪这条链条的上游：先是算力不够，然后是数据不够，然后是电力不够，然后是芯片产能、冷却、电网、甚至发电厂的选址。每一个环节都被逐一点名、逐一投资、逐一突破。这些讨论都很重要，但它们共享一个方向：都在问“怎样生产出更多的智能”。

We are now in the midst of a bottleneck migration on an unprecedented scale. Over the past decade, discussion of artificial intelligence has almost entirely tracked the upstream end of the chain: first there was not enough compute, then not enough data, then not enough electricity, then chip capacity, cooling, the grid, even the siting of power plants. Each link has been named, funded and broken through in turn. These discussions matter, but they share a single direction: they all ask how to produce more intelligence.

很少有人问另一个方向的问题：这条链条的下游终点在哪里？智能被源源不断地生产出来，它最终流向什么地方，被什么东西接住？如果上游的瓶颈被一个接一个地打通，压力迟早会传导到链条的下游，传到接住这一切的那一端。那一端，那个此前从来不被视为问题、因为它“从来都是那样”的环节，就是人本身。更准确地说，是人作为体验者的能力。

Few people ask the question that points the other way: where does this chain end downstream? Intelligence is being produced in an endless stream; where does it finally flow, and what catches it? If the upstream bottlenecks are cleared one after another, pressure will sooner or later be transmitted downstream, to the end that receives it all. That end, the link never regarded as a problem because it has "always been like that," is the human being. More precisely, it is the human capacity to experience.

这就是本文要讨论的东西：下一个瓶颈。

This is what this essay is about: the next bottleneck.

二

II

先确认一个前提：一切价值链的终点，都是某个主体的体验。

First, a premise: the end point of every value chain is the experience of some subject.

这句话听上去抽象，其实很朴素。一座桥的价值，最终落在有人从桥上走过、省下了时间、见到了想见的人；一款药的价值，最终落在某个人的痛苦减轻了；一部电影、一顿饭、一套房子、一份工作、一个科学发现，追到最后，都要落进某个意识里，被某个人经历到。即使是那些看起来纯粹工具性的东西，比如一台服务器、一条供应链，它们之所以有价值，也只是因为它们最终服务于某些人的经历。如果宇宙里没有任何一个能体验的主体，就不会有任何东西是“好”的或“坏”的，只会有物质在运动。

This sounds abstract but is really quite plain. The value of a bridge finally lands on someone crossing it, saving time, reaching the person they wanted to see; the value of a medicine finally lands on someone's pain being eased. A film, a meal, a house, a job, a scientific discovery: followed to the end, each must fall into some consciousness and be lived through by someone. Even things that look purely instrumental, a server or a supply chain, have value only because they ultimately serve someone's experience. If the universe contained no subject capable of experience, nothing would be good or bad; there would only be matter in motion.

所以，当我们说AI正在让智能变得充裕时，我们其实在说：链条上游的产出能力正在暴涨。那么问题就来了：下游那个接收端，它的容量是多少？

So when we say AI is making intelligence abundant, what we are really saying is that output capacity at the upstream end of the chain is surging. The question then becomes: what is the capacity of the receiving end downstream?

答案令人不安。人类的体验能力，在过去几万年里几乎没有变化。我们拥有的高级情绪，屈指可数：敬畏、爱、审美愉悦、求知的喜悦、悲悯、怀旧、崇高感、幽默，以及它们之间有限的组合。我们的注意力一次只能深度处理极少的东西，我们的时间感由一套固定的神经节律锁死，我们的记忆以一种我们无法控制的方式保存和消退。这套体验结构是为更新世的小群体生活而演化出来的：它擅长感受饥饿与饱足、危险与安全、接纳与排斥，它的上限，大致就是一个狩猎采集者在篝火旁听一个好故事时所能达到的深度。

The answer is unsettling. Human experiential capacity has barely changed in tens of thousands of years. Our higher emotions can be counted on two hands: awe, love, aesthetic pleasure, the joy of understanding, compassion, nostalgia, the sublime, humour, and a limited set of combinations among them. Our attention can deeply process only a very few things at a time; our sense of time is locked to a fixed set of neural rhythms; our memories are kept and fade in ways we cannot control. This experiential structure evolved for small-group life in the Pleistocene. It is good at registering hunger and satiety, danger and safety, acceptance and rejection, and its ceiling is roughly the depth a hunter-gatherer could reach listening to a good story by the fire.

在过去，这不构成问题，因为世界的复杂度与人的体验能力大致匹配。人类创造的一切文明成果，本来就是由人做出来、为人做出来的，它们天然被裁剪成人类能够消化的尺寸。最宏大的大教堂，也是按照人能仰望的角度建造的；最复杂的交响乐，也是在人的听觉所能分辨的音域和节奏里写成的。生产者和体验者是同一种生物，所以产出永远不会超出接收能力太多。

In the past this was no problem, because the complexity of the world roughly matched human experiential capacity. Everything human civilization created was made by humans and for humans, and so was naturally cut to a size humans could digest. The grandest cathedral was built at an angle a person could gaze up at; the most complex symphony was written within the range and rhythms the human ear can distinguish. Producer and experiencer were the same species, so output never exceeded receiving capacity by much.

这个匹配正在被打破。当AI开始以指数级的速度提升客观世界的复杂度，它创造出的东西不再需要被裁剪成人类的尺寸。如果觉得这个描述难以理解的话，可以想象这样一幅图景：AI能够写出宏大的交响乐，而人类的感知能力还停留在狗的水平。

That match is now breaking. As AI begins to raise the complexity of the objective world at an exponential rate, what it creates no longer needs to be cut to human size. If this description is hard to grasp, picture this: AI can write grand symphonies, while human perception remains at the level of a dog.

这个比喻值得多停留一会儿。一只狗坐在音乐厅里，它听得到声音，甚至可能对某些音高或节奏产生反应。但它不会因为和声的解决而感到释然，不会在主题再现时认出那是开头那个旋律，不会感受到一个乐章的结构如何在时间中展开。这不是因为狗不够努力，也不是因为它不够聪明。是它根本不具备欣赏和声、结构与主题发展的那套感知结构。对它来说，交响乐就只是声音。

The metaphor deserves a moment longer. A dog sitting in a concert hall hears the sound and may even react to certain pitches or rhythms. But it will not feel relief when a harmony resolves, will not recognize the opening melody when the theme returns, will not sense how the structure of a movement unfolds in time. This is not because the dog is not trying hard enough, nor because it is not clever enough. It simply lacks the perceptual structure needed to appreciate harmony, form and thematic development. To the dog, the symphony is just sound.

这幅图景并不完全是想象。2026年10月6日，OpenAI在GitHub上一次性公开了722篇数学手稿。它们出自一个尚未发布的内部前沿模型，涵盖372个结果族，横跨数论、几何、分析、理论计算机科学与数学物理，其中不乏触及著名猜想的结果。为了得到这些结果，模型被投放了大约四千个问题，平均每个结果所耗费的算力，大约相当于几个小时的机器思考。

This picture is not entirely imaginary. On 6 October 2026, OpenAI published 722 mathematical manuscripts on GitHub in a single release. They came from an unreleased internal frontier model and covered 372 families of results across number theory, geometry, analysis, theoretical computer science and mathematical physics, several touching on well-known conjectures. To obtain them, the model had been set roughly four thousand problems; the average result consumed compute equivalent to a few hours of machine thinking.

随后几天，数学界的反应近乎一场地震。牛津大学的厄休拉·马丁批评说，这是把一份凌乱的初稿扔给整个学界，等着别人来收拾；哥伦比亚大学的数论学家迈克尔·哈里斯干脆表示，自己不会去读这些论文；陶哲轩一方面承认其中可能有真正有价值的想法，另一方面对缺少人类可读的阐释深感沮丧。菲尔兹奖得主迪米尼尔-科潘的反应最为直白：他说冲击巨大，自己几乎瘫痪了。也有数学家站在另一端，认为无论如何评价，这都是数学史上最重要的时刻。

In the days that followed, the reaction among mathematicians amounted to an earthquake. Ursula Martin of Oxford complained that the community had been handed a messy first draft and left to clean it up; Michael Harris, a number theorist at Columbia, said flatly that he would not read the papers; Terence Tao conceded that some may contain genuinely valuable ideas, while expressing deep frustration at the lack of human-readable exposition. The Fields medallist Hugo Duminil-Copin was the most direct: he said the shock was immense and that he felt almost paralysed. Others stood at the opposite pole, holding that, however one judged it, this was the most significant moment in the history of mathematics.

