Text/ Chen Xiaoping

Generative artificial intelligence represented by big models is bringing major changes and posing new and huge challenges to artificial intelligence governance. On the one hand, new phenomena and new characteristics have emerged in traditional AI ethics issues such as user privacy, data security, and algorithm fairness and transparency; on the other hand, new AI ethics issues have emerged rapidly. Adjustment of traditional governance models and exploration of new governance models have become the top priority.

Let’s look at a case first. For thousands of years, literary creation has been taken over by the authors and readers have not participated at all. The big model suddenly makes "online interactive works" technical feasibility. The content of this work is generated online through the interaction between the author and the reader. Each reader and the author jointly create a unique "personal version of the "same work". ”, most of the content generation can only be undertaken by the big model, and human authors and readers guide the big model to work through questions and prompts.

Online interactive works will eliminate the living space of traditional literature, subvert traditional creative methods, allow AI to "sky" in the production of spiritual products, end the history of human beings "covering the sky with one hand", and even create the possibility of "reverse guest as the main", bringing A series of unprecedented and far-reaching opportunities and challenges. For example, large models will be "unintentionally" faked due to technical defects, and may be deliberately faked by others; if there is excessive fiction in online interactive works, it can induce teenagers and children to recognize the truth and falsehood, and their recognition of history and reality. Knowing chaos; the proliferation of fiction can lead to serious degradation of human cognitive ability and survival ability... It is worth noting that the content generation ability of large models is not only suitable for literary and artistic creation, but also generally applicable to the generation of various texts, pictures and videos , and the many links in the education and service industries are enough to lead to a huge subversion of human society. For example, the current education system with employment and skills training as the basic goals will suffer an impact; the material and intangible production methods of society are facing drastic changes; under the efficiency-driven economic growth model, extraordinary efficiency growth accelerates the accumulation and sustainability of negative energy under the efficiency-driven economic growth model. If AI cannot open up new economic territory and maintain moderate employment stability, more and more people will be forced to become bystanders and epiphysicians...

This topic explores AI governance under major changes from different perspectives. "Big Model: From Basic Research to Governance Challenges" analyzes the technical characteristics and corresponding governance challenges of the big model from the perspective of basic research. "Research and Prospect for the Alignment of Values ​​of Large Language Models" explores the technological path of alignment of large models and human values ​​from three levels: universal values, group values ​​and individual values. "Analysis of ethical issues of generative artificial intelligence from the perspective of moral action" explores governance issues such as value embedding, responsibility correspondence, and ethical evaluation from the perspective of management. "A preliminary study on the first principle of large models and artificial intelligence ethics" examines the sustainability of AI development from the perspective of philosophy and social sciences, and proposes that "the purpose of the survival and development of human species should be "for the purpose of the survival and development of human species." "Ethical Challenges and Standardization Countermeasures for the Development of Large Model Industrialization" is based on an industrial perspective, analyzes ethical risks in several highly sensitive areas and explains governance countermeasures with standardization as the core.

The CAAI Artificial Intelligence Ethics and Governance Working Committee thanks the author of the article for his careful dedication and the editors of the topic. We look forward to readers' advice and jointly explore the future of AI governance. (to be continued)

Selected from "China Artificial Intelligence Society Communication"

Volume 14, 2024, Issue 1

Special topic on the new challenges of artificial intelligence ethics governance

Scan the QR code to join us