Abstract:

With the formal recognition of data as the fifth factor of production and the transformative breakthroughs ingenerative artificial intelligence（ AI）， the integration of“ Data + AI” has entered a phase of profound convergence，becoming a critical research focus across industries. Addressing the insufficient exploration of their synergisticmechanisms， the bidirectional empowerment dynamics and implementation pathways were investigated， aiming toprovide theoretical frameworks and empirical validation for accelerating AI iteration， unlocking data factor value，and advancing intelligent service deployment. The bidirectional collaborative mechanism between data for AI and AIfor data were systematically examined. The relationship between them had evolved from a unidirectional dependencyto a state of deep bidirectional synergy， forming a virtuous interactive cycle and entering an application-drivenintegration phase. In the practical context， taking the National Agricultural Library as a case study， implementationstrategies were proposed for the layout and construction of“ Data + AI” infrastructure. It outlined a practical pathwaycentered on a high-quality knowledge base that serves as the foundational pillar-full-chain intelligent enginesenabling end-to-end data processing-multi-scenario service capabilities tailored to diverse application needs. Thisframework ultimately generated an intelligent flywheel effect through the continuous interplay of “data-models-applications”. Based on the data-centric AI development paradigm， it emphasized the complementary construction ofhigh-quality big data and small data， co-evolved the general capabilities of large models and the specializedcapabilities of small models， deepened the integrated two-way empowerment mechanism， and thereby facilitated theefficient implementation of intelligent application scenarios across various domains.

Key words: generative AI, data elements, empowerment mechanism, data-centric, data-intelligence integration, agricultural knowledge big model

摘要：

伴随数据增列为第五生产要素与生成式人工智能 （artificial intelligence，AI）技术的跨越式突破，数据和AI进入融合共生的剧烈变革期，“数据+AI”深度交叉融合成为各行业的研究焦点。针对二者协同机制的解析不足，探讨其双向赋能的机制与实践路径，旨在为AI快速迭代升级、数据要素价值加速释放及行业智能服务发展落地提供理论支撑与实证支持。从数据赋能AI和AI赋能数据维度阐释其双向协同机制，二者关系从单向依赖转为双向深度协同发展，形成良性互动循环，进入由应用驱动的融合发展阶段。在实践层面，以国家农业图书馆为例，提出“数据+AI”基础设施布局与建设响应策略，规划出高质量知识底座-全链条智能引擎-多元场景服务为核心的实践路径，最终形成“数据-模型-应用”的智慧飞轮效应。基于以数据为中心的AI开发范式，强调高质量大数据和小数据互补性建设，协同进化大模型通用能力和小模型专用能力，深化双向赋能一体化机制，助力智能化应用场景高效落地。

关键词: 生成式AI, 数据要素, 赋能机制, 以数据为中心, 数智融合, 农知大模型