这些争论底下有一个共同的事实：产出第一次彻底压过了接收。722篇手稿中，只有大约162篇的主要结果经过了Lean形式化验证，其余的都需要人去读、去核对、去理解；发布第二天，就有三篇因为一个符号错误被撤回。按照正常的学术节奏，整个数学界都不可能在短时间内消化这样体量的专门论文。如今在OpenAI工作的年轻数学家萨瓦尼对此看得很清楚：人类数学家的主要瓶颈，正在从生产证明转向吸收和理解AI产出的数学。一周之后，OpenAI拿出五百万美元，资助西蒙斯基金会的一项新计划，专门帮助数学家审视、理解并继承AI产出的结果。为了让人类读懂机器的证明而专门设立资助，这件事本身就说明了瓶颈落在了哪里。

Beneath these arguments lies a common fact: for the first time, output has decisively overwhelmed reception. Of the 722 manuscripts, only about 162 had their main results formally verified in Lean; the rest must be read, checked and understood by people. The day after the release, three were withdrawn because of a single sign error. At the normal pace of scholarship, the mathematical community could not possibly digest so many specialized papers in a short time. Mehtaab Sawhney, a young mathematician now working at OpenAI, saw this clearly: the main bottleneck for human mathematicians is shifting from producing proofs to absorbing and understanding mathematics produced by AI. A week later, OpenAI committed five million dollars to a new Simons Foundation initiative dedicated to helping mathematicians examine, understand and build on results produced by AI. That funding had to be set up specifically so that humans could read machine proofs says, by itself, where the bottleneck now lies.

更值得注意的是，这件事发生在数学。数学向来被认为是人类心智最接近纯粹结构的地方，从事它的也是人类中最擅长感受抽象结构的那一群人。如果连他们都跟不上，这个落差就不是努力程度的问题，而是结构性的问题。知道一个命题成立，和理解它为什么成立、感受到它美在哪里，是两回事，而后者恰恰是数学家几百年来真正珍视的东西。音乐厅里的那只狗，在这里已经换成了人类之中最好的听众。

More striking still is that this happened in mathematics. Mathematics has long been regarded as the place where the human mind comes closest to pure structure, and those who practise it are the people best able to feel abstract structure. If even they cannot keep up, the gap is not a matter of effort but a structural one. Knowing that a proposition is true is one thing; understanding why it is true, and sensing where its beauty lies, is another, and the latter is exactly what mathematicians have prized for centuries. The dog in the concert hall has here been replaced by the finest listeners humanity has.

如果未来人类与AI所构建的世界之间的关系变成这样，那么真正的危机就不是人类会不会被淘汰、会不会失业，而是一种更隐蔽也更深的东西：人类即使活着、安全、物质极大丰富，也会被排除在意义之外。世界越来越丰盛，人却只能感受到其中最浅的一层。这是一种存在意义上的贫困，而且它可以和物质的富足同时发生。

If the relationship between humanity and the world built by AI comes to look like this, the real crisis will not be whether humans are made obsolete or lose their jobs. It will be something more hidden and deeper: even alive, safe and materially rich beyond measure, human beings will be shut out of meaning. The world grows ever richer, while people can feel only its shallowest layer. This is poverty in the existential sense, and it can coexist with material abundance.

三

III

有人会说：人的体验能力并不是固定的，它一直在随着文明一起成长。这个说法是对的，而且恰恰是理解本文论点的关键。

Someone will object that human experiential capacity is not fixed, that it has always grown together with civilization. That is true, and it is precisely the key to this essay's argument.

举一个具体的例子。在十八世纪之前，欧洲人普遍不觉得高山是美的。阿尔卑斯山在旅行者的记述里，常常是险恶、荒凉、令人厌恶的障碍，是需要尽快穿越的地方。十七世纪末，托马斯·伯内特在他那本影响很大的地球理论著作里，把山脉描述为大洪水之后世界的废墟与残骸。一百年后，事情完全变了。1757年，伯克出版了关于崇高与美的哲学探讨；随后康德把崇高写进了他的判断力批判；浪漫主义诗人成群结队地登山，在悬崖与冰川面前体验到一种混合着恐惧、敬畏与升华的强烈情感。美国学者马乔丽·尼科尔森后来专门写了一本书，追溯这场从“山岳的阴郁”到“山岳的荣光”的转变。

Take a concrete example. Before the eighteenth century, Europeans generally did not find high mountains beautiful. In travellers' accounts the Alps were often a menacing, desolate, repellent obstacle, a place to be crossed as quickly as possible. In the late seventeenth century Thomas Burnet, in his influential theory of the earth, described mountain ranges as the ruins and wreckage of the world after the Flood. A hundred years later everything had changed. In 1757 Burke published his philosophical enquiry into the sublime and the beautiful; Kant then wrote the sublime into his critique of judgment; Romantic poets climbed mountains in droves and, before cliffs and glaciers, experienced an intense emotion mixing fear, awe and elevation. The American scholar Marjorie Hope Nicolson later devoted a whole book to tracing this shift from "mountain gloom" to "mountain glory."

山没有变，变的是人看山的方式。更准确地说，是人的体验结构里多出了一个新的维度：“崇高”这种情感，从一个模糊的可能性，变成了一个可以被稳定唤起、被命名、被传授、被追求的体验形态。

The mountains did not change; the way people saw them did. More precisely, a new dimension was added to human experiential structure: the sublime went from a vague possibility to a form of experience that could be reliably evoked, named, taught and pursued.

类似的事情在艺术史上反复发生。1913年，斯特拉文斯基的《春之祭》在巴黎首演，引发了著名的骚乱；今天它是每个交响乐团的常备曲目，听众会为它的节奏张力而兴奋。并不是乐谱变了，而是听众的耳朵被训练出来了。阅读也是一样：认知神经科学的研究表明，学会读写会在相当程度上重塑大脑的视觉加工区域。我们以为是自然而然的阅读体验，其实是一项文化技术对神经结构的改写。

Similar things have happened again and again in the history of art. In 1913 the Paris premiere of Stravinsky's The Rite of Spring provoked a famous riot; today it is a staple of every symphony orchestra, and audiences thrill to its rhythmic tension. It was not the score that changed but the listeners' ears that were trained. Reading is the same: cognitive neuroscience has shown that learning to read substantially reshapes the brain's visual processing areas. The experience of reading, which we take to be natural, is in fact a cultural technology rewriting neural structure.

这些例子说明了两件事。

These examples show two things.

第一，体验结构是可塑的。人能感受什么、怎样感受，并不完全由出厂设置决定，它可以被扩展，可以长出新的维度。这是整个构想的可能性基础。

First, experiential structure is plastic. What people can feel, and how, is not wholly fixed by factory settings; it can be expanded and can grow new dimensions. This is the ground of possibility for the entire vision.

第二，到目前为止，这种扩展的方式极其原始。它依靠文化的缓慢积累，依靠少数敏感者的偶然开拓，依靠一代人又一代人的潜移默化。从伯内特到浪漫主义，花了一百年，而它所扩展的，也只是在原有神经硬件允许的范围内重新排列组合。文化可以教会人欣赏山岳，但没法让人同时听懂一千个声部，没法让人感受到一种进化从未准备过的情感。

Second, the means of such expansion have so far been extremely primitive. They rely on the slow accumulation of culture, on the chance explorations of a sensitive few, on gradual influence passed from generation to generation. Getting from Burnet to Romanticism took a hundred years, and what it expanded was only a rearrangement within the limits of the existing neural hardware. Culture can teach people to appreciate mountains, but it cannot let them follow a thousand voices at once, or feel an emotion evolution never prepared them for.

换句话说，人类一直在做体验者的升级，只是不知道自己在做这件事，也没有任何工程手段。它像是工业革命之前的手工纺纱：确实在生产，但产能被身体牢牢限定。当上游的产出开始指数级增长，这种手工式的扩展就再也跟不上了。

In other words, humanity has always been upgrading the experiencer, only without knowing it and without any engineering means. It resembles hand-spinning before the industrial revolution: real production, but with capacity firmly bounded by the body. Once upstream output begins to grow exponentially, this handcraft kind of expansion can no longer keep pace.

