Large language models (LLMs) and dialogue agents have existed for years, but the release of recent GPT models has been a watershed moment for artificial intelligence (AI) research and society at large. Immediately recognized for its generative capabilities and versatility, ChatGPT's impressive proficiency across technical and creative domains led to its widespread adoption. While society grapples with the emerging cultural impacts of ChatGPT, critiques of ChatGPT's impact within the machine learning community have coalesced around its performance or other conventional Responsible AI evaluations relating to bias, toxicity, and 'hallucination.' We argue that these latter critiques draw heavily on a particular conceptualization of the 'human-centered' framework, which tends to cast atomized individuals as the key recipients of both the benefits and detriments of technology. In this article, we direct attention to another dimension of LLMs and dialogue agents' impact: their effect on social groups, institutions, and accompanying norms and practices. By illustrating ChatGPT's social impact through three disruptive events, we challenge individualistic approaches in AI development and contribute to ongoing debates around the ethical and responsible implementation of AI systems. We hope this effort will call attention to more comprehensive and longitudinal evaluation tools and compel technologists to go beyond human-centered thinking and ground their efforts through social-centered AI.

翻译：大型语言模型（LLM）和对话式人工智能代理已存在多年，但近期GPT模型的发布成为人工智能（AI）研究乃至整个社会的分水岭时刻。ChatGPT凭借其强大的生成能力和跨技术、创意领域的卓越表现迅速获得广泛采用。当社会在应对ChatGPT带来的新兴文化冲击时，机器学习领域对其影响的批判主要聚焦于性能表现或与偏见、毒性及"幻觉"相关的传统负责任AI评估。我们认为，后者批判在很大程度上依赖于"以人为本"框架的特定概念化——该框架倾向于将原子化的个体视为技术利弊的主要承载者。本文致力于关注LLM与对话式代理影响的另一维度：它们对社会群体、制度体系及相应规范与实践的效应。通过三个颠覆性事件阐释ChatGPT的社会影响，我们挑战了AI开发中的个体主义方法论，并为围绕AI系统伦理与负责任实施的前沿辩论提供新视角。期望本研究能引发对更全面、纵向评估工具的重视，促使技术工作者超越人类中心思维，以社会中心AI为根基推进实践。