Civilizing and Humanizing Artificial Intelligence in the Age of Large Language Models

The advent of large language models has significantly influenced the development of artificial intelligence (AI), expanding its reach beyond researchers to the general public via practical tools. These models demonstrate remarkable capabilities in complex tasks but also present limitations and risks of misuse. In response, regulatory frameworks from the United States and the European Union have proposed measures to ensure AI safety, introducing notions such as Constitutional AI to ensure adherence to these guidelines. This special issue introduces and covers the following two interweaving research directions toward AI safety. First, civilizing AI to balance AI’s ability to generate human-like outputs and mitigate adverse behaviors such as hallucinations and unintended biases, and second, humanizing AI to align AI systems’ behavior with human ethics, sociocultural norms, values, and regulations, ensuring they meet societal expectations similar to those for human roles such as drivers or health professionals.