这正是问题被逼到台前的原因。如果世界的复杂度增长以年计，而人的体验能力扩展以世纪计，那么两者之间的落差只会越来越大。要缩小这个落差，靠的不能再是等待，而必须是一种自觉的、有方向的、可以设计的升级。

That is exactly why the question is being forced into the open. If the complexity of the world grows on a scale of years while human experiential capacity expands on a scale of centuries, the gap between them can only widen. Narrowing it can no longer rely on waiting; it must be a deliberate, directed, designable upgrade.

四

IV

说到“升级人类”，大多数人首先想到的是超人类主义的那一套经典图景：延长寿命，逆转衰老，用义肢和外骨骼增强体能，用基因编辑消除疾病，用脑机接口提升记忆容量和运算速度，乃至最终把意识上传到硅基载体上。这些设想在过去几十年里构成了超人类主义的主流叙事，它们的共同关键词是：更长、更强、更快、更多。

Mention "upgrading humans" and most people first think of transhumanism's classic picture: extending life, reversing ageing, enhancing the body with prosthetics and exoskeletons, eliminating disease with gene editing, boosting memory and processing speed with brain–computer interfaces, and finally uploading consciousness to a silicon substrate. For decades these ideas have made up transhumanism's mainstream narrative, and their shared keywords are: longer, stronger, faster, more.

这些都有价值，但它们共享一个很少被拿出来审视的隐含前提：体验者本身不变。

All of this has value, but it shares an implicit premise that is rarely examined: the experiencer itself stays the same.

一个活到三百岁的人，如果他的体验结构和今天一样，那么他得到的是三百年的同一种生活：同样会无聊，同样会焦虑，同样会嫉妒，同样会在中年时感到意义的流失，同样会被少数几种情绪反复支配。寿命被延长了，可是被延长的那段时间里装着什么并没有改变。更强壮的身体、更快的计算速度，也是同样的道理：它们是更好的工具，但使用工具的那个“我”，还是原来那个“我”。

A person who lives to three hundred, with the same experiential structure as today, gets three hundred years of the same life: the same boredom, the same anxiety, the same envy, the same midlife drain of meaning, the same domination by a handful of emotions. Life has been extended, but what fills the extended time has not changed. The same goes for a stronger body or faster computation: they are better tools, but the "I" who uses the tools is still the same "I."

用一句话概括：更长的寿命，同一种无聊。

In a phrase: a longer life, the same boredom.

本文所说的超人类主义，重心在另一处。它不关心人能跑多快、活多久、力气多大，它关心的是人能感受什么、怎样感受，以及从中获得什么意义。如果说传统超人类主义升级的是体验者手里的工具，那么这里要升级的是体验者本身。

The transhumanism of this essay has its centre of gravity elsewhere. It is not concerned with how fast people can run, how long they live or how strong they are; it is concerned with what people can feel, how they feel it, and what meaning they draw from it. If traditional transhumanism upgrades the tools in the experiencer's hands, what is to be upgraded here is the experiencer itself.

这和我前一段时间所作的《现象学工程》有相通之处。所谓现象学工程，是把经验的结构当作可以设计、制造和验证的工程对象，它关心的是如何为一个体验者构造一段有结构、有纵深的经验。本文的问题可以看作它的另一面：如果说现象学工程问的是“给体验者什么样的体验”，那么体验者升级问的是“体验者能够承接什么样的体验”。前者设计的是输入，后者设计的是接收端。两者迟早会在同一张桌子上相遇，因为一段再精妙的经验，如果超出了接收者的结构，就和那场狗听到的交响乐没有区别。

This connects with my earlier essay, Phenomenological Engineering. Phenomenological engineering treats the structure of experience as an engineering object that can be designed, manufactured and verified; it is concerned with how to construct, for an experiencer, a stretch of experience with structure and depth. The question of this essay can be seen as its other side: if phenomenological engineering asks what experience to give the experiencer, upgrading the experiencer asks what experience the experiencer is able to receive. The former designs the input; the latter designs the receiving end. The two will sooner or later meet at the same table, because an experience, however exquisite, that exceeds the structure of its receiver is no different from the symphony the dog heard.

为什么说这件事应该在2026年开始被严肃讨论？因为几年前，它还只是一种遐想：AI的能力远没有逼近人类顶尖水平，世界的复杂度也还没有显著地溢出人的接收能力。今天情况不同了。当AI在数学、科学、写作和设计等领域接连逼近甚至越过人类的上限，“产出过剩而体验不足”的落差，第一次从思想实验变成了可以看到轮廓的现实。瓶颈已经开始暴露，而按照历史的经验，一个瓶颈从暴露到成为压倒一切的问题，往往比人们预想的要快。

Why should this be discussed seriously in 2026? Because a few years ago it was only a daydream: AI was nowhere near top human level, and the complexity of the world had not yet noticeably overflowed human receiving capacity. Things are different now. As AI approaches and even surpasses the human ceiling in mathematics, science, writing and design, the gap between surplus output and insufficient experience has, for the first time, moved from thought experiment to a reality whose outline can be seen. The bottleneck has begun to be exposed, and historical experience suggests that the time from a bottleneck's exposure to its becoming the overriding problem is often shorter than people expect.

接下来，本文把体验者升级的愿景分成三个层次来讨论：功能层，处理体验的基础设施；激励层，处理体验被什么驱动；存在层，处理体验最终为了什么。三者是递进的：前一层是后一层的前提，后一层是前一层的目的。

What follows divides the vision of upgrading the experiencer into three layers: the functional layer, which deals with the infrastructure of experience; the motivational layer, which deals with what drives experience; and the existential layer, which deals with what experience is ultimately for. They are cumulative: each layer is the precondition of the next, and each next layer is the purpose of the one before.

五

V

功能层处理的是意识运作的底层机制。如果把体验者比作一台接收设备，这一层改的就是它的基础规格：时间、焦点、处理、存储。

The functional layer deals with the underlying mechanics of consciousness. If the experiencer is thought of as a receiving device, this layer alters its basic specifications: time, focus, processing, storage.

时间流逝感知。 主观时间从来不是均匀的。车祸发生的那一秒会被拉得很长，深度沉浸于一项工作时几个小时会凭空消失，而等待一个消息时十分钟像一整个下午。心理学家契克森米哈赖在研究“心流”时发现，时间感的扭曲是这种状态最稳定的特征之一。胡塞尔则早在一百多年前就指出，当下的经验总是带着刚刚过去的余韵和对下一刻的预期，时间感正是这样被一层层织出来的。

The perception of time passing. Subjective time has never been uniform. The second of a car crash stretches out; hours vanish in deep absorption in work; ten minutes waiting for a message feel like a whole afternoon. The psychologist Mihaly Csikszentmihalyi, studying flow, found that distorted time is among the most stable features of that state. More than a century ago Husserl had already observed that present experience always carries the trailing resonance of what has just passed and the anticipation of the next moment, and that the sense of time is woven layer by layer in just this way.

这意味着，时间感是一种结构。如果这种结构可以被主动调节，人就能在珍贵的时刻把时间拉长，让一次日落、一段谈话、一次突破性的理解获得更大的主观纵深；也能在不可避免的痛苦与枯燥中把时间压缩。这比单纯的延寿更直接：延寿增加的是客观年数，而时间感知的改造增加的是“活着的感受”本身。我们可以称之为主观寿命。一个人客观上活八十年，主观上却可能活出远比这长得多、也远比这密的一生。

This means that the sense of time is a structure. If that structure could be actively adjusted, people could stretch time in precious moments, giving a sunset, a conversation, a breakthrough of understanding greater subjective depth, and compress it through unavoidable pain and tedium. This is more direct than simple life extension: life extension adds objective years, while altering time perception adds to the feeling of being alive itself. Call it subjective lifespan. A person who lives eighty years objectively might, subjectively, live a life far longer and far denser.

注意力。 威廉·詹姆斯在1890年写道，他的经验就是他同意去注意的东西。这句话道出了注意力的根本地位：它是体验的入口，你注意到什么，就体验到什么。而人类的注意力有两个结构性缺陷。一是太窄，一次只能深度处理极少的对象，丰富的世界只能被一小块一小块地看见。二是太容易被劫持，恐惧、焦虑、新奇刺激和精心设计的推荐算法，都能轻易地把它从人真正想关注的东西上拉走。今天的注意力经济本身就是一场针对这第二个缺陷的大规模开采。

Attention. William James wrote in 1890 that his experience was what he agreed to attend to. The remark captures the fundamental place of attention: it is the gateway to experience; what you attend to is what you experience. Human attention has two structural defects. First, it is too narrow: it can deeply process only very few objects at once, so the rich world can be seen only one small patch at a time. Second, it is too easily hijacked: fear, anxiety, novelty and carefully designed recommendation algorithms can all pull it away from what a person really wants to attend to. Today's attention economy is itself a large-scale extraction operation aimed at this second defect.

