联合实验室专家陈雷：希望大湾区成为数据安全使用典范

Core Insights - Generative artificial intelligence is a core driving force behind the new technological and industrial revolution, serving as an important engine for cultivating new productive forces and injecting new momentum into high-quality economic development [1] - The establishment of the Guangdong-Hong Kong-Macao Greater Bay Area Generative Artificial Intelligence Safety Development Joint Laboratory aims to create an innovative ecosystem that integrates government, industry, academia, research, and application, focusing on enterprise development, industry implementation, and safety regulation [1][2] Group 1: Data Quality and Integration - Data quality is crucial in the training of large models, and the Joint Laboratory is expected to integrate research capabilities from various universities to guide the correct use of data from policy and institutional levels [2][3] - The laboratory aims to create a big data platform by consolidating data from various universities in the Greater Bay Area, facilitating model testing and performance evaluation [3][9] Group 2: Data Security and Governance - Data security poses significant challenges, requiring a balance between data integration and safety, with suggestions for using techniques like homomorphic encryption and privacy computing for AI training [5][8] - The laboratory is expected to establish a collaborative governance system for data security and AI safety, promoting data sharing while ensuring privacy protection [5][7] Group 3: AI Talent Development - The Joint Laboratory is seen as a platform for AI talent cultivation, advocating for a "1+1" model where students spend one year in academic learning and another year in industry practice to bridge the gap between education and real-world application [6][7] Group 4: Future Expectations - There is an expectation for the Joint Laboratory to form more alliances and agreements to promote data security usage, creating a virtuous cycle of data sharing and safety in the Greater Bay Area [7] - The laboratory is also anticipated to focus on the safety of large models, ensuring their outputs are reliable and controllable while addressing new risks such as data poisoning [8][9]