Abstract:

【Objective】 analyze cognition and application status of Generative Generative Artificial Intelligence (Generative AI), tools. 【Methods】 Based on a questionnaire survey of 208 researchers from the Xinjiang Academy of Agricultural Sciences, this study analyzed the cognition and application status of generative AI tools in agricultural research. 【Results】 Agricultural researchers' cognition of generative AI remains at a basic level, with information primarily obtained through non-systematic channels.In terms of application, generative AI is concentrated in information retrieval and literature organization in the early stages of research, but its penetration in value-creation stages of mid-to-late research is insufficient.Notably, it is rarely applied in fields directly related to Xinjiang's agricultural production practices, such as field management plan generation.Key obstacles to its adoption include AI hallucinations, inadequate professional knowledge bases, poor localization adaptation, and privacy/data security risks inherent in AI tools.Additionally, researchers face challenges in technical training and organizational support. 【Conclusion】 1.Accelerate localized deployment and development by creating region-specific agricultural large language models and establishing localized knowledge bases.2.Standardize AI applications in agricultural research by formulating application guidelines and establishing a data classification protection system.3.Strengthen research integrity and data security by establishing research integrity committees and increasing investments in data security technologies.

These measures can promote the deep integration of generative AI into agricultural research in Xinjiang, facilitating the region's agricultural modernization.

Key words: artificial intelligence; agricultural research; process management; data security

摘要：

【目的】 生成式人工智能作为引领新一轮科技革命的核心技术,正深度重塑农业科研与应用范式,分析农业科研人员对生成式人工智能(Generative)AI工具的认知与应用现状。【方法】 基于对新疆维吾尔自治区农业科学院208名科研人员的问卷调查,分析生成式人工智能工具在农业科研中的认知与应用现状。【结果】 农业科研人员对生成式AI的认知多停留在基础层面,主要通过非系统化渠道获取信息。生成式AI集中于科研前期的信息检索和文献整理,而在科研中后期的价值创造环节渗透不足,尤其在田间管理方案生成等与新疆农业生产实践直接相关的领域应用较少。AI工具自身存在的AI幻觉、专业知识库不足、本地化适配差以及隐私与数据风险等问题,成为科研人员使用的主要障碍。此外,科研人员在技术培训和组织支持方面也面临困境。【结论】 一是加速本地化部署与开发,开发具有本地化特色的农业大模型,建立本地化知识库;二是规范AI在农业科研中的应用,制定应用规范和指南,建立数据分级保护制度;三是强化科研诚信与数据安全防护,设立科研诚信和管理委员会,加大数据安全防护技术投入。可推动生成式AI在新疆农业科研中的深度应用,助力新疆农业现代化发展。

关键词: 人工智能, 农业科研, 过程管理, 数据安全