注意力的改造因此有两个方向：一是主权，让人能够真正决定意识的焦点落在哪里，不再被默认的警报系统和外部的诱饵牵着走；二是带宽，让人能够同时在多个层面保持清醒，像一位指挥家同时听见整支乐队。

Altering attention therefore has two directions. One is sovereignty: letting people truly decide where the focus of consciousness falls, no longer led by the default alarm system and external lures. The other is bandwidth: letting people stay awake on several levels at once, like a conductor hearing the whole orchestra at the same time.

融合AI作为潜意识与逻辑分析机制。 人脑本来就不是一台只靠意识运作的机器。我们的大部分认知是在意识之外完成的：模式识别、直觉判断、情绪预警、语言的自动生成，这些后台进程在我们察觉之前就已经把结果递交上来。卡尼曼所说的快思考，就是这套后台系统的一个侧面。意识更像是一个前台，接收、整合、选择，然后体验。

Integrating AI as a subconscious and analytical mechanism. The human brain was never a machine that runs on consciousness alone. Most of our cognition takes place outside awareness: pattern recognition, intuitive judgment, emotional warning, the automatic generation of speech. These background processes deliver their results before we notice. What Kahneman called fast thinking is one face of this background system. Consciousness is more like a front desk: it receives, integrates, chooses, and then experiences.

本文设想的是让AI成为这套后台系统的一部分。它承担复杂的推理与分析，然后把结果以直觉、理解或洞见的形式交给意识层。人不需要一步步地推演一个复杂系统，就能“感觉到”它的结构；不需要逐条核对证据，就能对一个判断产生有根据的确信。这样，意识得以从繁重的计算中解放出来，专注于它最擅长、也最不可替代的事情：体验与选择。人和AI的关系也由此改变，从“使用一个外部工具”变成“共同构成同一个心智”。

What this essay envisions is making AI part of that background system. It would take on complex reasoning and analysis and hand the results to consciousness in the form of intuition, understanding or insight. One would not need to work through a complex system step by step to feel its structure, nor check evidence item by item to arrive at a well-founded conviction. Consciousness would thus be freed from heavy computation to concentrate on what it does best and what is least replaceable: experiencing and choosing. The relationship between human and AI changes accordingly, from using an external tool to jointly constituting a single mind.

记忆的精细化管理。 记忆决定了自我的连续性，也决定了过去以何种方式持续影响现在。神经科学在二十一世纪初的一个重要发现是记忆的再巩固：一段记忆在被提取时，会重新进入一种可塑的状态，然后被重新存储。也就是说，记忆从来不是一份被封存的档案，而是每次回忆都在被重写的文本。

Fine-grained management of memory. Memory determines the continuity of the self, and how the past goes on shaping the present. An important discovery of early twenty-first-century neuroscience is memory reconsolidation: when a memory is retrieved, it re-enters a malleable state and is then stored again. Memory, in other words, has never been a sealed archive; it is a text rewritten every time it is recalled.

既然它本来就在被重写，问题就只是由谁来写、以什么方式写。精细化的记忆管理，可以是让珍贵的经历保持鲜活，不被时间磨损成几个模糊的标签；可以是削弱创伤记忆的情绪强度，同时保留它的认知内容；也可以是有选择地重温、重组和重新诠释过去的经验。这等于给了人对自己生命叙事的编辑权。

Since it is being rewritten anyway, the only questions are who writes it and how. Fine-grained memory management could mean keeping precious experiences vivid rather than letting time wear them down into a few blurry labels; it could mean weakening the emotional intensity of traumatic memories while keeping their cognitive content; it could mean selectively revisiting, reorganizing and reinterpreting past experience. This amounts to giving people editorial rights over the narrative of their own lives.

需要说明的是，这四项并不是功能层改造的全部设想，只是众多可能方向中容易想到的一小部分。感官的种类与维度、情绪的颗粒度、自我边界的伸缩、多重意识流的并行，都可能成为同一层面上的改造对象。无论具体形态如何，它们本身都是后两层得以成立的基础设施。

It should be said that these four items are not the whole of what the functional layer might include; they are only a small part of the many possible directions, the ones that come to mind most readily. The kinds and dimensions of the senses, the granularity of emotion, the expansion and contraction of the boundaries of the self, the parallel running of multiple streams of consciousness, all could become objects of modification at the same level. Whatever their specific forms, they are all infrastructure on which the next two layers stand.

六

VI

如果说功能层改的是硬件规格，激励层改的就是目标函数。

If the functional layer alters hardware specifications, the motivational layer alters the objective function.

训练一个强化学习系统时，工程师最先要确定的东西叫奖励函数：系统在什么情况下得到正反馈，在什么情况下得到负反馈。这个函数一旦确定，系统的全部行为，包括它所有看似聪明的策略和所有令人费解的怪癖，都会围绕它展开。人类也有这样一个函数，只是它不是由我们自己写的，而是由几百万年的自然选择写好的。我们在什么时候感到满足，在什么时候感到痛苦，被什么吸引，被什么驱赶，这些几乎全部是出厂设置。

When training a reinforcement learning system, the first thing engineers must fix is the reward function: under what conditions the system receives positive feedback, and under what conditions negative. Once that function is set, all of the system's behaviour, including every seemingly clever strategy and every baffling quirk, unfolds around it. Humans have such a function too, except that we did not write it ourselves; it was written over millions of years of natural selection. When we feel satisfied, when we feel pain, what attracts us and what drives us away, nearly all of it is factory settings.

这里要讨论的愿景是：人应当能够自由设置自己的体验函数。

The vision under discussion here is that people should be able to set their own experience function freely.

要理解这为什么重要，可以看看出厂设置里那些最顽固的部分。

To see why this matters, look at the most stubborn parts of the factory settings.

恐惧。 演化精神病学家伦道夫·尼斯提出过一个“烟雾报警器原理”：对一个生物来说，把一根树枝误认作蛇的代价很小，把一条蛇误认作树枝的代价却可能是死亡，所以自然选择宁可让警报系统过度敏感。这套逻辑在草原上是合理的，可在今天，它会让人对一封措辞含糊的邮件产生与面对猛兽时相差无几的生理反应。

Fear. The evolutionary psychiatrist Randolph Nesse proposed a "smoke detector principle": for an organism, mistaking a stick for a snake costs little, while mistaking a snake for a stick may cost its life, so natural selection prefers an oversensitive alarm system. That logic made sense on the savanna, but today it can produce, in response to an ambiguously worded email, a physiological reaction little different from facing a predator.

习得性无助。 塞利格曼在1960年代的经典实验中发现，经历过无法逃避的电击的狗，后来即使面对可以逃脱的处境，也会放弃尝试。值得一提的是，他和梅尔在半个世纪之后修正了这一理论：神经科学的证据显示，被动和放弃其实是大脑在长期厌恶刺激下的默认反应，真正需要后天学会的，反而是“我能控制”的感觉。这个修正本身就说明了问题：放弃是被预装的，掌控感才是需要争取的。

Learned helplessness. In classic experiments in the 1960s, Martin Seligman found that dogs subjected to inescapable shocks would later give up trying even in situations where escape was possible. It is worth noting that half a century later he and Steven Maier revised the theory: neuroscientific evidence showed that passivity and giving up are in fact the brain's default response to prolonged aversive stimulation, and what has to be learned is the sense that "I can control this." The revision itself makes the point: giving up comes pre-installed; a sense of control is what must be won.

过度的竞争欲与承认渴求。 从黑格尔的“为承认而斗争”到福山所说的“血气”，哲学家们早就注意到，人类对地位与认可的渴望是一种近乎独立的驱动力。在小群体里，地位直接关系到资源和配偶，这种渴望有清楚的演化功能。但在今天，它表现为无休止的比较、社交媒体上的点赞焦虑、学术圈里的论文竞赛，以及一种无论取得多少成就都无法止息的不满足。

Excessive competitiveness and the craving for recognition. From Hegel's "struggle for recognition" to what Fukuyama called thymos, philosophers have long noticed that the human desire for status and recognition is an almost independent drive. In small groups, status bore directly on resources and mates, and the desire had a clear evolutionary function. Today it shows up as endless comparison, anxiety over likes on social media, the publication race in academia, and a dissatisfaction that no amount of achievement can quiet.

