书稿完成后的一点记录

A Few Notes on Finishing the Book Manuscript

2022年底的ChatGPT发布之时，与大多数人的震惊与好奇不一样，我却陷入了一种更复杂的忧惧之中。虽然GPT在那个时候只是一个聊天工具，其能力距离真正的智能还相差甚远，但我却有一种强烈的直觉：人工智能（AI）替代我作为科研工作者的核心竞争力的时候不远了。当然，大众媒体上有很多人都因为各种各样的目的在宣传这样一个论调："当你觉得AI将要替代你的工作，那说明你工作的层次可能有点低。"但我本能地反对这种论调背后所基于的假设——即人类有某些AI无法替代的高级能力。我在科研工作中除了思考问题外，还喜欢反思我思考问题的过程是怎样的；同时我还粗略地了解过科学史上的一些伟大思想的提出过程。正因如此，我相当确信的一点是：没有什么思想是AI提不出来的，只是当前的AI架构还无法支持而已。从2022年末到2024年末，我几乎是带着这种等待末日审判一般的心态在继续我的研究，强烈的无意义感随时都会来侵袭我的内心。

When ChatGPT was released at the end of 2022, unlike the shock and curiosity felt by most people, I fell into a more complex state of dread. Although GPT was only a chatbot at the time, its capabilities were still far from true intelligence, yet I had a strong intuition: the day when Artificial Intelligence (AI) would replace my core competence as a scientific researcher was not far off. Of course, many in the mass media, for various reasons, were promoting the argument that, "If you feel AI is about to replace your job, it probably means your job is of a somewhat low level." But I instinctively rejected the assumption underlying this argument—that humans possess certain high-level abilities that AI cannot replace. In my research, besides thinking about problems, I also enjoy reflecting on how my thinking process works. I also have a rough understanding of how some of the great ideas in the history of science came to be. For these reasons, I was quite certain of one thing: there is no idea that AI cannot conceive; it's just that the current AI architecture cannot yet support it. From the end of 2022 to the end of 2024, I continued my research almost with the mindset of awaiting a final judgment day, where a strong sense of meaninglessness could invade my heart at any moment.

2025年初的某一天，毫无征兆地，我内心中有一个声音突然响起："焦虑是没有意义的，这一天迟早会到来。与其困在其中，不如主动去探索些什么吧。"就是因为这个声音，我从床上爬起来，打开了Overleaf（一个文稿编辑的在线网站），并新建了一个书稿项目。我在2022年已经完成过我的研究领域内的一本专著，因此写书对我而言不是一个陌生的活动。我那时的设想是：通过书籍写作这个既需要耐心、也需要结构化地整理思路的活动，来对冲强烈的AI焦虑感。在ChatGPT的协助下，我用了几天的时间，整理出来了一些初步的想法，然后便开始推进这本有关AI的书籍。半年之后，这本书的初稿顺利完结了。直到现在我也没有起好本书的名字，因此就用"这本书"来代替称呼吧。

Then, one day in early 2025, without any warning, a voice suddenly arose within me: "Anxiety is meaningless. This day will come sooner or later. Instead of being trapped by it, why not proactively explore something?" It was because of this voice that I climbed out of bed, opened Overleaf (an online document editor), and created a new book project. I had already completed a monograph in my research field in 2022, so writing a book was not an unfamiliar activity for me. My idea at the time was to use book writing—an activity that requires both patience and the structured organization of thoughts—to counteract the intense AI anxiety. With the assistance of ChatGPT, I spent a few days organizing some preliminary ideas and then began to advance this book about AI. Six months later, the first draft of the book was successfully completed. I still haven't come up with a good title for it, so for now, I will just call it "this book."

虽然我平日的研究更多是偏技术性的工作，但本书却并不是技术性的。它涉及到了一些AI相关的技术内容以及设想，但它更多的内容却与社会、哲学、文明挂钩。可以说，这本书的写作过程，就是在AI这一主题的牵引下，把我多年来从各种渠道了解的或碎片化、或系统化的知识传承一整个体系的过程。写作过程并没有太多的艰辛，更多是一种融会贯通、乃至最终打通整个思想体系的无与伦比的体验。

Although my daily research is more technical in nature, this book is not. It touches upon some AI-related technical content and concepts, but it is more deeply connected to society, philosophy, and civilization. One could say that the process of writing this book, guided by the theme of AI, was a process of weaving together the fragmented or systematic knowledge I have accumulated over many years from various sources into a coherent whole. The writing process was not particularly arduous; rather, it was an unparalleled experience of synthesis, integration, and ultimately, of breaking through to a complete system of thought.

