生命进化的宏大视角:从氧气危机到内共生
如果你和我一样,可能会对当下万物的巨变感到一丝无所适从(unmoored)。AI、社会结构乃至世界秩序的更迭速度快得惊人。为了理解两个女儿即将面对的未来,我没有看向前方,而是追溯到了 20 亿年前。
在那个时代,地球生命主要是单细胞生物。直到细菌演化出了光合作用(photosynthesis)并产生氧气。当时氧气其实是“毒药”,它破坏了几乎所有生命赖以生存的微妙化学环境,地球环境的改变速度超过了生物的进化速度。这导致了地球史上第一次大规模灭绝。但在那个濒死的世界里,发生了一件非凡的事:一个大细胞吞噬了一个小细胞,但并没有消化它,而是选择了合并(merger)。
这个小细胞变成了现在的线粒体(mitochondria),成为了复杂细胞的能量工厂。这次合并创造了巨大的能量盈余,支撑了后续一切进化:更大的细胞、复杂的身体和大脑。生物学家将此类时刻称为重大转型(major transitions):原本独立的实体停止竞争,开始构建新的整体。从分子到细胞,从细胞到个体,再到个体组成社会,进化阶梯上的每一步都是通过“合并”实现的。现在,我们正处于下一个重大转型的门槛:人类与 AI 的合并。
Original English Source
So if you're like me, you might be feeling at least a bit unmoored by how fast everything is changing. AI, our society, the world order. And that's just since this morning. I have two young daughters, and like a lot of us, I've been trying to make sense of the future they're growing up into. And what helped me make sense of it actually wasn't looking forward. It was going back, all the way back.
You see, two billion years ago, life on Earth was mostly single-celled until bacteria figured out a new trick, photosynthesis, which makes oxygen. Now at the time, oxygen was poison. It shredded the delicate chemistry that nearly all life on Earth depended on, and the planet changed faster than life could keep up with. Some scientists call what followed the first mass extinction event in Earth's history. But somewhere in that dying world, an extraordinary thing happened. A larger cell swallowed a smaller one, and instead of digesting it, they merged. The smaller cell became what we now call the mitochondria, the little powerhouse inside almost every complex cell on earth. That merger created an energy surplus so vast, it funded everything that followed. Larger cells, bodies, brains. Every breath you take is still powered by the descendants of that ancient partnership. That one accident in a dying world is the reason everyone in this room is alive today. Biologists call these moments major transitions when separate entities stop competing and start building a new whole. Like how molecules became cells, cells became bodies and individuals became societies. Every rung on that ladder was climbed through mergers. Now we're on the cusp of the next major transition, the merger of humans and AI.
AI 竞赛的囚徒困境与“顶级智能”的威胁
当我提出“我们要吃掉 AI”时,你可能会觉得荒谬或紧张。我在 AI 领域工作了 15 年,创办过早期的 AI 公司并卖给了谷歌。在一次私人聚会上,我询问那些正在构建主流 AI 系统的人:有多少人相信 AI 在未来 20 年内有超过 10% 的概率杀死大部分人类?几乎所有人都举手了。
这些开发者深知系统的危险,但他们陷入了一场无法停下的竞赛:任何放慢速度的人都会被超越。如果一家公司停下来,市场就会被别人占领;如果一个国家停下来监管,另一个国家就会领先。AI 就是我们时代的“氧气危机”,无论我们是否准备好,它都在逼近。
核心问题不在于 AI 会抢走什么工作,而在于:如果 AI 与我们保持分离,会发生什么?目前 AI 存在于屏幕之后,当你合上电脑,它消失了,但它在背后持续进化。如果保持分离,AI 就不是你的工具,而是你的替代品(replacement)。历史告诉我们,当新的顶级智能(apex intelligence)出现时——比如当初的智人——我们会让所有竞争智能走向灭绝。如果不实现合并,AI 就不是合伙人,而是宿敌。
Original English Source
Now I know what you’re thinking, OK, maybe you’re rolling your eyes. Maybe you're laughing, maybe you feel nervous -- it’s OK. I felt all of those things the first time I heard myself say it. So let me explain. For 15 years, I worked on building AI. I started one of the early AI companies. I raised a quarter of a billion dollars to do it, and I sold my business to Google. And not long ago, I was at a private event with many of the leaders building the AIs we all use every day. People you'd recognize. And I asked them how many believe there's more than a 10 percent chance that AI kills most of humanity in the next 20 years. Almost every hand went up. The people building these systems know how dangerous they are, but they're trapped in a race where anyone who slows down gets overtaken by someone who doesn't. If one company pauses for safety, another one takes the market. If a country stops to regulate, another one races ahead. Every AI founder has had the same conversation with themselves late at night. You lie there and you think, if I don't build this, someone worse will.
