Group 1 - The "Phoenix Bay Area Finance Forum 2025" was held in Guangzhou, focusing on the theme "New Pattern, New Path" and gathering global elites from politics, business, and academia to explore development opportunities amidst changing circumstances [1] Group 2 - Li Ke, co-founder and CEO of Haitan Ruisheng, emphasized the importance of cross-language model training in AI globalization, identifying the lack of Chinese corpus as a significant challenge [3] - Li Ke proposed two solutions to address the shortage of Chinese data: leveraging large language model technology to connect different languages and sourcing high-quality Chinese data [3] - The CEO expressed a desire to collaborate with Phoenix TV to enhance training data quality by utilizing their high-quality data assets [3] Group 3 - Li Ke noted that the production of data in human society is growing at an unprecedented rate, indicating a continuous explosion of data volume [4] - He highlighted the critical importance of data quality, pointing out the severe homogenization of data on the internet and the need for more high-quality data sources [4] - The CEO reiterated the value of Phoenix TV's data, which is characterized by high production quality and the encapsulation of knowledge and insights from the creators [4] - Li Ke articulated a vision for collaboration with Phoenix TV to build a foundational data structure for an intelligent world [4]
李科:跨语言模型训练中文语料不足难题可解,高质量数据源是关键