这些机制有一个共同点：它们只是一套为远古环境优化过的遗留代码。它们在当时是解决方案，在今天常常变成了问题本身。而一旦我们意识到它们只是默认设置，一个自然的问题就会出现：默认设置是可以改的吗？

These mechanisms have one thing in common: they are only legacy code optimized for an ancient environment. Then they were solutions; today they often become the problem itself. And once we realize they are merely default settings, a natural question arises: can the defaults be changed?

这里必须立刻澄清一个最容易产生的误解。改写体验函数，并不意味着消除一切负面体验。悲伤、紧张、挣扎、失落，这些体验本身都可以有深刻的价值。要清除的不是这些体验，而是它们的强制性。关键的区别在于“被迫”与“选择”。今天的人类，并没有选择恐惧，而是无法不恐惧；并没有选择比较，而是无法停止比较。基因和环境塑造共同替我们做了决定，而我们只能承受这个决定的后果。激励层的愿景，是把这个决定权交还给每一个主体：你可以选择体验悲伤，可以选择保留竞争的刺激，甚至可以选择保留恐惧，只要那是你自己的设定。

One very easy misunderstanding must be cleared up at once. Rewriting the experience function does not mean eliminating all negative experience. Sadness, tension, struggle and loss can all have profound value in themselves. What is to be removed is not these experiences but their compulsoriness. The key distinction is between being forced and choosing. People today do not choose fear; they cannot help being afraid. They do not choose comparison; they cannot stop comparing. Genes and environment have made the decision for us, and we can only bear its consequences. The vision of the motivational layer is to return that decision to each subject: you may choose to feel sadness, you may choose to keep the thrill of competition, you may even choose to keep fear, so long as it is your own setting.

七

VII

前两层都是手段。功能层让体验的基础设施更强，激励层让体验的驱动方式更自由，但它们本身都还没有回答一个问题：这一切是为了什么？

The first two layers are means. The functional layer strengthens the infrastructure of experience and the motivational layer frees the way experience is driven, but neither by itself answers the question: what is all this for?

存在层给出的回答是：升级到一种新的体验结构，使人在未来AI建构的超级复杂世界中，仍然能够获得丰盈的意义感。这才是整个构想真正的终极目标。

The existential layer answers: to upgrade to a new experiential structure that allows people, in the hyper-complex world AI will build, still to find an abundant sense of meaning. This is the true ultimate aim of the whole vision.

要理解这一层，需要先说清楚意义感是怎样产生的。一个常被忽视的事实是，意义不只取决于世界给了我们什么，同时也取决于我们能从中接住什么。同一盘棋，在新手眼里是一堆棋子的移动，在大师眼里却可能是一次令人屏息的弃子，一种在十几步之后才会展开的美。棋盘上的信息完全相同，差别全在观看者的结构里。专业训练之所以能深刻地改变一个人的生活，很大程度上就是因为它在这个人的体验结构里长出了新的接收器，让原本沉默的东西开始说话。

To understand this layer, one must first be clear about how the sense of meaning arises. An often overlooked fact is that meaning depends not only on what the world gives us but also on what we are able to catch from it. The same game of chess is, to a novice, a set of pieces being moved, and to a grandmaster may be a breathtaking sacrifice, a beauty that will unfold only a dozen moves later. The information on the board is identical; the difference lies entirely in the structure of the one looking. Professional training can transform a person's life largely because it grows new receptors in that person's experiential structure, letting things that were silent begin to speak.

意义感可以被理解为一种匹配：世界的复杂度和体验者的结构恰好咬合，于是复杂变成了丰富。结构太简单，世界就是一片混沌；结构足够精细，同样的世界就变得层次分明、意蕴无穷。

The sense of meaning can be understood as a kind of fit: the complexity of the world and the structure of the experiencer engage exactly, and complexity becomes richness. If the structure is too simple, the world is chaos; if it is fine enough, the same world becomes layered and inexhaustible in significance.

这正是AI时代深层次的危险所在。当世界的复杂度以指数级增长，而体验者的结构停在原地，匹配就会断裂。断裂之后，人面对的将是一个越来越不可理解、越来越与己无关的世界。AI或许可以向人解释它所创造的一切，但被解释并不等于被感受。一只狗也许可以通过训练对交响乐的某个段落作出反应，但它永远不会因为那个段落而流泪。

This is where the deeper danger of the AI era lies. When the complexity of the world grows exponentially and the structure of the experiencer stays where it is, the fit breaks. Once it breaks, what people face will be a world ever less intelligible and ever less their concern. AI may be able to explain everything it creates, but being explained is not the same as being felt. A dog might be trained to respond to some passage of a symphony, but it will never weep because of that passage.

所以存在层的升级，目标不是让人理解AI，而是让人能够感受一个更复杂的世界。这意味着发展出新的情绪、新的审美维度、新的意义形态。

So the aim of upgrading the existential layer is not to make people understand AI but to make them able to feel a more complex world. That means developing new emotions, new aesthetic dimensions, new forms of meaning.

这件事的难度在于，它在原理上就超出了我们今天的想象。托马斯·内格尔在1974年那篇著名的论文里追问：做一只蝙蝠是什么感觉？他的结论是，即使我们知道关于蝙蝠回声定位的全部物理事实，我们仍然无法从内部想象那种体验，因为我们没有对应的体验结构。新的情绪也是一样。一个天生的红绿色盲无法想象红与绿的差别是什么样子，我们今天也无法想象一种尚未被演化出来的情感是什么样子。我们只能说：它可能存在，它的空间很可能远比我们今天能体验的那几种情绪的组合要大得多。

The difficulty is that this lies, in principle, beyond what we can imagine today. In his famous 1974 paper Thomas Nagel asked what it is like to be a bat. His conclusion was that even if we knew every physical fact about bat echolocation, we still could not imagine that experience from the inside, because we lack the corresponding experiential structure. New emotions are the same. Someone born red–green colour-blind cannot imagine what the difference between red and green is like, and we cannot imagine today what an emotion not yet evolved would be like. We can only say that it may exist, and that its space is very likely far larger than the combinations of the few emotions we can feel today.

人类目前拥有的高级情绪，大致相当于调色盘上的几种颜色。几千年来，诗人、作曲家、哲学家和修行者用这几种颜色调出了无比丰富的作品，这本身是了不起的成就。但如果调色盘本身可以扩展，如果可以增加原色，增加维度，那么可能的体验空间将以组合的方式爆炸性地增长。

The higher emotions humans now possess are roughly like a few colours on a palette. For thousands of years poets, composers, philosophers and contemplatives have mixed from those colours works of boundless richness, and that is a remarkable achievement in itself. But if the palette itself could be expanded, if primary colours and dimensions could be added, the space of possible experience would grow explosively, combinatorially.

这就是“丰盈”的含义。它不是更多的快乐，也不是更强的刺激，而是更多层次、更多维度、更深纵深的意义感。在一个由AI构建的高复杂度世界里，只有拥有这样的体验结构，人才能继续作为在场的体验者，而不是一个站在音乐厅里、只听得见声音的旁观者。

This is what abundance means. It is not more pleasure, nor stronger stimulation, but a sense of meaning with more layers, more dimensions and greater depth. In a high-complexity world built by AI, only with such an experiential structure can people remain present as experiencers, rather than bystanders standing in a concert hall hearing only sound.

八

VIII

讨论意识层面的超人类主义，绕不开一个著名的先例。1995年，英国哲学家大卫·皮尔斯发表了《享乐主义律令》，主张人类应当利用生物技术彻底废除一切形式的痛苦，让未来的意识只在不同梯度的幸福之间变化。这份宣言在超人类主义思想史上地位很高，也常常被当作“体验改造”这一方向的代名词。

Any discussion of transhumanism at the level of consciousness has to reckon with a famous precedent. In 1995 the British philosopher David Pearce published The Hedonistic Imperative, arguing that humanity should use biotechnology to abolish suffering in all its forms, so that future consciousness would vary only across gradients of well-being. The manifesto holds a high place in the history of transhumanist thought and is often treated as a byword for the project of modifying experience.

本文的立场与它有本质的不同。

The position of this essay differs from it fundamentally.

