On July 4, 2024, the "Forum on Frontier Artificial Intelligence Technologies: Governance Challenges and Responses Measures," organized by Tsinghua University's Institute for AI International Governance (I-AIIG), was successfully held in the Shanghai World Expo Center.
The keynote session of the first segment of the Forum was chaired by Xue Lan, Distinguished Professor of Arts, Humanities and Social Sciences and Dean of I-AIIG at Tsinghua University.
Turing Award winner and academician of the Chinese Academy of Sciences, Andrew Chi-Chih Yao, who is also the Dean of the Institute for Interdisciplinary Information Sciences at Tsinghua University and the Dean of the Shanghai Qi Zhi Institute, delivered a keynote speech titled "Research Directions in AI Safety Governance." He began by discussing the risks associated with advanced large models, including information, physical, and biological intelligence risks, and existential threats. He extended the discussion to the systemic safety of AI and the importance of ensuring human interests, advocating for governance classified by both technical and pathway dimensions.
Yao provided examples from data security research in the era of large models, highlighting that these models tend to remember private data while research on protecting user privacy is still in its early stages. He emphasized the need to develop a core set of technologies to ensure the data security of large models. Additionally, he used examples of Beneficial AGI (Artificial General Intelligence) and Provably Safe AGI to introduce the grand theoretical framework for AI safety governance.
He stressed the urgency of developing AI alignment and corresponding evaluation methods, leveraging China's strengths to establish an AI governance framework suited to the national context. Yao underscored the growing importance of AI safety governance, noting that it has evolved into a new interdisciplinary academic field that urgently requires foundational theoretical research to build safe AI systems.
In his video speech, United Nations Under-Secretary-General and Rector of the United Nations University, Tshilidzi Marwala, highlighted that the tremendous breakthroughs in AI technology are driving a global transformation, bringing with it safety risks and governance challenges. To promote effective AI governance, it is essential to align with the principles of the United Nations Charter and those of human rights. AI governance model must based on sound values, on top of these values are human behaviors, mechanisms for incentives and disincentives, and institutional governance structures, followed by policies, regulations, and standards, and last comes laws.
Marwala reiterated the UN's commitment to AI governance for the benefit of humanity, noting that in 2021, UNESCO released recommendations on AI ethics, which will serve as the foundation for the forthcoming "Global Digital Compact." In April 2024, the United Nations University launched the Global AI Network, comprising over 40 founding members from more than 19 countries and regions. The I-AIIG at Tsinghua University is one of its founding members, and Marwala expressed his hope for active participation from Chinese scholars and practitioners. “AI, as a revolutionary force, places a responsibility on everyone to shape its future to benefit all of humanity,” he said.
Professor Guo Yike, Provost of Hong Kong University of Science and Technology, Academician of the Royal Academy of Engineering, Academician of the European Academy of Sciences, Academician of the Hong Kong Academy of Engineering Sciences and Director of the Data Science Institute at Imperial College London, delivered a keynote speech titled "Data Security Governance for Generative AI: Building an Era of Responsible AI" through his AI-generated digital avatar. In his speech, he emphasized that data governance is fundamental to the development of technology in the AI era and is crucial for the robust growth of the digital economy. He highlighted that access control, permission management, and monitoring and auditing of sensitive data are essential for ensuring data security.
As AI legislation progresses in various countries, it reflects the international community's growing focus on data security governance. “It is particularly important to strengthen international cooperation and exchanges, collectively explore best practices for data security governance, share experiences and lessons, and promote the harmonization of laws and regulations,” concluded Guo.
The panel discussion themed on international AI governance was chaired by Zhang Linghan, a Professor at the Data Law Research Institute of China University of Political Science and Law and a member of the UN High-Level Advisory Group on AI.
Gao Qiqi, Dean of the School of Political Science and Dean of the AI and Big Data Index Research Institute at East China University of Political Science and Law, highlighted three major risks in AI development: unemployment, disorder, and loss of control. “It is necessary to balance global and national governance,” he emphasized.
“AI governance is a cross-border issue, and certain challenges can only be resolved through enhanced international cooperation,” said Duncan Cass-Beggs, Executive Director of the Global AI Risk Initiative at the Centre for International Governance Innovation (CIGI). He called for the United Nations to introduce globally binding agreements and, with the consent of all parties, establish a unified framework along with corresponding regulatory measures to prevent some countries from undermining others.
Liu Jiaqi, Chief Expert in AI Governance and Industrial Development at Huawei Strategic Research Institute, noted that there is a broad consensus across sectors on accelerating AI development, with governance serving to safeguard this development. He also stressed the necessity of building a platform that is globally representative, allows for full participation from multiple stakeholders, and addresses the concerns and voices of all parties involved.
Matt Sheehan, a fellow at the Carnegie Endowment for International Peace, pointed out that the U.S. academic community lacks computing power and platforms, and scholars have limited channels for participating in AI policy-making, whereas Chinese scholars play a more significant role in AI governance. “AI governance issues need to be addressed domestically first and called for ensuring that all societal sectors have the resources to participate in AI governance,” he stated.
Regarding the capacity gap between the Global North and South, Alexander Kriebitz, a lecturer at the Technical University of Munich noted that it requires not only focusing on AI development, improving infrastructure, and promoting AI applications but also interacting with the public through education to enhance their trust in AI.
“AI in China is advancing towards empowering various industries and it's crucial to leverage the role of the industry in AI governance,” said Gu Dengchen, Deputy Director of the Policy and Law Research Office at Alibaba Group. He pointed out that many current AI governance frameworks are still at the domestic discussion stage and primarily remain as conceptual ideas. There is a need to focus on solving common human challenges, particularly the fundamental issue of aligning large models.
