美核武专家紧急呼吁：绝不能这么做！

Core Viewpoint - The article emphasizes the importance of not allowing artificial intelligence to control nuclear weapons early warning systems, despite the consensus among nuclear powers that humans should retain ultimate decision-making authority regarding nuclear weapon use [1][2]. Group 1: Importance of Human Oversight - Erin D. Dumbacher highlights a historical incident during the Cold War where a false alarm in the Soviet nuclear warning system was correctly identified by an officer, preventing a potential nuclear disaster [1]. - The article stresses that the current advancements in artificial intelligence pose risks to nuclear safety, particularly in the context of early warning systems [1][4]. Group 2: Risks of AI in Nuclear Context - The article discusses how AI technology facilitates the creation of deepfakes, which can mislead decision-makers, including high-ranking officials like the U.S. President [4]. - There is a concern that AI could produce false information or "algorithmic hallucinations," which could interfere with human judgment in critical situations [4]. Group 3: Recommendations for AI Regulation - Dumbacher suggests that if the U.S. government pursues military applications of AI, strict limitations should be imposed regarding nuclear weapons, including enhanced information verification processes [5]. - The article advocates for training individuals to remain vigilant against misleading AI-generated information and calls for regulatory measures on presidential authority concerning nuclear weapon use [5].