On April 22nd 2024, the Simon Institute for Longterm Governance (SI) hosted a set of briefings for UN diplomats in New York entitled, “Where is AI at? Implications of frontier research and development for multilateral governance.” The briefings were hosted in collaboration with the Permanent Mission of Costa Rica, and drew 40 mission representatives and tech experts from 30+ countries. Diplomats left with a better understanding of the current state of frontier AI, insight into the evolving AI governance landscape, and a clearer grasp of the proposals relevant to AI in the Global Digital Compact (GDC).
The briefings began with a presentation from AI experts Kristian Rönn and Oliver Edholm, the respective CEOs and co-founders of Normative and Depict.ai. They explained scaling laws and the relationship between computing power and model capability, discussed the striking progress frontier AI has made in recent years, and provided a live demonstration of what current-day AI systems are already capable of. They ended their presentation by outlining potential commitments an international AI decision-making body could make. These included:
- Committing to a system of licensing and risk mitigation measures above agreed upon compute thresholds;
- Committing to an international auditing system to prevent unlicensed actors from accessing compute beyond thresholds.
Following this, Belinda Cleeland, Director of Policy at SI, provided an overview of the existing AI governance landscape. She discussed AI’s potential impact on the SDGs, existing national, regional, and international frameworks for AI governance, gaps in the current AI governance ecosystem, and the potential role of the UN in addressing these gaps. She also provided a break-down of relevant AI-related proposals in the recent Zero Draft of the GDC, and outlined open questions and recommendations for future iterations. You can find key insights from this portion of the workshop in our briefing note here.
The briefings each ended with an audience Q&A. Participants asked about the various AI-related institutional proposals included in the GDC Zero Draft, seeking clarification on their intended form and function. Other questions emerged on how to ensure the safety of open-source AI models, how best to ensure accountability of the private sector, and the relationship between AI governance and data governance.
For more information on these briefings, or to inquire about further support on the subject, please don’t hesitate to reach out to Belinda Cleeland at belinda@simoninstitute.ch.