在书稿完成后，我对于AI的"大图景"层面的理解也暂时定型。如果说这个图景中有什么值得在此记录的侧面的话，那么大概可以有以下的三个。

After the manuscript was completed, my understanding of the "big picture" of AI also temporarily solidified. If there are any aspects of this picture worth recording here, they would probably be the following three.

1、在完成这一书稿后，我更加坚定了我的判断：AI的思考能力（以及基于其之上的对物理世界的改造能力）有着近乎无穷的潜力。原因不在于这半年间大语言模型的能力增长了多少，而是在于我更加深入地分析了人的思维过程，并且得到一个结论：它可以被形式化。无论是多么神秘的思维活动，想象也好，直觉也好，灵感也好，都可以在一个形式化的框架内进行阐释。而人脑思维过程的可形式化也就注定了AI迟早会对其进行能力上的超越。当然，当前的大语言模型的架构可能还无法复现人类的某些能力，但这并不构成AI无法在能力上超越人类的理由。

1. After completing this manuscript, I became even more firm in my judgment: AI's ability to think (and, based on that, its ability to transform the physical world) has nearly infinite potential. The reason for this is not how much the capabilities of large language models have grown in the past six months, but rather that I have more deeply analyzed the human thought process and reached a conclusion: it can be formalized. No matter how mysterious a mental activity may be—be it imagination, intuition, or inspiration—it can be explained within a formal framework. And the formalizability of the human brain's thought process means that it is inevitable that AI will one day surpass it in capability. Of course, the current architecture of large language models may not yet be able to replicate certain human abilities, but this does not constitute a reason why AI cannot surpass humanity in capability.

2、按照当前的发展速度推演，AI在能力上超越顶尖人类的时间点可能很快就要到来。而人类做好了这个准备吗？我想大概是没有的。硅谷的科技领袖们大多数也相信AI很快就能超越人类，但出于显而易见的原因，他们中更多都持有"技术发展会带来更好的生活"的观点。在短期内，我是严重怀疑这一点的。作为一种无比强大的工具，AI既可以创造美好，也可以带来灾难——这取决于人类社会的基本组织模式是建立在信任与合作之上，还是建立在竞争与对抗之上。至少对我而言，这个问题的答案几乎是显而易见的。如果说现代文明相比古代文明有什么更优越的地方的话，那一定是它通过完善的法律和市场机制，建立了一个可以带来物质和技术进步的竞技场；但请注意，它终究是一个竞技场。AI带来的威胁决不仅仅是科幻作品中的"机器觉醒"这种拟人化想象，更有可能是竞技场中的木剑被换成了喷火器。

2. Extrapolating from the current pace of development, the point at which AI surpasses top humans in capability may arrive very soon. Is humanity prepared for this? I think not. Most tech leaders in Silicon Valley also believe that AI will soon surpass humans, but for obvious reasons, most of them hold the view that "technological development will bring a better life." In the short term, I seriously doubt this. As an incredibly powerful tool, AI can create beauty as well as disaster—it all depends on whether the basic organizational model of human society is built on trust and cooperation, or on competition and confrontation. For me, at least, the answer to this question is almost self-evident. If modern civilization has any superiority over ancient civilizations, it must be that it has established an arena that fosters material and technological progress through well-developed legal and market mechanisms. But please note, it is, in the end, an arena. The threat posed by AI is not just the anthropomorphic fantasy of "machine awakening" from science fiction, but more likely the scenario where the wooden swords in the arena are replaced with flamethrowers.

3、那么，结合上面的两点，是不是AI技术对于人类自身能力的超越，再结合人类社会的竞技场本质，必然导向的结论就是人类的终结呢？这也是这本书所想要触及到的最深远的层次。在写作本书之前，我的理解（或者说是担忧）是：当AI在能力上全面超越人类后，人类存在的必要性也就没有了；至少说，未来历史的主角就不再是人类了。但在写作本书的过程中，我渐渐意识到自己犯了一个重要的错误：上述的理解完全是在现代社会某种"不言自明"的逻辑基座之上得到的线性推演，即人的价值在很大程度上都是由他的某种"工具性能力"来定义的——这也是人类社会作为一个竞技场的必然结果。但人的一个更加根本的价值，即其作为体验主体的价值，是AI所无法替代的。令人遗憾的是，在工具理性甚嚣尘上的时代，这一更加根本的价值反而被掩盖了。因此，即使面临着AI冲击的风险，我相信人类文明在未来的AI时代依然有一次伟大的机会——在AI的协助下，从"工具性文明"升级为"主体性文明"的机会。若这一愿景能够得以实现，那么人类将会真正迎来一个每个人都能够去创造自己的人生故事的时代。当然，人类是否能够把握住这次机会，现在是完全无法预料的；但至少可以讲，被更强大的AI彻底取代、或用AI走向自我毁灭，并不是人类必然的归宿。

