The Government Knows AGI is Coming

《埃兹拉·克莱因秀》节目采访了拜登白宫前人工智能特别顾问本·布坎南，讨论了人工智能的快速发展及其对国家安全、经济和社会的潜在影响。 克莱因开篇指出，越来越多的AI专家和政府官员达成共识，认为通用人工智能（AGI），即能够执行人类能够执行的任何认知任务的人工智能，很可能在未来两到三年内出现。 布坎南证实了这一评估，他表示，具有非凡能力的AI系统即将到来，有可能在特朗普的第二个任期内出现。他强调了率先实现AGI的国家将获得的经济、军事和情报优势，并指出国家安全问题是美国努力在AI开发方面保持领先地位的主要驱动因素。 他指出，人工智能的独特之处在于，它是一种主要由私营部门开发出来的革命性技术，缺乏像核武器或互联网等过去的技术进步中看到的政府监督和塑造影响力。 谈话探讨了如果中国率先实现AGI可能带来的潜在危险。 布坎南强调了它对网络战、情报分析以及数字系统日益脆弱的影响。 他承认与人工智能驱动的监控国家相关的风险以及个人权利可能受到侵蚀，尤其是在中国等专制国家。 然而，他也看到了民主国家在加强刑事司法和减少偏见方面的潜在好处，尽管这些好处并非必然。 讨论还触及了人工智能实验室自身容易受到黑客攻击和知识产权盗窃的问题，因为它们是私营实体，没有政府设施的强大安全协议。 布坎南解释了他如何试图向这些实验室发出信号，以帮助美国政府帮助他们完成任务。 一个关键的争议点围绕着对中国先进AI芯片的出口管制。 虽然承认这些管制被认为是具有对抗性的，并且可能激励中国加大对国内AI开发的投资，但布坎南认为这些管制对于维持美国的领先地位是必要的。 他还谈到了围绕中国AI公司DeepSeek最近突破的争论，认为虽然其成就令人印象深刻，但并未从根本上改变美国公司在AI领域具有优势的现有分析。 对话随后转向了关于人工智能监管的争论。 克莱因指出，人工智能界的一些人士，如风险投资家马克·安德森，担心过度监管可能会扼杀创新，并有利于大公司，而人工智能安全界的其他人则担心，如果缺乏监管，可能会出现安全方面的逐底竞争。 布坎南为拜登政府的做法辩护，强调其努力促进竞争，培育充满活力的AI生态系统，同时解决对AI安全的担忧。 然而，特朗普政府已经废除了拜登政府发布的行政命令。 布坎南强调了人工智能政策需要两党合作，强调了保护在人工智能创新方面处于领先地位的美国公司的重要性。 他还讨论了人工智能日益增长的能源需求，以及国内能源基础设施发展对于支持该行业增长的重要性。 克莱因向布坎南施压，询问人工智能对劳动力市场的潜在影响。 虽然布坎南承认可能存在重大颠覆和转型，但他强调必须确保工人在谈判桌上占有一席之地，并在转型过程中受到保护。 他还讨论了人工智能赋予个人权力和提高其能动性的潜力，从而促进更具活力的经济。 最终，讨论揭示了一个充满巨大潜力和重大风险的复杂局面。 虽然承认存在不确定性以及缺乏明确的政策答案，但布坎南坚持认为，美国政府必须优先保持其在人工智能领域的领导地位，促进创新，并解决可能出现的潜在社会和经济混乱。 他认为，安全和机会之间的平衡至关重要，美国必须建立能够管理这项快速发展的技术，同时维护美国价值观和经济利益的机构。

The Ezra Klein Show features a discussion with Ben Buchanan, former special advisor for artificial intelligence in the Biden White House, about the rapid advancement of AI and its potential implications for national security, the economy, and society. Klein opens by noting a growing consensus among AI experts and government officials that artificial general intelligence (AGI), AI capable of performing any cognitive task a human can, is likely to arrive within the next two to three years. Buchanan confirms this assessment, stating that extraordinarily capable AI systems are imminent, potentially during a second Trump term. He emphasizes the economic, military, and intelligence advantages that would accrue to the nation that achieves AGI first, citing national security concerns as the primary driver for US efforts to lead in AI development. He notes the unique position of AI as a revolutionary technology primarily developed by the private sector, lacking the governmental oversight and shaping influence seen with past technological advancements like nuclear weapons or the internet. The conversation explores the potential dangers if China achieves AGI first. Buchanan highlights the implications for cyber warfare, intelligence analysis, and the increased vulnerability of digital systems. He acknowledges the risks associated with AI-powered surveillance states and the potential for the erosion of individual rights, especially in autocratic nations like China. However, he also sees potential benefits for democracies in terms of enhanced criminal justice and reduced bias, although these benefits are not guaranteed. The discussion touches on the vulnerability of AI labs themselves to hacking and intellectual property theft, given that they are privately run entities without the robust security protocols of government facilities. Buchanan explains how he tried to signal to these labs, to help the US government to help them in their mission. A key point of contention revolves around export controls on advanced AI chips to China. While acknowledging that these controls are perceived as antagonistic and might incentivize China to invest more heavily in domestic AI development, Buchanan defends them as necessary to maintain a US lead. He also addresses the debate surrounding the recent breakthroughs by Chinese AI firm DeepSeek, arguing that while its achievements are impressive, they do not fundamentally alter the existing analysis of US companies having an advantage in AI. The conversation then shifts to the debate surrounding the regulation of AI. Klein points to concerns from some in the AI community, like venture capitalist Marc Andreessen, that overregulation could stifle innovation and favor large companies, while others in the AI safety community fear a race to the bottom in terms of safety if regulations are lacking. Buchanan defends the Biden administration's approach, emphasizing its efforts to promote competition and foster a dynamic AI ecosystem while addressing concerns about AI safety. However, the Trump administration has repealed the executive order made by the Biden administration. Buchanan highlights the need for a bipartisan approach to AI policy, emphasizing the importance of protecting American companies that are leading in AI innovation. He also discusses the increasing energy demands of AI and the importance of domestic energy infrastructure development to support the industry's growth. Klein presses Buchanan on the potential impact of AI on labor markets. While Buchanan acknowledges the potential for significant disruptions and transitions, he emphasizes the importance of ensuring that workers have a seat at the table and are protected during this transformation. He also discusses the potential for AI to empower individuals and increase their agency, leading to a more dynamic economy. Ultimately, the discussion reveals a complex landscape with immense potential and significant risks. While acknowledging the uncertainties and the lack of clear policy answers, Buchanan maintains that the US government must prioritize maintaining its leadership in AI, fostering innovation, and addressing the potential societal and economic disruptions that may arise. He argues that a balance between safety and opportunity is crucial and that the US must build institutions capable of managing this rapidly evolving technology while safeguarding American values and economic interests.