30 Months on the Exponential (in Charts)
打开互动全文版（逐段中英对照 + 针对本篇提问）→本文借助由智能体 AI 生成的图表，回顾了作者在两本书《共同智能》与《共同存在》之间的 30 个月里 AI 的进展。文章认为，AI 能力呈指数级但不均衡地提升：自 GPT-4 发布以来，已有 180 个模型超越它，基准任务的时间跨度屡次翻倍，成本大幅下降，而长篇写作等某些领域仍然薄弱。作者主张，这些进步正在改变人们与 AI 协作的方式，从基于聊天机器人的半人马与赛博格协作，转向一种“魔法师”模式——用户召唤智能体、分派任务并评估结果。对读者而言，结论是：理解这些趋势会让当下与近期未来不再令人意外；而人类通过深度知识、广度知识、品味与能动性仍保有优势，这应指导人们何时把工作交给 AI、何时避免这样做。
This article reviews the 30 months of AI progress between the author's two books, Co-Intelligence and Co-Existence, using charts generated with the help of agentic AI. It argues that AI capability has advanced exponentially but unevenly: since GPT-4's release, 180 models now outperform it, benchmark task horizons have doubled repeatedly, and costs have fallen sharply, while some areas such as long-form writing remain weak. The author contends that these gains are shifting how people work with AI, from chatbot-based centaur and cyborg collaboration toward a conjurer model in which users summon agents, assign them work, and evaluate the results. The conclusion for readers is that understanding these trends makes the present and near future less surprising, and that humans retain an advantage through deep knowledge, wide knowledge, taste, and agency, which should guide when to delegate work to AI and when to avoid it.
我的第二本关于 AI 的书《共存》（_Co-Existence_）今天出版（你可以在这里获取），距离 2024 年初《共同智能》（_Co-Intelligence_）的发布已有 30 个月。回顾这两本书之间的间隔，是了解 AI 进展的好方法。而且，多亏了 AI，我可以用一些非常精美的图表来描述这一进展。
My second book on AI, _Co-Existence,_ is out today (and you can get it here), 30 months after the release of _Co-Intelligence_ in early 2024. Looking back across the gap between these books is a good way to get a sense of AI progress. And, thanks to AI, I can describe this progress with some quite lovely charts.
制作图表本身就是事情发展程度的一个很好的指标。在聊天机器人时代，即《共同智能》的时代，我不得不在聊天窗口中与 AI 来回协作，引导它完成每一个新步骤。在我于《共存》中描述的当前智能体式 AI 时代，我把工作委托出去，让 AI 来处理。在这种情况下，我使用了 Claude Projects 中的 Opus 5.5，它允许你与一个“编排器”智能体对话，该智能体会根据需要衍生出额外的智能体，给它们指令并将信息转达给你。完成工作很直接。我给 AI 几句建议，它就在几分钟内完成了原本需要我数小时的工作（所有研究、所有图表、所有布局、所有图像、所有引用……）。而且由于我不必亲自做这些工作，我让 AI 为我制作了许多选项。我挑选了一些，并对其他选项给出了模糊的反应，这会让人类团队发疯（“让它更好”）。它甚至发现我数学错了。我原本打算把这篇帖子称为“指数上的 28 个月”。
Making the charts is itself a good indicator of how far things have come. In the era of chatbots, the era of _Co-Intelligence_, I would have had to work back-and-forth with an AI in a chat window, guiding it through each new step. In the current era of agentic AI that I cover in _Co-Existence_, I delegated the work and let the AI handle it. In this case, I used Opus 5.5 in Claude Projects, which allows you to talk to a single “orchestrator” agent that then spins off additional agents as needed, giving them instructions and relaying information back to you. Getting the work done was straightforward. I gave the AI a couple sentences of suggestions and it did what would have taken me hours of work (all the research, all the graphing, all the layouts, all the images, all the citations…) in minutes. And since I didn’t have to do the work, I had the AI make many options for me. I picked some and gave handwavy reactions to others that would drive a human team mad (“make it better”). It even caught that I’d gotten the math wrong. I was going to call this post “28 months on the exponential.”
一些与 Claude 交互以制作本文图表的例子。注意它纠正了我。
Some examples of interacting with Claude to make the graphics for this post. Notice that it corrected me.