AI is the oxygen crisis of our era, and it's coming, whether we're ready or not. So what do we do? When a lot of people think about AI, they think about what it will do to us, what jobs it will take, what we should do to slow it down or regulate it. And those are important questions, but they're actually downstream of a much deeper question, which is what happens if AI stays separate from us? Right now, your AI lives on the other side of a screen. You ask it a question: it answers. You close your laptop and it's gone. But while your laptop is closed, the AI keeps getting better at your job. And if we stay separate, the AI is not your tool. It's your replacement. One that gets smarter and faster and cheaper every week. It doesn't take much to notice what happened the last time a new apex intelligence arrived here on Earth. That intelligence was us. And since we got here, we've driven to extinction every competing intelligence between Homo sapiens and Pan troglodytes. Today, we keep our closest animal ancestors, chimpanzees, in reserves for their protection from us. Without a merger, AI isn’t a partner -- it’s a rival.
隐形共生:人类与技术边界的逐步消解
事实上,“合并”并不是我们需要决定是否开始的事情,而是我们需要意识到自己已经身处其中。你什么时候不再记电话号码的?它们悄无声息地从你的大脑移到了口袋里。接着是日历,然后是各种细小的判断决策。当某些东西离开你的大脑,更好的东西取而代之:你不再校对拼写,转而专注于写作。
观察工具是如何不断向我们靠近的:从远处的大型机(mainframe),到桌面 PC,再到口袋里的智能手机、手腕上的手表、脸上的眼镜。每一步都在逼近思维的速度。边界已经模糊。瘫痪患者开始用思想打字;Neuralink 的植入者 Noland Arbaugh 说,使用它就像在使用“原力”,机器不再像机器,更像是他身体的一部分。
甚至 Face ID 技术也在演进,未来可能通过耳机或眼镜识别皮下极其微小的肌肉运动,从而实现“读心”。目前的脑机接口有 1,000 个连接,很快会有 1 万、10 万甚至 100 万。在 100 万个连接的量级,我们不再是修复丧失的功能,而是增加从未有过的能力:一下午学会一门语言,或者与朋友共享一段真实如亲历的记忆。这种未来不需要全新的技术,只需要现有技术的迭代。这种深度的整合(integration)将使我们很难分辨哪里是人类思维的终点,哪里是 AI 的起点。
Original English Source
The good news is a merger isn't something we need to decide to start. It's something we need to notice that we are already in. When did you stop remembering phone numbers? There was no moment you decided to forget them. They just moved from your head to your pocket. Your calendar probably went next. Then little judgment calls you used to make for yourself. The tool was great at it, so you let the tool do it. And while something left your head, a better thing took its place. You stopped checking your spelling and you started writing. You stopped remembering how to get there, and you started thinking about what you’d say when you arrived. And notice how we keep pulling these tools closer to us. The mainframe was in a whole other building. We put the PC on our desk, the smartphone in our pocket, the smartwatch on our wrist, smart glasses on our face. Every step closer to our minds, closer to the speed of thought.
And even that boundary is starting to blur. Right now, paralyzed patients are typing with their thoughts. Neural implants are restoring speech, vision and hearing to people who've lost them. Noland Arbaugh, the first person to receive a Neuralink brain implant, says that using it feels like using the force. The machine doesn’t feel like a machine -- it feels like him. And you may not realize it, but a technology we all use every day is learning to hear our thoughts. The Face ID system used to unlock your phone is being repositioned into headphones and glasses, where it can recognize microscopic muscle movements just beneath our skin, movements imperceptible to the human eye. The system that first learned to recognize us is now starting to see inside. Today, a brain implant has about 1,000 connections into the brain, and soon it will have 10,000, and then 100,000 and then a million. At 1,000 connections, you can restore movement. At 10,000, speech. At a million connections, you stop restoring what was lost and you start adding what was never there. Imagine learning a language in an afternoon, a new skill overnight. Maybe even sharing a memory with a friend and having it feel just as real to them as it felt to you. The thing about this future is it doesn't require new technology. It just requires more of the same technology. Someone you work with will get it first, and you'll hold out for a while, the way you did with the smartphone, but eventually you won't. The advantages of integration will be hard to compete with.
