作　者：王天恩
关键词：数字智能；ChatGPT；Sora；通用性；通用化；人工智能
摘　要：

以ChatGPT 和Sora 等为标志，人工智能全面步入数字智能发展阶段。数字智能基于统计模型通过大数据纳入人类语境/世界情境进行自然语言和视频处理，获得了在既定大数据范围内甚至超越人类个体的自然语言和视频处理及超强泛化能力，促成了人工智能通用性的空前发展，引发了生成式人工智能是否步入通用人工智能轨道甚至已经通用化的观点对立，关于数字智能的发展由此陷入通用性和通用化构成的通用迷思。数字智能的通用迷思根源于其通用性极度扩展，而人工智能的“通用性”和“通用化”是两个不能通约的概念，即使数字智能通用性无限叠加，也不会走向人工智能的通用化，其中涉及两种不同范式的根本转换。作为信息编码层次的人工智能，数字智能的意义不在于人工智能的通用化，而在于从自然语言到广义语言、从大数据纳入人类语境到纳入世界情境的通用性发展。

 

The General Myth of Digital Intelligence

Wang Tianen（Shanghai University）

Abstract Marked by advancements such as ChatGPT and Sora, artificial intelligence (AI) has fully entered the stage of digital intelligence development. Digital intelligence, by processing natural language and video upon statistical models through integrating vast amounts of data into human contexts and world scenarios, has acquired the processing capabilities that, within the scope of the given big data, can even surpass those of individual humans, along with an extraordinary capacity for generalization. This has led to unprecedented advancements in the generality of AI and sparked debates about whether generative AI is on the path to artificial general intelligence(AGI), or even has already achieved generalization, thus creating a generalization myth surrounding the development of digital intelligence. The myth stems from the extreme expansion of its generality, while the concepts of AI generality and generalization are fundamentally incommensurable. Even if the generality of digital intelligence continues to grow, it will not lead to the generalization of AI, which will involve a fundamental shift between two different paradigms. The significance of digital intelligence, as an AI operating on the level of information encoding, lies not in the generalization of AI, but in its development of its generality, which spans from natural language to language in broader sense and from incorporating big data into human contexts to encompassing world scenarios.

Key words  digital intelligence; ChatGPT; Sora; generality; generalization; artificial intelligence

 

■  作者简介 王天恩，上海大学智能时代的马克思主义研究中心、智能哲学与文化研究院教授，上海 200444。

[PDF]（下载数：）