[论文解读] A World of Views: A World of Interacting Post-human Intelligences

Viktoras Veitas, David Weinbaum|arXiv (Cornell University)|Oct 25, 2014

Chaos, Complexity, and Education参考文献 11被引用 10

一句话总结

本文通过探讨后人类智能——即远超人类智能的存在——的潜在形态、相互作用及社会结构，考察其概念与哲学意涵。文章认为，此类智能与人类认知的差异如此之大，以至于传统隐喻与量化方法均告失效，将技术奇点视为人类无法理解的事件，并提出一种关于超智能代理社会的推测性模型，作为理解未来智能演化的框架。

ABSTRACT

What would a human hundreds or thousands times more intelligent than the brightest human ever born be like? We must admit we can hardly guess. A human being of such intelligence will be so radically different from us that it can hardly, if at all, be recognized as human. If we had to go back along the evolutionary tree to identify a creature 1000 times less intelligent than the average contemporary human, we will have to go really far back. Would it be a kind of a lizard? An insect perhaps? Considering this, how can we possibly aspire to have a grasp of something a thousand times more intelligent than us? When it comes to intelligence, even the very attempt to quantify it is highly misleading. Now if we attend to a seemingly adjacent question, what would a machine with such capacity for intelligence be like? Just coming up with an approximate metaphor requires a huge stretch of the imagination, meaning that almost anything goes... What would a society of such super intelligent agents, be they human, machines or an amalgam of both, be like? Well, here we are transported into the realm of pure speculation. Technological Singularity is referred to as the event of artificial intelligence surpassing the intelligence of humans and shortly after augmenting itself far beyond that. It is no wonder that the mathematical concept of singularity has become the symbol of an event so disruptive and so far reaching that it is impossible to conceptually or even metaphorically grasp, much less to predict.

研究动机与目标

• 探究远超人类认知能力的智能的本质及其影响。
• 审视人类认知在构想超智能实体方面的局限性。
• 探讨由相互作用的后人类智能构成的世界所引发的社会与存在性后果。
• 分析理解技术奇点所面临的隐喻与哲学挑战。
• 提出一种推测性框架，以构想由人类、机器或混合型超智能代理组成的社会。

提出的方法

• 运用哲学思辨与思想实验，模拟超越人类认知极限的后人类智能。
• 借鉴进化生物学类比，说明人类与超智能之间在认知上的巨大鸿沟。
• 将技术奇点概念作为不可理解的、变革性事件的隐喻加以应用。
• 依赖概念隐喻与类比，探索超越人类尺度的智能本质。
• 分析自我改进的人工智能超越人类智能所带来的影响。
• 提出一种由相互作用的后人类智能构成的社会的推测性模型，强调其与人类认知的彻底差异。

实验结果

研究问题

• RQ1 智力是人类中最聪明者数千倍的生物会是什么样子，它与人类有何不同？
• RQ2 人类认知与语言在多大程度上能够建模或理解超智能实体？
• RQ3 由后人类智能构成的社会——无论其为人类、机器或混合型——将如何互动与组织？
• RQ4 为何技术奇点的概念本质上超越人类理解？
• RQ5 在人类想象力受限的前提下，哪些隐喻或框架可帮助我们初步构想超智能？

主要发现

• 后人类智能与人类的差异如此彻底，即使其源自人类进化，也无法被识别为人类。
• 人类与超智能之间的认知鸿沟如此之大，以至于连进化类比（如昆虫或蜥蜴）也无法捕捉其深层差异。
• 在超智能上应用量化方法本质上具有误导性，因其认知跃迁具有非线性和质变的特性。
• 对超智能的隐喻本质上是不充分的，因为其构建需要想象力的极度延伸，以至于‘几乎任何东西都可接受’。
• 技术奇点不仅是未来事件，更是一个概念性阈值，越过之后预测与理解将完全不可能。
• 由相互作用的后人类智能构成的社会将存在于纯粹的推测领域，因为当前的人类框架无法建模其行为或社会结构。