社会化人工智能的适应性

当前学界对社会化人工智能的探讨较少关涉人工智能所面对的社会化是一个怎样的议题,并缺乏明确的判定标准。随着人工智能的技术进步,人工智能越来越深地嵌入到人、机、环境三重交互的社会化环境中,实现人工智能的社会化是人工智能发展的核心诉求之一,即从求解结构化的简单命题过渡到动态适应非结构化的不完全信息动态博弈命题。具身人工智能凭借身体优势,是实现人工智能从实验室到社会化的可能路径之一,其多功能性和具身交互能力为应对社会化不完全信息动态博弈提供了在环境中摄取多样信息的能力,以及解决复杂问题的多样性方案。通过举例深度进化强化学习,证明具身性为人工智能提供了与现实交互和获取多样信息的通道;社会化算法和多功能方案则证明,具身人工智能的发展应当增强自身的多模态信息处理能力、解决环境中的多模态任务,实现其在社会化环境中适应性的飞速发展。因此应当大力发展如传感器、云计算、材料科学、运行算法等具身人工智能相关产业,同时更重要的是,应形成具身人工智能构造理论。

Research on adaptive of social artificial intelligence

To realize the socialization of AI is one of the core demands of AI development.Embodied artificial intelligence is the most possible path to realize artificial intelligence from laboratory to socialization. its versatility and embodied interaction ability are the best solution to deal with the dynamic game of social incomplete information. Deep evolutionary reinforcement learning proves that embodiment provides AI with a channel to interact with reality and obtain diverse information. Social algorithms and multi-function solutions prove that the best way to achieve embodied AI is to solve multiple tasks, which can greatly improve its adaptability in society.