从具身智能到具身智能体

    From Embodied Intelligence to Embodied Intelligent Agents

    • 摘要: 目前的人工智能的理论框架和算法充其量能实现一种具有多模态感知、可多场景执行任务的弱具身智能体。要实现真正的具身人工智能体,需要基于真正的具身智能的理论,探索具身人工智能的理论框架及实践进路。从梅洛·庞蒂的具身主体理论和瓦雷拉的生成认知可以发展出新的具身智能观,也为强具身智能体的发展提供理论基础和指导性原则。在此基础上,借助卡尔·弗里斯顿的自由能原理及主动推理理论进一步夯实具身智能的科学基础和技术原理。在这种哲学-科学的交叉视野中,可以进一步设想一种基于物理学、生物学和认知科学的具身人工智能体的技术原理和发展进路。

       

      Abstract: At present, the theoretical framework and algorithms of artificial intelligence can at best achieve the implementation of weakly embodied intelligence agent with multimodal perception and the ability to perform tasks in multiple scenarios. To achieve true embodied artificial intelligence, it is necessary to explore the theoretical framework and practical approach of embodied artificial intelligence based on the theory of true embodied intelligence. From Merleau Ponty's theory of embodied agents and Varela's generative cognition, a new view of embodied intelligence can be developed, providing a theoretical basis and guiding principles for the development of strong embodied intelligent agents. On this basis, with the help of Karl Friston's free energy principle and active Inference theory, the scientific foundation and technical principles of embodied intelligence have been further strengthened. In this interdisciplinary perspective of philosophy and science, a technological principle and possible approach for embodied artificial intelligence based on physics, biology, and cognitive science can be further conceived.

       

    /

    返回文章
    返回