探索ChatGPT模型能力:工业应用的前景和挑战  

Exploring the Capabilities of the ChatGPT Model:Prospects and Challenges inIndustrial Applications

作　　者：黄勃[1] 李文超 刘进[2] 刘瑾[1] 王文广 杨勇 赵晓丽[1] HUANG Bo;LI Wenchao;LIU Jin;LIU Jin;WANG Wenguang;YANG Yong;ZHAO Xiaoli(School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;School of Computer Science,Wuhan University,Wuhan 430072,Hubei,China;DataGrand Inc.,Shanghai 201203,China;Shanghai Baosight Software Co.Ltd,Shanghai 201203,China)

机构地区：[1]上海工程技术大学电子电气工程学院,上海201620 [2]武汉大学计算机学院,湖北武汉430072 [3]达观数据有限公司,上海201203 [4]上海宝信软件股份有限公司,上海201203

出　　处：《武汉大学学报（理学版）》2024年第3期267-280,共14页Journal of Wuhan University:Natural Science Edition

基　　金：科技部科技创新2030“新一代人工智能”重大项目(2020AAA0109300)。

摘　　要：人工智能技术的发展给人们的生活带来了极大的便利,自然语言处理技术在其中发挥着重要作用。基于Transformer神经网络架构的聊天生成预训练转换器(chat generative pre-trained transformer,ChatGPT),具有强大的语言理解和文本生成能力,可以根据用户输入的信息和问题,快速生成高度匹配用户意图的响应,其精度和反应速度相当出色。本文结合人工智能技术应用于工业领域的特征,介绍了ChatGPT在众多自然语言处理任务中展现出的非凡能力,展望了ChatGPT在工业应用中研发设计、生产运维和经营管理等环节的应用可能性,列举了当前ChatGPT应用于工业领域存在多方面限制,并提出改进方向,以便更好地满足工业应用的需要。The development of artificial intelligence technology has brought significant convenience to peoples lives,with natural language processing playing a crucial role.Based on the Transformer neural network architecture,the chat generative pre-trained transformer(ChatGPT),possesses robust language understanding and text generation capabilities.It can rapidly generate responses that closely align with user intent,based on the information and questions inputted by users.Its precision and response speed are remarkably outstanding.This article,integrating the characteristics of artificial intelligence applications in the industrial field,introduces the exceptional abilities of ChatGPT in a multitude of natural language processing tasks.It looks forward to the potential applications of ChatGPT in industrial contexts,such as in research and design,production operations,and business management.The article also enumerates the various limitations currently faced by ChatGPT in industrial applications and suggests several directions for improvement to meet the needs of industrial applications.

关 键 词：ChatGPT 深度学习 TRANSFORMER 大型语言模型 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]