生成式人工智能的主体性问题

Reconceptualizing Agency with Generative Artificial Intelligence

生成式人工智能在与人类交互中,表现出与以往人工智能截然不同的自主交互能力,呈现出一定的主体性特征,形成了全新的人机交互形态,给已有的主体性概念带来了挑战。从主体交互的视角看,生物有机体的主体性由自组织驱动的个体与环境交互所定义,人类主体性由情绪、语言和文化驱动的个体与自然、社会文化环境交互所定义,生成式人工智能在与人类交互中呈现的主体性则是由语言生成驱动的交互能力所定义。由此,主体性可视为参与交互实体的潜在行动能力,由交互的驱动力因素所决定。这种交互主体性不再是传统的基于主体来推定主体性,也不需要所有交互实体都具有意向性,而是包含了交互实体的目标导向性和自主性,展现出一种新的人工主体性形式,为主体性概念赋予了新的内涵。

Generative Artificial Intelligence,in its interaction with humans,demonstrates an autonomous interaction capability that is distinctly different from previous forms of artificial intelligence,thus exhibiting the characteristics of agency and forming a new pattern of human-machine interaction.This challenges the traditional concept of agency.From an interactive perspective,the agency of biological organisms is defined the interactions between the individual and the environment,driven by selforganization.Human agency is defined by the individual's interaction with the natural and socio-cultural environment,driven by emotions,language,and culture.The agency exhibited by Generative Artificial Intelligence in its interaction with humans is defined by the interaction capability driven by language generation.Agency can thus be viewed as the potential action capacity of the entities participating in the interaction,determined by the driving forces of the interaction.This form of interactive agency is not derived from the concept of the agent in the traditional sense,nor does it necessitate that all participants in the interaction possess intentionality.Instead,interactive agency encompasses the goal-directedness and autonomy of the entities engaged in the interaction,representing a new form of artificial agency and providing new semantics to the concept of agency.