AI Governance and Capabilities See Major Advances
As Dartmouth College marked 70 years since the term 'AI' was coined, the United Nations created an international scientific panel to provide evidence-based recommendations on the technology's societal and ethical implications. The move comes as new benchmarks from a Stanford index show a dramatic leap in AI's reasoning and problem-solving capabilities, increasing the urgency for global governance frameworks.
The original 1956 Dartmouth Summer Research Project on Artificial Intelligence was a two-month, 10-man study that operated on the principle that "every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." This foundational event, which coined the term 'AI', brought together pioneers like John McCarthy and Claude Shannon to lay the intellectual groundwork for the field. Dartmouth's commemorative 70th-anniversary event, "The Dartmouth Conference, Revisited," scheduled for October 29-30, 2026, aims to connect that founding vision to a new mandate for responsible innovation. The focus is on clarifying where AI can assist human judgment and creativity and where ethical guardrails are essential. The new 40-member UN panel is formally named the Independent International Scientific Panel on Artificial Intelligence and will provide annual, evidence-based reports on AI's opportunities and risks. The panel, which includes experts from academia, the private sector, and civil society, is designed to function similarly to the Intergovernmental Panel on Climate Change (IPCC), offering a trusted scientific reference for policymakers. The UN General Assembly approved the panel's creation with 117 countries in favor, though the United States and Paraguay voted against it. The panel's members, selected from over 2,600 candidates, include prominent figures such as AI pioneer Yoshua Bengio, Joelle Barral of Google's DeepMind, and Nobel Peace Prize laureate Maria Ressa. The urgency for such governance is underscored by rapid capability leaps. On the GPQA benchmark, a set of PhD-level science questions, AI performance jumped by 48.9 percentage points in just one year. On SWE-bench, which involves solving real-world coding issues from GitHub, AI success rates soared from just 4.4% in 2023 to 71.7% in 2024. In the realm of complex mathematical reasoning, an AI system developed by Google DeepMind solved four out of six problems at the International Mathematical Olympiad (IMO), achieving the equivalent of a silver medal for the first time. Another system, TongGeometry, later surpassed this by solving all 30 problems from a benchmark of past IMO geometry questions.
