摘要
本文基于链式孵化视角,利用2015~2024年中国31省(区、市)面板数据,采用双重机器学习模型检验大模型技术对人工智能(AI)产业创新的影响机制。研究发现:大模型技术对AI产业创新有显著正向总效应,其中3112%需通过“大模型技术→产业生态体系→AI产业创新”链式路径实现,阻断任一环节则间接效应降至0,印证链式孵化的必要性。产业生态为核心中介,呈技术强驱动、生态弱转化特征,生态向创新的转化效率为关键瓶颈。进一步分析表明,大模型技术与产业生态存在非线性协同效应,技术对创新的边际效应随生态成熟度递增,且低生态地区技术已具备独立驱动创新的核心引擎作用。区域异质性显著,东部链式路径传导完整,间接效应占比2961%,中部因生态中介缺失致传导失效;西部因技术-生态适配偏差出现负向效应;东北协同机制尚未建立。由此提出突破生态→创新转化瓶颈、构建算力-孵化-开源联动基建、实施梯度区域政策、防范生态垄断风险等建议,以完善链式孵化机制,推动AI产业创新。
From the chained incubation perspective,this study utilizes panel data from China's 31 provinces and municipalities from 2015 to 2024 and employs Double Machine Learning(DML)to examine how large model technology impacts AI industry innovation.Results show significant positive total effects,with 3112%transmitted through“technology→ecosystem→innovation”pathways.Blocking any link reduces the indirect effect to zero,confirming chained incubation necessity.The ecosystem acts as the core intermediary,exhibiting strong technological drive and weak ecological transformation characteristics,with the conversion efficiency from ecosystem to innovation being the key bottleneck.Further analysis indicates that large model technology and the industrial ecosystem exhibit non-linear synergistic effects,with the marginal effect of technology on innovation increasing as the ecosystem matures.In regions with low ecosystem maturity,technology has already become the core engine driving innovation independently.Regional heterogeneity is significant:in the eastern region,the chain-based transmission pathway is intact,with indirect effects accounting for 2961%;in the central region,transmission fails due to the absence of an ecosystem intermediary;the western region exhibits negative effects due to technical-ecological mismatches;and the northeastern region lacks established collaborative mechanisms.Based on these findings,the study proposes recommendations to address the bottleneck in ecological-to-innovation conversion,including establishing a linked infrastructure of computing power,incubation,and open-source collaboration,implementing tiered regional policies,and mitigating risks of ecological monopolization,to enhance chain incubation mechanisms and drive the innovative development of the AI industry.
作者
郑庆花
肖翠仙
Zheng Qinghua;Xiao Cuixian(School of Economics,Wuzhou University,Wuzhou 543002,China)
出处
《工业技术经济》
2025年第10期32-42,共11页
Journal of Industrial Technology and Economy
基金
广西哲学社会科学规划研究课题“数字赋能广西产业链创新链深度融合的动力机制与路径研究”(项目编号:22FJY024)
梧州学院博士基金项目“数字经济促进现代化产业体系建设的效应测度研究”(项目编号:2023A001)。
关键词
大模型
链式孵化
人工智能产业创新
双重机器学习
产业生态
转化效率
系统性创新
非线性协同效应
large-scale model
chained incubation
artificial intelligence industrial innovation
double machine learning
industrial ecosystem
conversion efficiency
systematic innovation
nonlinear synergy effect
作者简介
郑庆花,梧州学院经济学院讲师。研究方向:数据治理;通讯作者:肖翠仙,梧州学院经济学院教授,博士。研究方向:产业经济。
