Abstract:

In recent years， artificial intelligence （AI）， driven by the three engines of algorithm， data and computing power， has developed rapidly， and achieved many remarkable achievements in frontier fields， such as AlphaFold3， nuclear fusion intelligent control， and novel coronavirus drug design. AI-driven scientific research （AI for Science， AI4S） has solved the bottleneck of scientific data analysis with high dimensions， large scale and span， and limited scientific experiments that restrict large-scale interdisciplinary scientific research activities， and promotes scientific research towards a new model with “platform collaboration” as the main feature. This paper reviewed the international situation of AI4S， the current development status of China’s smart agriculture and the practical dilemma， and combined literature， statistical data， and investigation case analysis to put forward the practical path to promote AI4S application in China’s agricultural development. AI4S would become a powerful engine to transform agricultural production from the traditional model of “looking at the sky， looking at the land and looking at the crops” to models such as intelligent perception， intelligent decision-making and visual management， and promote scientific research from the single-fighting “small farmer’s workshop” model to the platform scientific research of “Android model”. Researchers could share basic computing power， models， algorithms， databases and knowledge bases. On this platform， the research and development and application of the whole life cycle of the agricultural industry chain would be focused on， and scientific research innovation and application of results be accelerated through the “snowball effect”. Using artificial intelligence technology to realize digitalization， networking and intelligence of agricultural production， and support online iteration of theory-experiment. It is also necessary to improve the high-quality agricultural science digital resource system， appropriately advance AI key technologies and infrastructure， optimize the cross-innovation research ecology under the new paradigm， formulate agricultural data management norms， strengthen policy creation and supporting policy mechanisms to break through data barriers and to promote artificial intelligence + agriculture， and to strengthen agricultural science and technology innovation from the source.

Key words: AI for Science, AI4S, artificial intelligence, smart agriculture, scientific research paradigm

摘要：

近年来，在算法、数据、算力三大引擎驱动下，人工智能（artificial intelligence，AI）发展迅速，并在AlphaFold3、核聚变智能控制、新冠药物设计等前沿领域取得诸多令人瞩目的成果。AI驱动的科学研究（AI for Science，AI4S）解决了科学数据分析维度高、尺度跨度大以及局限性科研实验制约大规模跨学科科研活动的瓶颈问题，促进科学研究迈向以“平台协作”为主要特征的新模式。分析了AI4S的国际态势，梳理了当前我国农业数字化发展现状及现实困境，将文献、统计数据、调研案例分析相结合，提出推动AI4S赋能我国农业发展的实践路径。AI4S将成为撬动农业生产从“看天、看地、看庄稼”的传统模式向智能感知、智能决策、可视化管理等模式转变的强力引擎，推动科学研究从单打独斗的“小农作坊模式”迈向“安卓模式”的平台科研。在此平台上，科研人员共享算力、模型、算法、数据库和知识库等基础设施，围绕农业全产业链全生命周期研发应用，通过“滚雪球效应”加速科研创新和成果应用。利用AI技术赋能农业生产数字化、网络化和智能化，为支撑理论-实验的在线迭代，还需要完善高质量农业科学数据资源体系、适度超前推进AI关键技术与基础设施、优化新范式下的交叉创新科研生态、加强农业数据安全监管、制定完善的配套政策和激励机制等措施来打通数据壁垒，推动AI+农业落地，从源头强化农业科技创新，推动农业强国建设。

关键词: AI for Science, 人工智能, 智慧农业, 科研范式