要点二:有价值的观点都是反共识的,而这恰恰是 GenAI 无法提供的
我的理解
有影响力的工作几乎必然需要反共识且正确的判断——共识机会已经被人占走了,超额价值只来自于那些尚未被广泛认知的正确押注。GenAI 天然无法提供反共识观点:训练在互联网平均知识上的模型会系统性地压低低概率的卓越洞见,而这类洞见恰恰是最有价值的。这反而是个好消息:只要你能做到反共识且正确,你就是不可替代的。实践公式是:把重复繁琐的工作最大化地委托给 GenAI,用腾出的精力发展和构建属于自己的反共识判断。
相关链接
- Ch05-L04 研究 生成式 AI 建立在共识之上 — 共识机制是反共识价值命题的技术根源
- Ch05-L07 可落地的指南 哪些任务该交给 GenAI — 委托重复工作才能腾出发展反共识判断的精力
- Ch05-L05 猜想 GenAI 与人类有何不同 — 人类「尤里卡」与批判性思维正是反共识能力的来源
- Ch04-L04 学习要点 1 有选择地把任务委托给 AI — 委托策略是实现反共识公式的配套实践
原文
Lesson 44 of 68 要点二:有价值的观点都是反共识的,而这恰恰是 GenAI 无法提供的 / Takeaway 2: Valuable opinions are contrarian, which GenAI cannot offer
我们工作中有许多任务看起来很有影响力,但真正的影响力来自于推动进展,而进展通常需要你持有反共识的观点,并且是正确的。原因很简单——如果一个机会是共识,那它大概率已经被人抢走了。
一些有价值的工作示例:
发明了一种简化的方法来完成原本复杂的任务——对“哪些方法是必要的”持反共识观点
实现了别人认为不可能的事——对愿景持反共识观点
砍掉了一个浪费资源的项目——对价值持反共识观点
把现有业务应用到新的领域——对机会持反共识观点
这也是为什么当你向 GPT 提一些泛泛的问题,比如“我怎么才能变富”或“怎样让我的产品更好”时,GPT 可以给你一大堆看似有用的分析,却给不出任何真正有用的洞见。问题的核心在于:你这个问题的答案并不存在于 GPT 训练所依赖的互联网集体知识中。即使存在,只要它不是最常见的结果,GPT 也不会赋予它较高的概率,也就不会把它作为默认答案展示出来。
这是好事。这意味着,只要我们能做到反共识且正确,我们就是不可替代的。希望我们的课程能够帮助你做到这一点。公式其实很简单——
尽可能把重复、繁琐、非创造性的工作交给 GenAI
培养你的反共识观点,并把它构建出来
在动手构建中学习
English Original
There are lots of tasks in our work that seem impactful, but true impact comes from making progress, and progress usually requires holding a contrarian view and being correct. The reason is simple – if an opportunity is a consensus, it’s probably been taken already.
Some examples of valuable work
Invented a simplified method to complete a previously complex task – contrarian view on which methods are necessary
Achieved something that other people deem impossible – contrarian view on the vision
Killed a wasteful project – contrarian view on value
Applied the business to a new domain – contrarian view on opportunities
That’s why when you ask GPT generic questions like “how do I get rich” or “how to make my products better”, GPT can give you a lot of seemingly useful analysis, but not any actual useful idea. The core of this problem is that the answer to your question doesn’t exist in the collective knowledge of the Internet, which GPT is trained on. Or, even if it does exist, but is not the most common result, GPT won’t assign a high probability to it, and won’t show it as its default answer.
That’s good. That means we are irreplaceable, if we can be contrarian and correct. Hope our course can help you get there. The formula is simple –
Delegate the repetitive, tedious, non creative jobs to GenAI as much as possible
Develop your contrarian view and build it
Learn through building