皮尔斯的构想里隐含着一个统一的范式：存在一种对所有意识都普遍适用的“好”，那就是幸福；痛苦则是普遍的“坏”，应当被消灭。而本文认为，未来的体验结构没有统一范式。体验什么，是每一个主体自己的自由。

Pearce's vision contains an implicit single paradigm: there is a good universally valid for all consciousness, namely happiness, and suffering is a universal bad to be eliminated. This essay holds instead that future experiential structures have no single paradigm. What to experience is each subject's own freedom.

这是两种截然不同的世界图景。在统一范式下，体验者升级是一项以某个目标为终点的工程，所有人最终都应该抵达同一个地方。在多元范式下，体验者升级是一种可能性空间的打开，它扩展的是选择。

These are two utterly different pictures of the world. Under a single paradigm, upgrading the experiencer is a project with a fixed goal as its end point, and everyone should finally arrive at the same place. Under a pluralist paradigm, upgrading the experiencer is the opening of a space of possibility; what it expands is choice.

超人类主义传统中有一个相关的概念，叫“形态自由”。它最早由马克斯·莫尔提出，后来安德斯·桑德伯格为它做了系统的辩护，其核心是：每个人都有权按照自己的意愿改造或者不改造自己的身体。本文所说的，可以看作是把这个原则从身体推进到体验：一种体验自由。你有权决定自己的时间感、注意力、动机结构和情感维度，也有权决定不去改变它们。

The transhumanist tradition has a related concept called morphological freedom. First proposed by Max More and later systematically defended by Anders Sandberg, its core is that everyone has the right to modify, or not modify, their own body as they wish. What this essay proposes can be seen as carrying the principle from the body to experience: a freedom of experience. You have the right to decide your own sense of time, your attention, your motivational structure and your emotional dimensions, and the right to decide not to change them.

这意味着几件事。

This means several things.

有人想追求丰盈，想发展出能够感受AI时代复杂交响的新结构，这是一种选择。有人说：我不想要那种丰盈，我就想在今天这个有局限的人脑里去感受世界，去爱、去痛、去在有限中体会有限的美。这同样是一种完整的、值得尊重的选择。甚至有人想把自己改造得更有局限一点，更专注于某一种简单的体验，像一个只弹原声吉他的乐手拒绝所有电子效果器，那也完全可以。

Someone who wants to pursue abundance, to develop new structures able to feel the complex symphonies of the AI age, is making one choice. Someone who says, I don't want that abundance, I just want to feel the world inside today's limited human brain, to love, to hurt, to find within finitude a finite beauty, is making a choice equally complete and equally worthy of respect. Someone who even wants to make themselves somewhat more limited, more focused on one simple kind of experience, like a guitarist who plays only acoustic and refuses every electronic effect, may do that too.

多元范式还有一层更深的意义。人类历史上反复出现的教训之一是：一旦某个单一的指标被确立为唯一目标，并被大规模地优化，结果往往是灾难性的。今天的注意力经济就是一个现成的例子：它把停留时长和点击率当作唯一的目标函数，结果是十亿人的注意力被碎片化。如果体验者升级被纳入一个统一的“幸福最大化”范式，它很可能重蹈同样的覆辙，只是规模更大、程度更深。体验的多样性，就像生物的多样性一样，本身就是一种抗风险的结构。

Pluralism has a deeper significance as well. One lesson that recurs throughout human history is that once a single metric is established as the sole goal and optimized at scale, the results are often disastrous. Today's attention economy is a ready example: it treats time on site and click-through rate as its only objective function, and the result is the fragmented attention of billions. If upgrading the experiencer were placed under a single paradigm of maximizing happiness, it would very likely repeat the same mistake, only on a larger scale and to a greater depth. Diversity of experience, like biological diversity, is itself a risk-resistant structure.

所以，这里要打开是一片广阔的可能性地形。每一个主体都可以在其中选择自己的位置，甚至在不同的时期选择不同的位置。未来的体验世界，很可能会像今天的生物圈一样分化出无数种形态，而它们之间并没有一个公认的高下排序。

What is to be opened here, then, is a vast landscape of possibility. Each subject can choose its own position within it, and even different positions at different times. The future world of experience may well diversify into countless forms, like today's biosphere, with no accepted ranking among them.

九

IX

这样一个激进的构想，在当下必然会招来大量的反对。下面先把反对分成两类，因为它们的性质完全不同。

So radical a vision is bound, at present, to draw a great deal of opposition. Let us first divide the objections into two kinds, because their character is entirely different.

第一类反对关心的是“怎么做”。改造是否基于充分的知情同意？改造是否可逆，如果一个人后悔了，能否回到原来的状态？如何防止有人在别人不知情的情况下被改造，或者被社会压力、经济压力逼着去改造？改造过程中，主体的自主性和自我修正能力如何被保全？这些问题是真实的，而且非常重要。但它们不是反对这件事的理由，而是这件事的工程规范。就像《现象学工程》里所讲到的那样，桥会塌是结构力学存在的理由。

The first kind is concerned with how. Is modification based on fully informed consent? Is it reversible; if someone regrets it, can they return to their original state? How do we prevent people from being modified without their knowledge, or pushed into it by social or economic pressure? How are the subject's autonomy and capacity for self-correction preserved during modification? These questions are real and very important. But they are not reasons against the project; they are its engineering specifications. As Phenomenological Engineering put it, collapsing bridges are the reason structural mechanics exists.

第二类反对关心的是“该不该做”，而且它们往往有一个共同的出发点：今天人类的体验结构。这一类反对里，有很多其实是无效的。原因不在于它们的逻辑有错误，而在于它们的前提本身就是这场升级可能改变的东西。用一个可能被改变的结构所产生的判断，去评判对这个结构的改变，这在方法上是循环的。

The second kind is concerned with whether, and it usually shares a common starting point: the experiential structure of humans today. Many objections of this kind are in fact invalid. Not because their logic is faulty, but because their premises are themselves among the things this upgrade might change. To judge a change to a structure by judgments that structure produces, when the structure itself may be changed, is methodologically circular.

以“社会公平伦理”为例。最常见的反对是：如果只有一部分人完成了体验升级，岂不是制造了一种新的、更深的不平等？这个担忧听上去理所当然，但值得追问一步：公平伦理为什么对我们如此重要？

Take the ethics of social fairness. The most common objection runs: if only some people complete the experiential upgrade, won't that create a new and deeper inequality? The worry sounds self-evident, but it is worth asking one step further: why does the ethics of fairness matter so much to us?

一个相当有力的解释是：因为人脑天生就爱比较。我们对自己处境的感受，很大程度上不取决于绝对水平，而取决于和他人的相对位置。输给别人会带来真实而强烈的痛苦，这种痛苦又会转化为怨恨、冲突乃至暴力。为了不让输家太惨，为了让群体能够稳定地合作下去，人类在漫长的历史中发展出了一整套关于公平的规范。从这个角度看，公平伦理在相当程度上是一个补丁，它是为了应对“比较心理”这个演化遗产而被打上去的。

One quite powerful explanation is that the human brain is born to compare. How we feel about our situation depends largely not on absolute level but on our position relative to others. Losing to others brings real and intense pain, and that pain turns into resentment, conflict and even violence. To keep the losers from faring too badly, and to let groups go on cooperating stably, humanity developed over its long history a whole set of norms about fairness. Seen this way, the ethics of fairness is to a considerable degree a patch, applied to cope with the evolutionary legacy of comparative psychology.

这不是说公平是错的，也不是说它不重要。在今天的人类世界里，它是必不可少的。问题在于，它之所以重要的理由，依赖于一个心理事实，而这个心理事实恰恰在激励层改造的范围之内。如果一个主体已经不再被过度竞争欲和承认渴求所驱动，如果比较本身不再生成痛苦，那么建立在这种痛苦之上的伦理框架，它的适用条件就发生了变化。

This is not to say fairness is wrong, nor that it is unimportant. In today's human world it is indispensable. The point is that the reason it matters depends on a psychological fact, and that psychological fact lies precisely within the scope of the motivational layer's modifications. If a subject is no longer driven by excessive competitiveness and the craving for recognition, if comparison itself no longer generates pain, then the ethical framework built on that pain has had its conditions of application changed.

社会公平并不是孤例。顺着同样的思路，至少还可以看到另外两种常见的反对。

Social fairness is not an isolated case. Following the same line of thought, at least two other common objections come into view.

