美国能源部:2025年《人工智能战略报告》，重定义国家核心能力

Core Insights - The U.S. Department of Energy (DOE) has released its 2025 Artificial Intelligence Strategy report, emphasizing AI as a core driver for national security, scientific discovery, and energy leadership [1] - The strategy aims to transform AI from specialized applications in national laboratories to a systematic, scalable enterprise capability across the DOE [1] National Security Applications - The strategy prioritizes national security, focusing on "high-consequence systems," particularly nuclear deterrence [2] - AI implementation is aimed at maintaining the U.S. deterrent edge, with the National Nuclear Security Administration (NNSA) leading efforts in nuclear stockpile management and predictive maintenance [2] Nuclear Non-Proliferation and Safety - The DOE is developing multimodal foundational models for nuclear non-proliferation, assessing risks from external proprietary models [3] - Generative AI is being introduced for engineering design in high-consequence systems, enhancing reliability and modernization of critical national security assets [3] - Tools based on large language models (LLMs) are being developed for intelligence and counterintelligence analysis to improve the timeliness and accuracy of intelligence products [3] AI Security Measures - The DOE recognizes the need for trust in AI applications within critical infrastructure, establishing an AI assurance testing platform to evaluate vulnerabilities and robustness [4] Computing and Data Infrastructure - The DOE's unmatched computing resources and vast scientific data are seen as foundational for its AI ambitions, with the report highlighting challenges such as data silos and legacy infrastructure [5][7] - The establishment of the "American Science Cloud" aims to facilitate data sharing and collaboration across government, academia, and the private sector [7] Energy Leadership and Governance - The strategy outlines how AI will permeate energy production, distribution, and regulation, supported by robust internal governance and workforce planning [8][9] - A proportional AI governance framework is proposed to ensure that oversight matches the risk level of AI applications [9] Conclusion - The DOE's 2025 AI Strategy represents a dual approach: a defensive strategy to protect high-consequence systems and an offensive strategy to leverage computing resources for scientific discovery and energy leadership [10] - The success of this strategy hinges on overcoming significant barriers such as data silos and bureaucratic inertia, emphasizing the need for organizational transformation [10]