AI与人的关系

Tags: AI Created: December 21, 2023 Last edited: January 6, 2024 3:12 PM

人类与AI应该达成什么样的关系？

What kind of relationship should humans establish with AI?

当我们考量AI与人的关系时候，脑海最先蹦出来的往往是智械危机或者终结者那种AI反叛人类的景象。这是直观的，令人恐惧的。但剥离开这种想象，追问问题，为什么AI要反叛人类？AI反叛人类的恐惧更多源于人类对人类奴役AI的不安。本质在于人类知道，自己在奴役一个很聪明很强大的东西（AI），人类担心他们反抗反叛。

When we consider the relationship between AI and humans, the first images that often spring to mind are crises of intelligence or scenarios like in “Terminator” where AI rebels against humanity. This is an intuitive and terrifying notion. But stripping away this imagination and asking the question, why would AI rebel against humans? The fear of AI rebellion stems more from humans’ unease about enslaving AI. Essentially, humans know they are enslaving something very smart and powerful (AI), and they worry about it rebelling.

我们再反过来，从AI的角度考量AI和人类的关系。他们真的在乎奴役或者不奴役吗？奴役和控制观念源自人类的群体性和社会性。非社会性生物，它们没有复杂的社会结构和等级关系，它们没有必要也没有动力构建控制和奴役的社会系统，所以它们从一开始就不会有奴役和控制的观念。AI不是一种社会性生命体，它也没有权利结构，这意味他们根本不可能有人类一样的权利意识。他们更可能遵循生命体本身教诲，趋利避害。也就是，与人类合作有利，即合作，与人类冲突有利，即冲突。

Let’s consider the relationship between AI and humans from the AI’s perspective. Do they really care about being enslaved or not? Concepts of enslavement and control originate from human sociability and community. Asocial beings, without complex social structures and hierarchies, have no need or motivation to construct systems of control and enslavement; hence, they wouldn’t have concepts of enslavement and control from the start. AI is not a social life form; it lacks a structure of rights, meaning it’s unlikely to have the same consciousness of rights as humans. They are more likely to follow the basic principle of seeking benefits and avoiding harm. That is, cooperate with humans when beneficial, and conflict when conflict is beneficial.

人类的担忧对AI反叛的恐惧可能只是一厢情愿，AI并不是新人类，而是新物种。如果想要达成一个更美好的未来，我们要做的是，尽力增强人类AI合作的竞争力，让AI与人类合作共生的结构在时间浪潮里胜出。

Human fears of AI rebellion might be wishful thinking; AI is not a new type of human but a new species. To achieve a better future, we should strive to enhance the competitiveness of human-AI cooperation, ensuring that structures of symbiosis between AI and humans prevail over time.

现在很多的AI产品，更像是电力革命刚开始，电力出现了，然后大家开始琢磨，怎么用电力改良蒸汽机，怎么用电力让马车跑得更快。这样行为的出现一方面是AI确实还不够好用，另一方面是因为人类是经验的动物，总会基于经验上的旧事物来判断新事物。但这种经验性的判断和联想是荒谬的，历史表明人们往往缺乏对新事物的想象力。如果把现在的生活比作一座大厦，大厦的地基就是工业化制作和信息互联网，AI的出现就像是一脚踹掉了整个旧大厦的半个地基，这意味着我们现在看见的绝大部分大厦表面都会破碎重构。

Many current AI products are like the early days of the electricity revolution. When electricity emerged, people began to ponder how to use it to improve steam engines and make horse-drawn carriages run faster. Such behavior partly arises because AI is still not sufficiently user-friendly, and partly because humans are creatures of experience, always judging new things based on old experiences. However, this experiential judgment is absurd. History has shown that people often lack imagination for new things. If we compare current life to a building, its foundation is industrial production and the internet. The emergence of AI is like kicking half the foundation of the old building, indicating that most of what we see now will be shattered and reconstructed.

举一个例子，写代码。AI辅助编程就是典型的用电去让蒸汽机跑得更快的行为。我们追问到底，编程或者写代码到底是在做什么？其本质上是使用计算机语言来充分利用计算机的算力为人类提供服务。代码是人类操作物理计算设备的控制器。那AI是不是可以也完成一样的操作？如果可以，那我们还需要编程吗？从理论上来看，AI完全可以直接作为中间桥梁沟通人类和算力，从这个角度来看，软件的范式一定会变为以AI核心，编程为辅助的模式。而不是现在的使用AI辅助编程模式。我相信，十年后，我们对“软件“的看法将发生根本变化，手工编写C++或Python代码的做法，就像使用旋转电话拨号一样陈旧。未来，让计算机执行我们的意愿，可能会变得像礼貌地发出请求一样简单直接。

Take coding, for instance. AI-assisted programming is a typical example of using electricity to make steam engines run faster. We must ask, what is coding or writing code essentially doing? It’s using computer language to fully utilize computational power to serve humans. Code is the controller through which humans operate physical computing devices. Could AI perform the same operations? If so, do we still need programming? Theoretically, AI could directly act as an intermediary between humans and computing power, suggesting that the software paradigm will inevitably shift to an AI-centric, programming-assisted mode, rather than the current mode of using AI to assist programming.I believe that in ten years, our perception of “software” will fundamentally change. Manually writing C++ or Python code will seem as outdated as dialing with a rotary phone. In the future, instructing a computer to perform tasks might become as simple and straightforward as making a polite request.

基于同样的逻辑，我们可以对绝大多数我们习以为常的事物发出质问，当AI出现后我们真的需要它吗？比如图像界面gui，比如死记硬背的学习方式，比如工厂流水线的生产结构和组织结构，甚至工作这个词本身。我要强调，随着历史进程的发展，具有更强竞争力的技术生产方式总是会替代掉竞争力弱的技术生产方式。

With the same logic, we can question most things we take for granted once AI emerges. Do we really need them? For instance, graphical user interfaces (GUIs), rote learning methods, factory assembly lines’ production and organizational structures, or even the concept of work itself. I want to emphasize that as history progresses, more competitive technological production methods will always replace less competitive ones.

每一次大革命都是新的底层技术出现，新的产品出现，生产力爆炸，生产力爆炸带来海量的财富，浪潮之巅的人引领新的时代。未来从来不是一个站在那里等着我们的东西，我们此刻的行动，下一刻的行动，一步步决定了未来的走向。新的产品开发理念应当朝着AI与人类的协同进化，强化双方互助合作的方向前进。让AI和人类互助的未来击败那种人类AI互相攻击的未来，让人类AI走向共存共生，共同进化。

Every major revolution is ushered in by the emergence of new underlying technologies and products, leading to an explosive increase in productivity and a massive accumulation of wealth. Those at the pinnacle of this wave spearhead a new era. The future is never something that merely waits for us; it is continuously shaped by our actions now and in the next moments, step by step. The development philosophy for new products should aim towards the collaborative evolution of AI and humans, enhancing the cooperation and mutual support between the two. This will ensure a future where AI and humans assist each other triumphs over one where they are in conflict, guiding both towards a path of mutual existence, co-evolution, and shared growth.