您当前的位置:首页 > 论文详情

表意Al:基于汉字形根体系的Token困境破解范式

Logographic AI: Resolving the Token Dilemma through Chinese Character Morpho-Root System

摘要: 当前人工智能(Al)面临字母符号体系引发的认知瓶颈:Token的离散性阻碍对物理世界的连续建模。本文提出基于汉字形根系统的中国范式AI理论,通过形义拓扑结构与卷积神经网络的天然适配性,实现符号-感知-物理的统一认知架构。研究表明:(1)汉字214个形根的视觉语义一体性突破传统Token的无意义性,其空间拓扑结构(如"氵+木=沐")直接编码物理约束;(2)形构熵驱动的推理机制支持跨模态联合预测,为构建"世界模型"提供非西方技术路径。本范式在古籍识别与机器人指令理解任务中验证了其技术优势,为突破AI认知天花板提供文明级解决方案。

Abstract: Current AI development faces fundamental cognitive bottlenecks caused by alphabetic token systems: The discreteness of tokens hinders continuous modeling of the physical world. This paper proposes a Chinese paradigm AI theory based on the morpho-root system of Chinese characters, achieving unified symbolic-perceptual-physical cognition through the natural compatibility between glyph topological structures and convolutional neural networks (CNNs). Research demonstrates: (1) The visual-semantic unity of 214 Chinese character roots (e.g., "氵" encoding fluid dynamics) overcomes the meaninglessness of conventional tokens; (2) Morpho-structural entropy-driven reasoning enables cross-modal joint prediction, providing a non-Western technological path for building "world models". Validated in classical text recognition and robotic command comprehension tasks, this paradigm offers a civilizational-level solution to break through AI’s cognitive ceiling.

版本历史

[V1] 2025-04-01 14:40:20 PSSXiv:202504.00172V1 下载全文
点击下载全文
在线阅读
许可声明
metrics指标
  •  点击量64
  •  下载量18
  • 评论量 0
评论
分享
邀请专家评阅
收藏