Think about what we even do when we use a computer today. You move a picture of an arrow around until it touches a picture of a folder. Inside, there are pictures of files. You click, you scroll, you drag. Sending a file to a colleague takes a whole minute. Two of those seconds were the decision. The rest of it is the equivalent of walking across your house to flip a light switch. With a system that can hear our thoughts, you skip the walk. The further this goes, the more deeply we integrate with AI, the harder it will be to tell where our thoughts end and AI begins. For example, what's the square root of 117 trillion? Go ahead, I'll wait. If you tried to answer that, you felt something. You felt a gap, a pause between the question arriving and anything starting to form. And you've lived your whole life inside that gap. Close it and the answer arrives instantly, the way you know your name, like a memory. One that bridges the distance between human and AI.
文明共存的终极法则:整体与部分的协同
我们选择合并,是因为被取代的后果更为可怕。但历史上的每一次重大转型都有一个先决条件:局部必须记住自己是整体的一部分。你体内的细胞生长通常是为了服务于你,但如果某个细胞忘记了它属于整体,开始无限制生长,那就是癌症(cancer)。癌症会在短期内成功,但最终会杀死宿主,随之而来的也是癌症自身的毁灭。
这个模式在所有尺度上重复。我们的文明(civilization)本身也是一种合并:它是数百万陌生人共同分享体制、牺牲和未来的隐形契约。随着 AI 的到来和世界的动荡,这个契约可能会磨损。失去生计的人会感到被遗弃,而保留生计的人可能会理所当然地袖手旁观。如果各部分在适应之前就破裂了,重大转型就会失败。
为了实现与 AI 的合并,我们必须首先保持彼此之间的合并。宇宙长久以来一直在进行这种演化试验:成功的合并留下了后代,失败的合并留下了化石。我希望我的女儿们,以及你们的后代,都能成为进化的赢家。请保持团结(hold together),不要沉溺于那种“你的对手沉没而你却能置身事外”的幻想。
Original English Source
Now I think we'll choose to merge because the alternative, being replaced, is far worse. But every major transition in the history of life has a condition: the parts have to remember that they are parts. A cell in your body wants to grow and replicate. And normally its growth serves you. Your cells grow so you can grow. But sometimes a cell forgets that it belongs to a whole. It starts growing without limit. And if your immune system fails to catch it, we call that cancer. The thing about untreated cancer is it succeeds for a while. The tumor grows, but eventually, the cancer kills the host, which kills the cancer. A part forgets the whole, and the whole dies, which kills the part. This pattern repeats at every scale.
Our civilization is itself a merger. It is the sometimes fragile, invisible agreement that millions of strangers will share institutions, sacrifices and a future. No one person built this system, and no one group controls it, but we all rely on it. And as AI arrives and the world gets more turbulent, every part of the society we depend on for our survival will be tempted to defect. People who lose their livelihoods will feel abandoned. People who keep theirs will feel entitled to look away. And bit by bit, the agreement frays. Major transitions fail when the parts break before they can adapt. And for us to make it to a merger with AI, we have to stay merged with each other. Major transitions fail when we don't make that leap. The thing about the future is we all have to share the same one, and we either all make it there together or we don't make it there at all.
Two billion years ago, the first merger gave us our cells, and the ones that followed gave us our bodies, our minds and our civilization. Every beautiful, difficult thing that followed, followed because the parts held together. Now our oxygen crisis is arriving, and it will not be gentle. Jobs will change and some will disappear. Institutions will shake and some will fail. Surviving this will take everything that we have. So here's what I'm asking from all of us, for every day from now on. Hold together. Do not indulge the fantasy that your side can let the other side sink and somehow stay dry. The universe has been doing this for a long time, and the mergers that worked left descendants. The ones that failed left fossils. I want my daughters to be descendants. And I want yours to be, too. Thank you.