技术洞察 3：从用户到构建者
我的理解
本课提出课程最核心的主题：从「被动用户」转变为「主动构建者」。以往我们被动依赖可用的应用，应用不够好就忍受，没有就放弃；AI 辅助编程让我们能在应用不足时改进它、应用不存在时自己构建，计算机从固定工具变成可主动挖掘价值的资产。这种心态转变的重要性甚至超过对提示工程的技术掌握——一个有构建者心态但零提示经验的人，往往能比熟知 AI 理论却没有动手意识的人获得更多价值。课程提示新用户通常经历「给太少太简单→给太多太难→退缩」几个阶段，而持续 tinkering（动手折腾）才是长期成长的唯一路径。
相关链接
- Ch01-L02 模块1中文 如何学习AI — 构建者心态的理论背景与比尔·盖茨 builder 案例
- Ch04-L01 模块 3 成为高效的 AI 管理者 — 构建者的进阶形态：将 AI 作为可委派任务的团队
- Ch02-L01 模块 2 深入解析使用 AI 构建的五个阶段 — 构建者在实践中会经历的五个具体阶段
原文
Lesson 6 of 68 技术洞察 3：从用户到构建者 / Technical Insight 3: From a user to a builder
能够用自然语言操作计算机是一种全新的范式，这很酷。但我们究竟该如何使用这一强大的工具？是不是干脆把所有事情都交给 AI，让自己彻底脱离鼠标？这是一个复杂的话题，取决于 AI 的能力以及每个人的工作流程。从宏观上看，AI 的新用户通常会经历几个阶段。
起初，我们交给 AI 的任务太少、太简单。随后，我们会逐渐过渡到给 AI 太多、太难的任务。当遇到困难时，我们可能会退回原地。关键在于持续动手尝试，因为这才是我们长期成长的方式。
本课程会兼顾两端，但本节以及前几个模块聚焦于最初的阶段——那时我们尚未意识到 AI 究竟能赋予我们多大的能力，更别提该在何处、如何使用它。
一个我们能够轻松编写程序的世界，与我们所熟悉的世界截然不同。想想我们目前使用计算机的方式，相当被动。我们能从计算机中获得多少价值，很大程度上取决于市面上有哪些应用可用。当某个应用能满足我们的需求时，我们就用它；当应用并不完全符合我们的需求、夹带广告或要求注册时，我们就忍受它；当根本没有可用的应用时，我们就接受现实并就此作罢。计算机对我们来说是一个既定的工具，我们只是被动地使用它。
但请回想一下我们在前几节课中所做的事情。JIRA 是一款出色的应用，但我们借助 AI 辅助编程让它变得更好用。是因为我们比 JIRA 的产品经理更聪明吗？未必。这是因为 JIRA 是为特定场景设计的，而编程则是通用的。这正是 AI 辅助编程在日常生活中拥有大量应用空间的原因——它打开了通往通用场景的大门。
除了带来与计算机沟通的新方式之外，这其实也意味着我们对计算机的态度发生了根本转变。在 AI 时代，当一个应用不完全符合我们的需求时，我们会去改进它；当根本没有可用的应用时，我们就自己构建一个。计算机突然变成了一种我们可以主动从中挖掘价值的资源。它成了一种资产、一位助手，正如我们在第三个模块中将谈到的，它甚至可以成为一支可以由我们委派任务的团队。借助 AI 辅助编程，我们不再只是计算机的使用者，而成为了构建者。
这种心态上的转变，其重要性可能甚至超过了对生成式 AI 的技术专长。假设有两个人：一个具备这种心态但毫无提示工程经验，另一个对 AI 研究非常了解却没有意识到自己可以成为构建者，那么前者反而更容易从 AI 中获得更多。这是因为生成式 AI 的使用极其讲究动手实践，而用生成式 AI 来构建东西，本身就是学习它的有效方式。对于生成式 AI 能做什么保持开放心态，保有好奇心，并真正尝试动手构建以完成你的任务，这就是你可以做的第一项练习。直接告诉 ChatGPT 你想构建什么，仅这个练习本身就能教会你很多。
希望这能让你更切身地感受到，作为一个构建者会有多么有趣、多么强大！当然，这并不意味着我们可以心安理得地跳过对 AI 工作原理的扎实理解。其中存在不少常见的误解和陷阱。建立对生成式 AI 内部机制的正确认识，能让你为一些意料之外、不尽如人意的行为做好准备，比如它会偷懒、会健忘、会忽略细节。这些内容将在下一个模块中介绍。
English Original
Having a new paradigm of operating computers using natural languages is cool. But how should we use this powerful tool? Should we just ask AI to do everything so we don’t need to touch our mouse? It’s a complicated topic that depends on AI’s capabilities and everyone’s workflow. From a high level, new users of AI will likely experience several stages.
First, we give AI too few and too easy tasks. Then, we gradually transition to giving AI too many and too hard tasks. When we encounter difficulties, we might step back. The key is to keep tinkering, as that’s how we grow in the long run.
This course will address both sides, but this lesson and the first few modules focus on the initial stage, when we don’t yet realize the extent of power we’ve gained from using AI, let alone where and how to use it.
The world where we can easily write programs is very different from the one we’re used to. Consider our current use of computers. It’s rather passive. How much we benefit from computers largely depends on the apps available to us. When there is an app that does what we want, we use it. When the app doesn’t do exactly what we want, has ads, or requires registration, we tolerate it. When there is no app available, we accept it and move on. Computers are a given tool to us, and we use them passively.
But think about what we just did in the previous lessons. JIRA is a great app, but we made it even better using AI-Assisted Programming. Is it because we’re smarter than the JIRA product managers? Not necessarily. It’s because JIRA is designed for specific use cases, while programming is general. That’s why AI-assisted programming has many applications in your daily life—it opens up the general use cases.
In addition to a new way of communicating with computers, it’s actually a mindset change in our attitude toward computers. In the AI era, when an app isn’t doing exactly what we want, we improve it. When there is no app available, we build one. Now computers suddenly become a resource from which we can proactively extract value. They become an asset, a helper, and, as we will touch on in the third module, a team we can delegate to. With AI-Assisted Programming, we are no longer mere users of computers; we become builders.
This mindset change is potentially more important than the technical expertise of GenAI. If there are two people, one with the mindset but no experience in prompt engineering, and the other who knows AI research very well but doesn’t realize they could be a builder, the former can easily get more from AI. This is because the usage of GenAI is very hands-on, and building things using GenAI is an effective way of learning it. Keeping an open mind about what can be done using GenAI, having curiosity, and actually trying to build things to complete your tasks is the first exercise you can do. Just tell ChatGPT what you want to build; that exercise itself can teach you a lot.
Hope this gives you a more tangible sense of how fun and powerful being a builder can be! Of course, this doesn’t mean we can safely skip a solid understanding of how AI works. There are quite a few common misconceptions and pitfalls. Setting up a correct understanding of the internal mechanisms of GenAI can prepare you for some unexpected and undesired behaviors, such as getting lazy, being forgetful, and failing to attend to details. These aspects will be introduced in the next module.