我在上一本关于人工智能的书中写下了这句话，而且自那以后，它在每一刻都持续字面成立。从 2023 年初到 2024 年初，有一个 AI 模型无可否认是世界上最好的：OpenAI 的 GPT-4。这是大多数人在使用 AI 的头几年里记得的 AI 模型，是我在写《共同智能》（*Co-Intelligence*）时思考的那个模型，也是让 ChatGPT 成为许多人默认 AI 的那个模型。自那以后，已有二十个不同的 AI 模型先后占据领先地位，它们全都来自 Anthropic、Google 或 OpenAI。
I wrote this in my last book on AI, and it has continued to be literally true at every moment since. From early 2023 through early 2024, there was a single AI model that was undeniably the best in the world: GPT-4 from OpenAI. This was the AI model that most people remember from the first couple years of AI use, it was the one I thought about while writing *Co-Intelligence*, and it was the one that gave ChatGPT its status as the default AI for many people. Since then, twenty different AI models have held the lead at one time or another, all from Anthropic, Google, or OpenAI.
哪家公司拥有世界上最好的人工智能模型。
Which company had the best AI model in the world.
但领先者的变化本身掩盖了一个更大的变化。现在总共有 180 个 AI 模型比 GPT-4 发布时更好，其中许多是开放权重模型，任何人都可以自由使用或修改。其中十九个可以在现代智能手机上运行。
But the changing lead alone hides an even bigger change. A total of 180 AI models are now better than GPT-4 was at its release, many of which are open weights models, which anyone is free to use or modify. Nineteen of them can run on a modern smartphone.
下面的数据来自 METR 长时程基准（METR Long Horizon Benchmark），它曾经是 AI 领域最著名的图表。该基准衡量的是：在单次提示下，AI 能以至少 50% 的成功率完成多少人类工作量。该基准上的进步惊人地规律。从 GPT-4 发布到 METR 测量结束，AI 能完成的工作量翻了八倍。2024 年，AI 能完成一个四分钟的任务。而在我于 2026 年 4 月提交《共存》（_Co-Existence_）草稿的那一刻，AI 已能完成 17 小时的人类工作。
The data below draws on the METR Long Horizon Benchmark, which used to be the most famous chart in AI. The benchmark measured the amount of human work that an AI can do with at least a 50% chance of success from a single prompt. Gains on this benchmark have been startlingly regular. From the release of GPT-4 to the end of the METR measure, the amount of work AI could do doubled eight times. In 2024, an AI could do a four-minute task. At the moment I turned in my draft of _Co-Existence_ in April of 2026, an AI could do 17 hours of human work.
METR 表示，用他们当前的方法已无法准确衡量 AI 能完成的工作量。其他研究则表明，这个数字可能是 17 周的工作量，甚至更多。
METR said they could no longer accurately measure the amount of work an AI could do using their current methods. Other studies suggest that number might be 17 weeks of work, or more.
而且，即便能力不断提升，价格却在下降。我过去最喜欢的 AI 基准——Google Proof Question and Answer Exam，它给 AI 出的题目常常难倒研究生——过去要让 AI 给出一个正确答案，需要花费 30 美分的 AI 使用费。如今，得到同样的结果只需花费 0.04 美分。
And, even as ability has increased, prices have dropped. My old favorite AI benchmark, the Google Proof Question and Answer Exam, which gives the AI problems that regularly stump graduate students, used to cost 30 cents in AI usage to get a single correct answer. Now, it costs four hundredths of a cent to get the same result.
尽管取得了所有这些进展，AI 能力的前沿依然是参差不齐的，与 30 个月前并无二致。有些领域始终顽固地薄弱（AI 至今仍是个相当糟糕的长文写作者），而另一些领域，比如解数学题，则在同期从糟糕透顶跃升为超越人类。而且，正如我在书中所讨论的，即便在 AI 表现出色的地方，微小的差距也会对人类产生巨大的需求，至少目前如此。
Despite all the progress, the frontier of AI ability is still jagged, just as it was 30 months ago. Some areas have stayed stubbornly weak (AI is still quite a bad long-form writer), while others, like solving math problems, went from terrible to superhuman in the same period. And, as I discuss in the book, even where AI is excellent, small gaps create large needs for humans, at least for now.
