突破大模型的天花板:物理语言在下一代人工智能中的作用  

Breaking the Ceiling of Large Language Models:The Role of Physical Language in Nextgeneration Artificial Intelligence

作　　者：杨华 陈俭 YANG Hua;CHEN Jian(Shenzhen ELS Technology CO.,Ltd.,Shenzhen Guangdong 518000)

机构地区：[1]深圳市企企通科技有限公司,广东深圳518000

出　　处：《中国科技纵横》2024年第15期20-23,共4页China Science & Technology Overview

摘　　要：本文探讨了大模型AI在自然语言处理和机器学习领域的成功应用及局限性,尤其是在智能转化率、精度、可控性和能耗方面的潜在上限。基于符号化系统的大模型AI可能正面临智能发展的天花板。为突破这一限制,本文提出了“物理语言”概念,物理语言是一种能够直接与自然规律对话和交互的语言系统,区别于传统符号化语言。物理语言通过直接表达和操控物理世界信息,为智能系统提供了更高效的处理方式,能够显著提升智能系统的转化率、精度和可控性,同时降低能耗。此外,本文进一步探讨了物理语言在工业控制、自动化系统、机器人技术等领域的应用,并论证了其在突破大模型AI局限性和推动通用人工智能(AGI)发展中的关键作用。This paper explores the successes and limitations of large model AI in the fields of natural language processing and machine learning,particularly in terms of its potential upper limits regarding intelligence transformation efficiency,accuracy,controllability,and energy consumption.The article points out that AI based on symbolic systems may be approaching an intelligence development ceiling.To address this limitation,the paper introduces the concept of"physical language,"defining it as a language system that can directly interact and communicate with the laws of nature,in contrast to traditional symbolic languages.Physical language enables more efficient processing by directly expressing and manipulating information from the physical world,significantly improving intelligence systems'transformation efficiency,accuracy,and controllability while reducing energy consumption.The article further explores the application of physical language in industrial control,automation systems,and robotics,and demonstrates its key role in overcoming the limitations of large model AI and advancing the development of Artificial General Intelligence(AGI).

关 键 词：大模型AI 物理语言 智能天花板 自然语言处理 符号化系统 AGI 因果推理 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]