UN Forms AI Panel as Models Surpass Human Benchmarks
The United Nations has announced the formation of an international scientific panel to analyze AI's societal impacts and offer policy recommendations. The move coincides with a 2025 Stanford Index report revealing that top AI models now match or exceed human performance on complex benchmarks. Researchers also recently gathered at Dartmouth to mark the 70th anniversary of the conference that coined the term "artificial intelligence."
The UN's new AI advisory body is comprised of 39 experts from 33 countries, including executives from OpenAI, Microsoft, and Sony. The panel's mandate is to analyze AI's risks and opportunities and advance recommendations for its international governance. An interim report was released in December 2024 after consultations with over 2,000 participants. This initiative comes as enterprise adoption of AI has surged, with 78% of organizations reporting its use in 2024, a significant increase from 55% the previous year. This boom is fueled by massive investment, with total corporate spending on AI reaching $252.3 billion in 2024. The United States continues to lead in private AI investment with $109.1 billion, nearly 12 times that of China. The performance of AI models has seen dramatic improvements on new, more challenging benchmarks. For instance, on the SWE-bench for coding challenges, AI performance jumped from solving 4.4% of problems in 2023 to 71.7% in 2024. In some timed tasks, top AI systems now outperform human experts. The 70th-anniversary event at Dartmouth featured a range of speakers from industry and academia, including executives from Google, OpenAI, and Anthropic. Discussions focused on the real-world implications of AI, from its role in national security to its potential to address staffing shortages in healthcare. A key trend in 2025 is the move from conversational AI to AI agents that can execute complex tasks. Models are increasingly multimodal by default, capable of understanding images and voice, not just text. Additionally, the rise of powerful, low-cost open-source models from China is shifting the global AI landscape and challenging U.S. dominance. The UN General Assembly recently approved a separate 40-member global scientific panel to provide independent, science-based insights into AI's impacts. This panel, which received overwhelming support with a 117-2 vote, will serve three-year terms and is expected to release its first report by July. The creation of this body was met with objections from the United States, which cited concerns about a lack of transparency in the selection process.
