Abstract: Large Language Models (LLMs), represented by ChatGPT, have found applications in certain research fields, raising interest in their impact on the careers of researchers. This paper uses GPT-4 to assess the potential career impact of large language models and LLM-driven systems on researchers, exploring the relationship between career characteristics and career impact in terms of knowledge, skills, education levels, and salary levels. The results indicate that researchers in different fields are affected by LLMs and LLM-driven systems to varying degrees. From a task perspective, LLMs have a significant impact on standardized and regulated tasks but less on non-standardized and specific operational tasks. In terms of knowledge, LLMs influence the humanities and sciences more than mechanical engineering disciplines. Regarding skills, LLMs have a greater effect on logical reasoning skills but less on interpersonal communication and mechanical operation skills. Additionally, LLMs tend to impact roles with higher education levels and salary levels more significantly. This demonstrates that LLMs can be utilized as productivity tools by researchers, helping to enhance labor efficiency and further accelerate the research process.

摘要： 以ChatGPT为代表的大语言模型已在部分科研领域得到应用，并引发了大语言模型对于科研人员职业影响的关注。本文探索利用GPT-4评估大语言模型或由大语言模型驱动的系统的应用对科研人员潜在职业影响，并从知识、技能、受教育水平、工资水平探究职业特征与职业影响之间的关系。结果表明：不同研究方向的科研人员会受到大语言模型及其驱动的系统不同程度的影响；从职业任务来看，大语言模型对于标准化、规则化的任务影响较大，对于非标准化、具体化的实操任务影响较小；从知识来看，大语言模型对于人文社科、理学类学科影响较大，对于机械工程类学科影响较小；从技能来看，大语言模型对于逻辑推理类技能影响较大，对于涉及到人际交往、机械操控类技能影响较小；从受教育水平和工资水平来看，大语言模型对于高受教育水平、高工资水平的职业影响程度更大。由此可见，大语言模型可以成为科研人员的生产工具应用到科研工作中，从而提高科研人员的劳动效率、进一步加快科研进程。