基于大型语言模型的工具对电池研究的机遇与挑战  

Problem and perspective for battery researcher based on large language model

作　　者：吴思远 王雪龙[1] 肖睿娟[1] 李泓[1] WU Siyuan;WANG Xuelong;XIAO Ruijuan;LI Hong(Institute of Physics,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]中国科学院物理研究所,北京100190

出　　处：《储能科学与技术》2023年第3期992-997,共6页Energy Storage Science and Technology

基　　金：中国科学院信息化专项(CAS-WX2021SF-0102)。

摘　　要：近期,ChatGPT和GPT-3等大型自然语言模型的出现在学术界引发巨大热议;此外,Nature出版集团指出可以使用ChatGPT辅助文章撰写,这表明人工智能特别是自然语言处理将在学术文献领域引起巨大改变。对于电池领域来说,目前这些工具在电池以及储能领域有什么作用,它们在电池领域存在哪些问题以及如何改进这些问题尚未有文章讨论。本文在文献自动化整理与模型试用的基础上归纳了电池领域开展信息自动整理归类的问题与挑战、面对大型语言模型电池领域特别是储能从业人员如何面对以及学习,强调由于一些术语未按照标准化书写导致电池领域获取高质量数据集存在较大阻碍,这些将限制着电池研究中引入大型语言模型技术的发展。The Natural Language Process(NLP)models such as ChatGPT and GPT-3 have been discussed recently in academia and the Nature Publishing Group allows the authors to use ChatGPT to assist academic research.This means machine learning especially NLP has been integrated into the academia and will change the research paradigm.It exists opportunity and challenge for the battery researchers especially in replacing monotonous repetitive work.What can the researchers do for batteries,how to construct and use it to assist battery researching and the problem existing in it have not been discussed in details.Based on it,we write this perspective to explain above questions especially the following:①The problems existing in NLP models;②What can the battery practitioners do to meet these opportunities and challenges;and③How to learn the basic knowledge and construct battery model.All discussions are based on our recent works and the use of models and we hope it will offer initial guidance for battery researchers.

关 键 词：电池 自然语言处理 自动化 

分 类 号：O6-39[理学—化学]