对话陈松蹊院士：中国急需加速构建高质量的科学数据集

Core Viewpoint - The need for high-quality data set construction in China is emphasized, with a call for scientists to adopt a public perspective and scientific vision to drive this initiative [1][4]. Group 1: High-Quality Data Sets - China possesses the capability and research strength to establish high-quality data sets, supported by both domestic and international observational data [1]. - Breakthroughs have been achieved in constructing high-quality marine data sets, with test results meeting or exceeding international standards [1][4]. - The construction of high-quality data sets requires leadership from relevant departments to organize scientists and ensure effective implementation [4]. Group 2: Statistical Analysis and Big Data - Traditional statistical methods face challenges in handling "ultra-high-dimensional" data, particularly in fields like genetics and geophysics, where data dimensions can reach millions while sample sizes remain small [1]. - There is a need for innovative hypothesis testing methods to address the complexities of high-dimensional data analysis [1]. - Statistical methods can serve as a common language across various fields, facilitating cross-domain applications of big data [2]. Group 3: Artificial Intelligence and Statistics - The integration of artificial intelligence (AI) and statistics is crucial, as AI models are fundamentally data-driven and share commonalities with statistical models [2]. - Simple statistical models should be prioritized before resorting to complex AI models, especially in scenarios with limited data [2]. - The importance of uncertainty measurement in AI and statistical methods is highlighted, as high uncertainty can render estimates meaningless [2]. Group 4: Talent Development - There is a significant talent gap in data analysis, including AI, necessitating enhanced training and educational programs [3]. - Tsinghua University is actively developing relevant undergraduate and master's programs to address this talent shortage [3].