3. So, combining the two points above, does the surpassing of human abilities by AI technology, combined with the competitive arena-like nature of human society, inevitably lead to the end of humanity? This is the deepest level this book seeks to touch upon. Before writing this book, my understanding (or rather, my fear) was that once AI comprehensively surpassed humanity in capability, the necessity for human existence would vanish; at the very least, humans would no longer be the protagonists of future history. But in the process of writing this book, I gradually realized that I had made a critical mistake: the aforementioned understanding was a linear extrapolation based on a certain "self-evident" logical foundation of modern society, namely, that a person's value is largely defined by their "instrumental ability"—an inevitable consequence of human society as a competitive arena. But a more fundamental value of a person, their value as a subject of experience, is something AI cannot replace. Regrettably, in an era where instrumental rationality reigns supreme, this more fundamental value has been obscured. Therefore, even in the face of the risks posed by AI, I believe human civilization still has a great opportunity in the coming AI era—the opportunity, with AI's assistance, to upgrade from an "instrumental civilization" to a "subjective civilization." If this vision can be realized, then humanity will truly enter an age where everyone can create their own life story. Of course, whether humanity can seize this opportunity is completely unpredictable right now; but at least it can be said that being completely replaced by a more powerful AI, or using AI to march toward self-destruction, is not our inevitable destiny.

幸运的是，随着本书的完成，困扰了我两年的AI焦虑也基本上烟消云散了。虽然这本书的主题是关于AI的，但这个写作过程也阴差阳错地给予了我一次彻底审视并重构自己的人生故事的机会。我现在依然相信，过不了多少年，我作为一个科研工作者所能产生的工具性价值会被AI所替代；AI将能够自主地设定研究问题，组织已有的思想并涌现出新的思想，生成研究方案，自动执行研究方案，并在这个过程中不断地精筛，最终得到一系列近乎完美的研究结果。但这又如何呢？研究这一活动的初衷本就在于对未知的探索，这是一个体验过程；是否是全世界第一个证明某个定理、研究成果是否能够换取领域内的泰斗地位、被引用了多少次，这些全都是在工具性异化之下才涌现出来的"指标"。更重要的是，即使是追寻这些外在指标，人最终也是希望通过这些指标来满足内心的某种体验维度。既然最终我们都是要通过行动来得到某种体验，那么AI时代人的终极课题也就呼之欲出了——如何能够在AI这一新工具的帮助下，对我们自己的主体性进行升级？这也是我们每个人都需要勇敢践行的课题。

Fortunately, with the completion of this book, the AI anxiety that had plagued me for two years has essentially vanished. Although the theme of the book is about AI, the writing process, by a twist of fate, also gave me a chance to thoroughly examine and reconstruct my own life story. I still believe that, in not too many years, the instrumental value I can produce as a researcher will be replaced by AI. AI will be able to autonomously set research questions, organize existing ideas and generate new ones, create research plans, automatically execute them, and continuously refine the process, ultimately obtaining a series of nearly perfect research results. But so what? The original purpose of research lies in the exploration of the unknown, which is an experiential process. Whether one is the first in the world to prove a certain theorem, whether research results can be exchanged for a preeminent status in the field, how many times one is cited—all these are "metrics" that emerge under instrumental alienation. More importantly, even when chasing these external metrics, what one ultimately hopes for is to satisfy some dimension of inner experience through them. Since we are ultimately seeking a certain experience through our actions, the ultimate task for humanity in the AI era becomes clear—how can we, with the help of this new tool, AI, upgrade our own subjectivity? This is a task that each of us must bravely undertake.

从某种意义上，最开始的那句话是对的："当你觉得AI将要替代你的工作，那说明你工作的层次可能有点低。"但它的正确性不是来自于AI无法取代人类的某些高级能力，而是来自于一个彻底的视角反转：我们工作将不再是为了和别人或和AI竞争，而是为了建构自己的独一无二的故事。

In a sense, that initial statement was right: "When you feel AI is about to replace your job, it probably means your job is of a somewhat low level." But its correctness does not come from the idea that AI cannot replace certain high-level human abilities. It comes from a complete reversal of perspective: our work will no longer be for the sake of competing with others or with AI, but for the sake of constructing our own unique story.