努力换来的意义。 另一种常见的反对是：如果满足感可以被直接设定，那么奋斗、等待和长期投入还有什么意义？意义难道不正是在一个人下了重注、熬过漫长的匮乏之后，才终于兑现的吗？一个为了一项事业坚持十年才看到结果的人，和一个直接被设定为感到满足的人相比，前者的满足似乎有一种后者无法拥有的分量。这种直觉很强烈，但值得追问：为什么意义必须以长期匮乏和沉重的下注为代价？

Meaning earned through effort. Another common objection runs: if satisfaction can be set directly, what is the point of striving, waiting and long-term commitment? Isn't meaning precisely what pays out only after a person has placed a heavy bet and endured a long stretch of deprivation? Compared with someone simply set to feel satisfied, a person who persists ten years in an undertaking before seeing results seems to have a satisfaction with a weight the other cannot possess. The intuition is strong, but it is worth asking: why must meaning be paid for with long deprivation and heavy bets?

一个重要的原因在于，人类的奖赏系统就是这样计算价值的。神经科学对多巴胺的研究表明，奖赏信号对应的不是结果本身，而是结果与预期之间的落差：越是长期得不到、越是不确定、投入越大，最终兑现时的满足就越强烈。在资源稀缺的环境里，这套机制能驱使个体为不确定的长期回报持续投入，于是“先匮乏、再下注、最后兑现”成了人类意义感最典型的形状。心理学所说的“努力合理化”，即人们会更珍视费了力气才得到的东西，也是这套机制留下的痕迹。换句话说，“意义必须靠长期匮乏之后的努力去换”这个判断，本身就是奖赏结构的产物。这不是说奋斗没有价值，一个人完全可以选择保留这种形状的意义；而是说，意义常常只能以这种形状出现，这一点并不是意义的本质，只是今天这套奖赏结构的特征。

An important reason is that this is how the human reward system computes value. Research on dopamine shows that the reward signal corresponds not to the outcome itself but to the gap between outcome and expectation: the longer something has been out of reach, the more uncertain it was, the more one invested, the more intense the satisfaction when it finally pays out. In an environment of scarcity this mechanism drives individuals to keep investing in uncertain long-term returns, and so "deprivation first, then the bet, then the payout" has become the most typical shape of the human sense of meaning. What psychology calls effort justification, the tendency to value more highly what was hard to obtain, is another trace of the same mechanism. In other words, the judgment that meaning must be bought with effort after long deprivation is itself a product of the reward structure. This is not to say striving has no value; a person may well choose to keep meaning of this shape. It is to say that the fact that meaning often can appear only in this shape is not the essence of meaning, merely a feature of today's reward structure.

自我同一性。 另一种反对更为根本：如果一个人改写了自己的时间感、动机和情感结构，那么改造之后的那个人还是他吗？这难道不是一种温和的自我毁灭？这个担忧背后有一个关于自我的预设：存在一个稳定而连续的“我”，它是值得被保存的那个单位，而任何深刻的改变都是对它的威胁。

Personal identity. Another objection goes deeper: if a person rewrites their sense of time, their motivations and their emotional structure, is the person after the modification still them? Isn't this a gentle form of self-destruction? Behind the worry lies a presupposition about the self: that there exists a stable and continuous "I" that is the unit worth preserving, and that any profound change is a threat to it.

但这个“我”是怎样被经验到的？它依赖于记忆以特定的方式保存和提取，依赖于一条把过去与现在串起来的叙事主线，也依赖于一种根植极深的自我保存本能，正是这种本能让“不再是我”听上去近乎死亡一样可怕。这三者，恰好分别落在记忆管理、叙事与激励结构的改造范围之内。况且，人本来就在不断变化：一个人十岁和五十岁时的体验结构，差别可能大到彼此难以理解，而我们并不认为这是一种死亡。我们对自我连续性有多执着，很大程度上取决于今天的体验结构如何编织自我感。一旦这种编织方式本身可以调整，“改造之后还是不是我”这个问题，也会随之换一种问法。

But how is this "I" experienced? It depends on memory being stored and retrieved in particular ways, on a narrative through-line stringing past and present together, and on a deeply rooted instinct of self-preservation, the very instinct that makes "no longer being me" sound almost as frightening as death. These three fall respectively within the scope of memory management, narrative and the modification of motivational structure. Besides, people are always changing: a person's experiential structure at ten and at fifty may differ so much that each would find the other hard to understand, and we do not regard this as a kind of death. How strongly we cling to the continuity of the self depends largely on how today's experiential structure weaves the sense of self. Once that weaving can itself be adjusted, the question of whether one is still oneself after modification will be posed differently as well.

这三个例子遵循的是同一个逻辑。公平之所以重要，依赖于比较会生成痛苦；“意义要靠努力去换”之所以成立，依赖于奖赏系统以匮乏和下注为前提来计算价值；同一性之所以重要，依赖于自我感被编织的方式，以及对消失的恐惧。它们在今天都成立，但它们赖以成立的前提，都恰好是这场升级可能改写的东西。

The three examples follow the same logic. Fairness matters because comparison generates pain; "meaning must be earned through effort" holds because the reward system computes value on the premise of deprivation and betting; identity matters because of the way the sense of self is woven, and the fear of disappearing. All of them hold today, but the premises on which they rest are precisely what this upgrade might rewrite.

尼采曾经追问道德的谱系：我们的善恶观念是从哪里来的？是在什么样的生存条件下被塑造出来的？体验者升级会把这个追问从历史学变成一个现实问题。当伦理所赖以成立的心理基础本身变得可以设计，伦理学就必须回答一个它从未真正面对过的问题：在体验结构可变的世界里，什么样的规范仍然成立？

Nietzsche once asked about the genealogy of morals: where do our ideas of good and evil come from? Under what conditions of life were they shaped? Upgrading the experiencer would turn that inquiry from a historical question into a practical one. When the psychological foundations on which ethics rests themselves become designable, ethics must answer a question it has never truly faced: in a world where experiential structure can vary, which norms still hold?

回答这个问题，需要的是一套更高维的伦理学。

Answering that question requires a higher-dimensional ethics.

迄今为止的伦理学，无论是功利主义、义务论还是美德伦理，都默认了一个固定的参数：人类的心理结构。人会痛苦，会比较，会恐惧，会渴望被承认，这些被当作给定的事实，伦理学则在这些事实之上寻找最好的相处方式。功利主义之所以能把一切折算成快乐与痛苦，是因为它假定所有人共享同一种痛苦的度量；康德之所以能诉诸普遍的理性法则，是因为他假定所有理性存在者有着相同的结构。这些理论彼此争论了几百年，但它们都在同一个平面上争论。

Ethics so far, whether utilitarian, deontological or virtue-based, has assumed one fixed parameter: human psychological structure. People suffer, compare, fear and crave recognition; these are treated as given facts, and ethics searches, on top of them, for the best way to live together. Utilitarianism can convert everything into pleasure and pain because it assumes everyone shares one measure of suffering; Kant can appeal to a universal law of reason because he assumes all rational beings share the same structure. These theories have argued with one another for centuries, but always on the same plane.

更高维的伦理学，是把这个参数本身放进考虑范围的伦理学。它要问的是：在所有可能的体验结构所构成的空间里，什么样的选择是被允许的，在什么条件下被允许，以及不同结构的存在者之间应当如何共存。在这个意义上，今天的伦理学并不会被废除，它只是这个更高维空间中的一个切片：在“人类现有心理结构”这一截面上，它依然有效，正如二维平面上的规则，是三维空间在那个截面上的投影。

A higher-dimensional ethics is one that brings that parameter itself into consideration. What it must ask is: within the space formed by all possible experiential structures, which choices are permissible, under what conditions, and how beings of different structures ought to coexist. In this sense today's ethics will not be abolished; it becomes a single slice of that higher-dimensional space. On the cross-section of "existing human psychological structure" it remains valid, just as the rules of a two-dimensional plane are the projection of three-dimensional space onto that cross-section.

