| 0 | 0 | 0 |
| 下载次数 | 被引频次 | 阅读次数 |
大语言模型以另类的数据算法处理能力雄踞“新”人工智能的学科中心,在学科内部掀起范式革命。当科学家用“新”人工智能来做科学时,这些具有另类数据算法处理能力的数智机器从科学研究的成果转变为科学研究的基础。这两股力量交织构造了人工智能驱动另类科学的最初形态。科学知识生产不再由科学家独立地在头脑中完成认知任务,而是将该认知任务分布在数智机器上甚至延展到社会试验中,最终与智动科学家形成全新的科技创新生态。这种新的共生关系,筑成了人工智能创新世的另类科学知识生产形态,观察其中的科学观和研究方法转型,有助于审慎地看待和思考这些变化带来的挑战和图景。
Abstract:Large language models are in the core role of the New AI discipline with their alternative data algorithm processing capabilities, triggering a paradigm revolution within the discipline. When scientists use New AI for science, these digital intelligent machines(DIMs) with these capabilities have shifted from the results of scientific research to the basis of scientific research. These two forces jointly constructed the primary form of the AI for alternative science. The scientific knowledge production no longer involves scientists independently completing cognitive tasks in their own minds. Instead, the cognitive tasks are distributed across DIMs and also extended to social experiments, forming a new technological innovation ecosystem with agentic scientist. This new symbiotic relationship has constructed the alternative scientific knowledge production form of this era. Observing these changes in the view of science and research methodology, may help to prudently view and think about the challenges and prospects brought by these changes.
[1] 陈小平.双智社会的颠覆性变革前瞻[J].自然辩证法研究,2025(9):26-33+50.
[2] Wei J,Yang Y,Zhang X,et al.From AI for Science to Agentic Science:A Survey on Autonomous Scientific Discovery[EB/OL].(2025-08-18)[2025-08-30].http://arXiv.org/abs/2508.14111v1.
[3] 马克思恩格斯文集:第1卷[M].北京:人民出版社,2009.
[4] 钟义信,何华灿,汪培庄,等.人工智能:范式革命与通用理论[M].北京:科学出版社,2025.
[5] [美]保罗·汉弗莱斯.延长的万物之尺:计算科学、经验主义与科学方法[M].苏湛,译.北京:人民出版社,2017.
[6] Hassabis D,Jumper J.Highly Accurate Protein Structure Prediction with AlphaFold[J].Nature,2021,596:583-589.
[7] Hey T,Tansley S,Tolle K.The Fourth Paradigm:Data-intensive Scientific Discovery[M].Washington:Microsoft Research,2009:xix.
[8] 王飞跃,缪青海.人工智能驱动的科学研究新范式:从AI4S到智能科学[J].中国科学院院刊,2023,38(4):536.
[9] 李正风.科学知识生产方式及其演变[M].北京:清华大学出版社,2006:1.
[10] Kitano H.Artificial Intelligence to Win the Nobel Prize and Beyond:Creating the Engine for Scientific Discovery[J].AI magazine,2016,37(1):39.
[11] Castelvecchi D.Researchers Build an‘AI Scientist’:What Can It Do?[J].Nature,2024,633(50):266.
[12] Nersessian N J.Conceptual Change in Science and in Science Education[J].Synthese,1989(80):166-168.
[13] 吴国盛.从求真的科学到求力的科学[J].中国高校社会科学,2016(1):43.
[14] 黄侃.智能社会的主体焦虑及其概念变革[J].西南民族大学学报(人文社会科学版),2018(7):76.
[15] [美]阿洛克·乔杜里,杰弗里·福克斯,[英]托尼海伊,等.智能辅助科研:深度学习革命[M].张浩,等译.北京:科学出版社,2025.
[16] Marr D.Artificial Intelligence:A Personal View[J].Artificial Intelligence,1977(9):39.
[17] Farrell H,Gopnik A,Shalizi C,et al.Large AI Models Are Cultural and Social Technologies[J].Science,2025,387(6739):1153.
[18] 陈小平.大模型关联度预测的形式化和语义解释研究[J].智能系统学报,2023(4):898.
[19] [美]D.普赖斯.小科学,大科学[M].宋剑耕,戴振飞,译.上海:世界科学社,1982.
[20] [美]迈克尔·吉本斯,卡米耶·利摩日,黑尔佳·诺沃提尼,等.知识生产的新模式:当代社会科学与研究的动力学[M].陈洪捷,现文钦,等译.北京:北京大学出版社,2011.
[21] 李德毅,谢耘.认知机器:智能的非生命实现[J].人民论坛,2025(15):19.
[22] Su H,Chen R,Tang S.et al.Many Heads Are Better Than One:Improved Scientific Idea Generation By A LLM-Based Multi-Agent System[EB/OL].(2024-10-12)[2025-08-30].Preprint arXiv:arXiv.2410.09403.
[23] Silver D,Sutton R.Welcome to The Era of Experience[EB/OL].(2025-04-10)[2025-08-30].http://aibreakdown.org/arxiv-paper-welcom-to-era-of-experience.
[24] Hummel J E,Heaton R F.From Basic Affordances to Symbolic Thought:A Computational Phylogenesis of Biological Intelligence[EB/OL].(2025-08-20)[2026-01-28].http:// dx.doi.org/10.1037/rev0000592,2025:25.
[25] 余鼎,李正风.生成式人工智能社会实验的伦理问题及治理[J].科学学研究,2024(1):6.
[26] 陈小平.大模型:人工智能思想及其社会实验[J].文化纵横,2023(6):76.
① 美国哲学家豪格兰德(John Haugeland)曾在Artificial Intelligence:the Very Idea一书中批评符号人工智能时,将其称为“好的老式人工智能”(Good Old Fashion AI),本文接受这种界说,简称为“老”人工智能。“新”人工智能主要是指以人工神经网络机器学习的大语言模型为代表的人工智能。
② 学术界有时将蛮力法翻译为强力法或暴力法。本文采纳了玛格丽特·博登(Margaret Boden)著《人工智能哲学》中译版刘西瑞和王汉琦的译法,参见该书第182页。
基本信息:
DOI:10.19484/j.cnki.1000-8934.2026.03.008
中图分类号:G301;TP18
引用信息:
[1]黄侃.人工智能驱动另类科学知识生产的现状和图景[J].自然辩证法研究,2026,42(03):89-97.DOI:10.19484/j.cnki.1000-8934.2026.03.008.
基金信息:
国家社会科学基金西部项目“智能科学的认知形态学研究”(23XZX007)
2026-03-18
2026-03-18