但 AI 的进步如此之大，以至于我们与它协作的方式必须改变。在《共同智能》（_Co-Intelligence_）一书中，我写到了半人马式工作（由 AI 与人类分工）和赛博格式工作（由 AI 与人类融合）。两者都让你保持在回路之中……但你一次只能身处有限的几个回路。既然我们有了真正的智能体，我便补充了第三种与 AI 协作的方式：召唤师。召唤师在需要时召唤智能体，给它们分派工作，并评估返回的结果。本文中的图表就是这样制作的：我召唤智能体，给出一些总体指导，并大量地说“把它做得更好”。这是使用 AI 最强大的方式，但随着我们失去对 AI 究竟在做什么来为我们取得这些结果的控制，它也带来了新的风险。
But AI's gains have been so large that how we work with it has to change. In _Co-Intelligence_, I wrote about centaur work (divided between AI and humans) and cyborg work (blended between AI and humans). Both keep you in the loop… but you can only be in so many loops at once. Now that we have real agents, I added a third way to work with AI: the conjurer. Conjurers summon agents when needed, assign them work, and evaluate what comes back. The charts in this post were made that way: I summoned agents, gave some general guidance, and I said "make it better" a lot. It is the most powerful way to use AI, and it carries new risks as we lose control over what, exactly, the AI is doing to get us those results.
老读者会知道，过去几年我一直在用“飞机上用 Wi-Fi 的水獭”这一提示词测试 AI 图像生成和电影工具。这是最好的图像生成器 Midjourney 在 30 个月前生成的结果。
Regular readers will know that I have been testing AI image generation and movie tools with the prompt “Otter on an Airplane Using Wifi” for the past few years. This is what the best image generator, Midjourney, produced 30 months ago.
所以今天我向 Opus 5.5 提出了以下要求（包括拼写错误，随着 AI 变得更好，提示词已经变得不那么重要了）：“我有一篇文章展示了过去 30 个月 AI 的发展，以纪念我的新书（在原书出版 30 个月后出版）。我想让你制作一个终极的飞机上用 Wi-Fi 的水獭视频。就做一些奇妙的东西，我知道你很有创造力，试试吧。”使用我订阅的谷歌和 ElevenLabs 的各种在线服务，它生成了以下内容。（打开声音）
So today I asked Opus 5.5 the following (typos and all, prompting has become much less important as AIs get better): “I have a post showing what happened over the past 30 months of AI in honor of my new book (published 30 months after the original). I want you to make the ultimate otter on an airplane using wifi video. Just something fantastic, I know you are creative, give it a go.” Using my subscriptions to various online services from Google and ElevenLabs, it produced the following. (Turn the sound on)
视频有一些瑕疵。但 AI 自己弄清楚了如何制作它，从脚本到图像到概念到音乐再到创建这一切的技术方面，都没有我的任何帮助。这就是变化所在。
The video has a few flaws. But the AI figured out how to make it, from the script to the images to the concept to the music to the technical aspects of creating all of this, without any help from me. That is what has changed.
由于所有这些进展，AI 在世界范围内只会成为一个更大的议题。自《Co-Intelligence》出版以来的成就与争议时间线，能让你大致了解正在发生多少事情（我省略了数百个重要事件，但即便借助 AI 的力量，我们都难以跟上）。
As a result of all of these gains, AI is only becoming a bigger issue in the world. This timeline of achievements and controversies since *Co-Intelligence* came out gives you some idea of how much is happening (I am leaving out hundreds of important events, but, even with the power of AI, we are all having trouble keeping up).
当然，AI 的进步速度对我这样的作者来说是个问题，这就是为什么我的新书并不试图追赶新闻或最新模型发布。相反，它关注的是本文中提到的趋势：AI 正以指数级但不均衡的方式改进，而我们使用它的方式正从聊天机器人转向能完成实际工作的智能体。这本书探讨的是这对必须与这些系统共存的人类意味着什么：我们仍在哪些方面具有优势（我认为这归结为深度知识、广度知识、品味和能动性），何时将工作交给 AI，以及何时完全避免使用它。如果你理解了这些趋势，现在（以及不久的将来）就会变得不那么令人惊讶。
Of course, the speed of AI advancement is a problem for an author like me, which is why my new book does not try to keep up with the news or the latest model releases. Rather, it focuses on trends like the ones in this post: AI is improving exponentially but unevenly, and the way we use it is shifting from chatbots to agents that do real work. The book is about what that means for the humans who have to co-exist with these systems: where we still have an advantage (I argue it comes down to deep knowledge, wide knowledge, taste, and agency), when to hand work to AI, and when to avoid it entirely. If you understand the trends, the present (and near future) becomes much less surprising.
《Co-Existence》今日出版。你可以在 Opus 5.5 重新设计的网站上获取。我告诉它要做得更好。
*Co-Existence* is out today. You can get it here, on a website Opus 5.5 redesigned. I told it to make it better.