通用人工智能的复杂面孔  

The Complex Faces of Artificial General Intelligence

作　　者：刘永谋[1] Liu Yongmou

机构地区：[1]中国人民大学哲学院

出　　处：《哲学动态》2024年第9期31-36,126,共7页Philosophical Trends

基　　金：国家社会科学基金重大项目“现代技术治理理论问题研究”(编号21&ZD064)的阶段性成果。

摘　　要：围绕“何为通用人工智能”,不同领域所形成的技术话语差别很大。既有的技术话语缺乏对通用人工智能的一致性定义和衡量标准,从而陷入“唯名论困境”之中。宣传话语对通用人工智能进行言过其实的社会推广,在推动了新科技和人工智能产业发展的同时,亦导致不少问题。人文话语以愿景预测为基础,作为批判法来反思通用人工智能可能导致的技术风险和社会风险。因此,相关政策的制定者和决策者必须仔细审度各种通用人工智能话语,寻找适合中国国情的辩证发展之路。Since 2022,the emergence of Generative Artificial Intelligence(GAI)driven by large language models(LLMS)like ChatGPT has had a significant impact with regard to human intelligence.The impressive capabilities of these technologies often evoke concerns about non-human superintelligence with self-awareness,highlighting both their potential benefits and considerable risks.Artificial General Intelligence(AGI)is a key area of focus in this context.Liu Yongmou observes that AGI discourses exhibit remarkable variation across different fields,particularly regarding its definition and measurement.The current technical discourses on AGI lack a consistent definition,resulting in a"nominalist dilemma".While AGI promotional discourse emphasizes the potential societal benefits and technological advancements,it also raises various issues.AGI humanities discourse critically assesses the technological and social risks associated with AGI,which urges policymakers and decision-makers to carefully evaluate these discourses to identify a developmental path suited to China's national conditions.Liu Wei identifies three fundamental characteristics of AGI:it must perform unlimited tasks beyond those predefined by humans,independently discover tasks in context,and be driven by autonomous values rather than data alone.He argues that current technical challenges include limitations in mathematical and physical tools,technological,biological,and social bottlenecks,and the discrepancies between facts and values.The reliance on data,algorithms,computing power,and knowledge represents only a partial view of intelligence,which,in an open environment,cannot achieve the flexible,intentional calculations seen in humans.Liu suggests that the future of AGI lies in developing an intelligent ecology that integrates humans,machines,and their environments.Yang Qingfeng proposes that multitasking,multi-contextuality,and multimodality are key characteristics of AGI.Multimodality is seen as essential for AGI by multimodalists,though they often overlook issu

关 键 词：通用人工智能 技术话语 宣传话语 人文话语 

分 类 号：B15[哲学宗教—世界哲学] TP18[自动化与计算机技术—控制理论与控制工程]