这样一套伦理学今天还远未成形，但至少可以看到它可能关心的几个方向。其一，异质共存：当两个存在者不再共享同一种痛苦的度量、同一种欲望的语法，传统伦理中那种“将心比心”的推理就失去了依托，需要一种不以共同心理为前提的相处原则。其二，结构选择的伦理：一个人可以改写自己，但他能否为尚未出生的人选择体验结构？能否改写到一种无法再改回来的状态？其三，那些在每一个切片上都成立的东西。有意思的是，前面提到的第一类反对，即知情同意、可逆性、主体的自主与自我修正能力，恰恰最有可能成为这样的不变量。它们不依赖于任何特定的心理内容，只关心主体是否仍然保有选择与修正的余地。旧伦理中那些以特定心理为前提的规范，可能会随着结构的改变而失效；而这些守护选择本身的规范，很可能会成为更高维伦理学的地基。

Such an ethics is still far from formed, but at least a few of its likely concerns can be seen. First, heterogeneous coexistence: when two beings no longer share one measure of suffering or one grammar of desire, the traditional ethical reasoning of putting oneself in another's place loses its footing, and a principle of coexistence is needed that does not presuppose a shared psychology. Second, the ethics of structural choice: a person may rewrite themselves, but may they choose an experiential structure for someone not yet born? May they rewrite themselves into a state from which there is no way back? Third, whatever holds on every slice. Interestingly, the first kind of objection discussed earlier, informed consent, reversibility, the subject's autonomy and capacity for self-correction, is the most likely candidate for such invariants. They depend on no particular psychological content; they care only whether the subject still retains room to choose and to correct. Norms of the old ethics that presuppose a particular psychology may lapse as structures change, while these norms that guard choice itself may well become the foundation of a higher-dimensional ethics.

这个问题今天还没有答案。但至少应该承认：拿旧框架下的答案去否决新框架下的问题，本身并不构成一个论证。

This question has no answer today. But it should at least be acknowledged that using answers from the old framework to veto questions from the new one does not, in itself, constitute an argument.

十

X

最后需要坦白一件事：几乎不可能全人类一次性地跳入这个体验升维的新世界。

One last admission is needed: it is almost impossible for all of humanity to leap at once into this new world of expanded experience.

每一次体验结构的扩展，最初都只属于少数人。最早在阿尔卑斯山的冰川前感到崇高的，是一小撮诗人和旅行者，在他们同时代的大多数人眼里，这种行为近乎怪癖。最早为《春之祭》喝彩的，也只是剧场里的一部分人，另一部分人在他们身边吹口哨、扔东西。新的体验形态总是先在少数人身上出现，然后经过漫长的时间，才可能扩散为一种普遍的能力，也可能永远不会。

Every expansion of experiential structure has at first belonged to a few. The first to feel the sublime before Alpine glaciers were a small band of poets and travellers, whose behaviour struck most of their contemporaries as close to eccentric. The first to cheer The Rite of Spring were only part of the theatre, while the rest whistled and threw things beside them. New forms of experience always appear first in a few people, and only after a long time may they spread into a general capacity, or perhaps never do.

体验者升级大概也会遵循同样的路径，只是落差可能更大。少数勇敢的体验者率先进入新的体验结构，而大多数人仍然带着旧的人类框架生活。两者之间的矛盾，很可能会成为一段时期内最核心的社会张力。这种张力不同于以往的贫富差距或观念分歧，因为它触及的是更底层的东西：当两群人的体验结构本身不同，他们甚至可能很难理解对方究竟在经历什么。勇敢者会觉得保守者固步自封（当然，更可能是根本就没有任何的看法），保守者会觉得勇敢者已经不再是“人”。在比较悲观的情况下，这段时期可能会呈现出《赛博朋克2077》里那种氛围：在夜之城里，改造技术被少数巨型公司垄断，升级成为一种消费品和一笔债务；被录制下来的他人体验，即所谓“超梦”，作为最廉价的娱乐在街头流通，越极端越好卖；有人在一次次改造中失去了与自身的联结，陷入所谓的“赛博精神病”；升级者与未升级者之间不再有共同的语言，只剩下猜忌、剥削和暴力。技术本身并不是那个世界的问题，问题在于它在没有任何体验层面的理论、规范与伦理承接的情况下，被直接交给了市场和强权。这段时期最终会如何展开，今天没有人知道。

Upgrading the experiencer will probably follow the same path, only with a larger gap. A brave few will enter new experiential structures first, while the majority go on living within the old human frame. The conflict between them may well become the central social tension of a whole period. This tension differs from earlier gaps of wealth or divergences of opinion, because it touches something more fundamental: when two groups differ in experiential structure itself, they may find it hard even to understand what the other is going through. The brave will think the conservatives stuck in their ways (or, more likely, will have no opinion of them at all), and the conservatives will think the brave are no longer "human." In a fairly pessimistic scenario, the period might take on the atmosphere depicted in Cyberpunk 2077. In its Night City, modification technology is monopolized by a few megacorporations, and upgrading becomes a consumer good and a debt; other people's recorded experiences, so-called braindances, circulate on the street as the cheapest entertainment, selling better the more extreme they are; some people, through modification after modification, lose their connection with themselves and fall into so-called cyberpsychosis; between the upgraded and the un-upgraded there is no longer any common language, only suspicion, exploitation and violence. Technology itself is not that world's problem; the problem is that it was handed straight to markets and power without any theory, norms or ethics at the level of experience to receive it. How this period will finally unfold, no one knows today.

所以本文只说愿景，不说预言。它不主张人类一定会顺利地进入这个新时代，也不提供一份时间表。本文只是试图指出一个方向：如果瓶颈的迁移是真实的，如果AI正在把上游的产出推向一个人类无法接收的量级，那么迟早有一天，人类必须面对那个此前从未被当作问题的环节。

So this essay speaks of vision, not prophecy. It does not claim that humanity will enter this new era smoothly, nor does it offer a timetable. It tries only to point out a direction: if the migration of bottlenecks is real, and if AI is pushing upstream output toward a scale humans cannot receive, then sooner or later humanity must face the link that has never before been treated as a problem.

最后回到《现象学工程》里提到的那台诺齐克体验机。它指出，人们之所以拒绝接上这台机器，根本原因不在于它是假的，而在于它是薄的：它只承诺单点强度的供给，却不交代这些体验如何衔接、如何累积，也不交代人在其中是否还保有行动与自我修正的余地。本文所做的，实际上就是进一步说明：诺齐克的想象为什么会薄。

Finally, back to Nozick's experience machine as discussed in Phenomenological Engineering. That essay argued that the fundamental reason people refuse to plug into the machine is not that it is fake but that it is thin: it promises only a supply of single-point intensity, with no account of how these experiences connect or accumulate, and no account of whether the person retains any room to act and to correct themselves. What this essay does is, in effect, explain why Nozick's imagining was thin.

因为诺齐克在设计这台机器时，默认了一个不变的体验者。他所能想象的“完美体验”，只能从今天人类体验结构所能接住的东西里挑选：更多的快乐，更大的成就感，更强烈的爱与被爱。被供给的永远是同一块调色盘上的那几种颜色，薄几乎是必然的结局。更关键的是，连“这台机器不值得接上”这个判断本身，也出自同一套结构。诺齐克给出的理由是，我们想要真正去做事，想要成为某种人，想要接触更深的实在；他把这些当作人类固有的体验判断，用来裁决机器的价值。换句话说，他用旧的体验者设计了机器，又用同一个旧的体验者审判了它。诺齐克让世界可以被任意编程，却没有想到体验者本身也可以被升级。

Because in designing the machine, Nozick assumed an unchanging experiencer. The "perfect experiences" he could imagine could only be picked from what today's human experiential structure can receive: more pleasure, greater achievement, more intense love and being loved. What is supplied is always the same few colours from the same palette, and thinness is nearly inevitable. More crucially, even the judgment that the machine is not worth plugging into comes from the same structure. Nozick's reasons were that we want actually to do things, to be a certain kind of person, to be in contact with a deeper reality; he treated these as inherent human experiential judgments and used them to rule on the machine's worth. In other words, he designed the machine with the old experiencer and then had the same old experiencer pass judgment on it. Nozick made the world infinitely programmable, but never thought that the experiencer itself could be upgraded.

这是本文与上一文交汇的地方。一台不薄的体验机，既需要一套懂得如何组织经验整体的工程，也需要一个能够承接更厚经验的体验者。前者回答的是给予什么，后者回答的是能接住什么。

This is where this essay and the previous one meet. An experience machine that is not thin needs both an engineering that knows how to organize experience as a whole and an experiencer able to receive thicker experience. The former answers what is given; the latter answers what can be caught.

说明：本文的行文设计得到了 Claude Opus 5.5 的协助。

Note: Claude Opus 5.5 assisted with the composition and structure of this essay.