AI促进科学产出和影响,但缩小研究课题的范围,研究发现。 AI boosts scientific output and impact but narrows research topics, study finds.

芝加哥大学和清华大学的研究发现,使用人工智能的研究人员出版的论文比论文多3.02倍,提前1.37年更快,比没有人工智能的同龄人多4.85倍,获得的引文也比没有人工智能的同龄人多4.85倍。 Researchers using AI publish 3.02 times more papers, advance 1.37 years faster, and receive 4.85 times more citations than peers without AI, a study by the University of Chicago and Tsinghua University finds. 通过分析六个自然科学领域的超过4100万篇科学论文,该团队使用基于谷歌的模型和人类验证识别了310,957篇人工智能增强的出版物. Analyzing over 41 million scientific papers across six natural science fields, the team identified 310,957 AI-augmented publications using a Google-based model and human verification. 虽然AI提高了生产率和承认度,但它也在缩小研究范围,将研究专题的范围缩小了近4.63%,这可能是由于AI依赖大型现有数据集限制在记录较少的地区进行勘探。 While AI boosts productivity and recognition, it is also narrowing the scope of research, reducing the range of topics studied by nearly 4.63 percent, likely due to AI’s reliance on large existing datasets that limit exploration in less-documented areas. 这些结果发表在《自然》上,是在AI日益融入全球科学,包括斯坦福第一届AI审查会议和澳大利亚国家AI计划的背景下产生的。 The findings, published in Nature, emerge amid growing global integration of AI in science, including Stanford’s first AI-reviewed conference and Australia’s National AI Plan.