Abstract:

[Significance] Artificial intelligence for science (AI4S), as an emerging paradigm that deeply integrates artificial intelligence(AI) with scientific research, has triggered profound transformations in research methodologies. By accelerating scientific discovery through AI technologies, it is driving a shift in scientific research from traditional approaches reliant on experience and intuition to methodologies co-driven by data and AI. This transition has spurred innovative breakthroughs across numerous scientific domains and presents new opportunities for the transformation of agricultural research. With its powerful capabilities in data processing, intelligent analysis, and pattern recognition, AI can transcend the cognitive limitations of researchers in the field and is gradually emerging as an indispensable tool in modern agricultural scientific research, injecting new impetus into the intelligent, efficient, and collaborative development of agricultural scientific research. [Progress] This paper systematically reviews the current advancements in AI4S and its implications for agricultural research. It reveals that AI4S has triggered a global race among countries around the world vying for the commanding heights of a new round of scientific and technological strategies. Developed nations in Europe and America, for instance, have laid out the frontier areas in AI4S and rolled out relevant policies. Meanwhile, some top universities and research institutions are accelerating related research, and tech giants are actively cultivating related industries to advance the application and deployment of AI technologies in scientific research. In recent years, AI4S has achieved remarkable development, showing great potential across multiple disciplines and finding widespread application in data mining, model construction, and result prediction. In the field of agricultural scientific research, AI4S has played an important role in accelerating multi-disciplinary integration, promoting the improvement of the scientific research efficiency, facilitating the breakthrough of complex problems, driving the transformation of the scientific research paradigm, and upgrading scientific research infrastructure. The continuous progress of information technology and synthetic biology has made the interdisciplinary integration of agriculture and multiple disciplines increasingly closer. The deep integration of AI and agricultural scientific research not only improves the application level of AI in the agricultural field but also drives the transformation of traditional agricultural scientific research models towards intelligence, data-driven, and collaborative directions, providing new possibilities for agricultural scientific and technological innovation. The new agricultural digital infrastructure is characterized by intelligent data collection, edge computing power deployment, high-throughput network transmission, and distributed storage architecture, aiming to break through the bottlenecks of traditional agricultural scientific research facilities in terms of real-time performance, collaboration, and scalability. Taking emerging disciplines such as Agrinformatics and climate-focused Agriculture-Forestry-AI (AgFoAI) as examples, they focus on using AI technology to analyze agricultural data, construct crop growth models, and climate change models, etc., to promote the development and innovation of agricultural scientific research. [Conclusions and Prospects] With its robust capabilities in data processing, intelligent analysis, and pattern recognition, AI is increasingly becoming an indispensable tool in modern agricultural scientific research. To address emerging demands, core domains, and research processes in agricultural research, the concept of agricultural intelligent research is proposed, characterized by human-machine collaboration and interdisciplinary integration. This paradigm employs advanced data analytics, pattern recognition, and predictive modeling to perform in-depth mining and precise interpretation of multidimensional, full-lifecycle, large-scale agricultural datasets. By comprehensively unraveling the intrinsic complexities and latent patterns of research subjects, it autonomously generates novel, scientifically grounded, and high-value research insights, thereby driving agricultural research toward greater intelligence, precision, and efficiency. The framework's core components encompass big science infrastructure (supporting large-scale collaborative research), big data resources (integrating heterogeneous agricultural datasets), advanced AI model algorithms (enabling complex simulations and predictions), and collaborative platforms (facilitating cross-disciplinary and cross-institutional synergy). Finally, in response to challenges related to data resources, model capabilities, research ecosystems, and talent development, actionable pathways and concrete recommendations are outlined from the perspectives of top-level strategic planning, critical technical ecosystems, collaborative innovation ecosystems, disciplinary system construction, and interdisciplinary talent cultivation, aiming to establish a new AI4S-oriented agricultural research framework.

Key words: AI4S, AI for agricultural science, system framework, implementation pathways, research paradigm