Social Sciences in China (Chinese Edition)

No. 3, 2024

Generalization of Artificial Intelligence and How to Achieve It

(Abstract)

Wang Tianen

The development of large language models, symbolized by ChatGPT, vividly illustrates the two distinct levels of artificial intelligence advancement thus far: artificial intelligence operating at the level of human knowledge and artificial intelligence functioning at the level of information encoding. As an artificial intelligence system operating at the level of information encoding, ChatGPT embodies not only the successful integration of big data paradigms into human contexts and its profound impact on the digitization of machine learning, but also signifies a strategic retreat of natural language machine understanding. The development of artificial intelligence has now entered a phase characterized by the deep integration of humans, machines, and big data, where the universality and generalizability of artificial intelligence are incommensurable. This marks the emergence of a third level of artificial intelligence development, namely artificial intelligence at the information level. The unfolding of these three levels of AI development paints a comprehensive picture of the human-machine developmental relationship: the progression of artificial intelligence mirrors the inverse trajectory of biological intelligence evolution, while the bidirectional interaction between the two establishes a universal mechanism for artificial intelligence advancement.