黄仁勋开通社交账号,首条“我的一封公开信”撕开AI路线之争(中英全文)

【相关阅读】[英伟达CEO黄仁勋:“AI末日论”该翻篇了,行业别再吓人了](https://finance.sina.com.cn/tech/roll/2026-07-25/doc-iniizkxf1068619.shtml) [黄仁勋开通社交账号,首条推文发了啥?](https://finance.sina.com.cn/wm/2026-07-25/doc...

【相关阅读】[英伟达CEO黄仁勋:“AI末日论”该翻篇了,行业别再吓人了](https://finance.sina.com.cn/tech/roll/2026-07-25/doc-iniizkxf1068619.shtml) [黄仁勋开通社交账号,首条推文发了啥?](https://finance.sina.com.cn/wm/2026-07-25/doc-iniizerm1366387.shtml) [一夜涨粉70万!黄仁勋首条推文引爆全网,网友:这是要当网红了?](https://finance.sina.com.cn/stock/bxjj/2026-07-25/doc-iniizerm1355030.shtml) 编辑:米奇 来源:财经会议圈 7月24日晚间,英伟达首席执行官黄仁勋在社交平台 X 发布个人入驻后的第一条推文,并未推介公司芯片与产品,而是直接附上题为《开放权重与美国在AI领域的领导地位》的联合公开信,为开放权重AI模型背书。 瞬间引爆全球科技圈与资本市场。 本次共有英伟达、微软、Meta、IBM 等25 家美国顶级科技企业、投资机构与开源基金会联合署名,成为今年立场最鲜明、影响最深远的 AI 行业宣言。 这封公开信直面当下美国 AI 监管争议,核心推翻了市场固有认知: 一国 AI 竞争力,不在于能否垄断最强闭源大模型,而在于能否搭建开放、可扩散、可自主掌控的全民 AI 生态。 文章以 80 年代开源软件革命为历史参照,论证开放权重模型是技术普及、产业竞争、网络安全与数字主权的核心基石。 公开信明确表态: 盲目封禁开放权重 AI,只会扼杀创新、转移产业优势、制造技术垄断风险。相比于封闭模型的单点隐患,开源体系能够依靠全球开发者共同审查、迭代修复漏洞,反而具备更强安全韧性。 同时,开放模型大幅降低 AI 创业与产业落地门槛,让初创企业、高校与传统行业无需从零训练模型,真正实现技术普惠。 信中最终给出时代定论: AI 未来必须双轨并行,顶尖闭源模型保障极致性能,顶尖开放模型保障生态活力。 附上公开信完整英文原文 + 权威中文精译,全文无删减、无篡改 黄仁勋 X(原 Twitter) 英文原文 For my first post, I‘m sharing a letter @NVIDIAsigned on why open models matter.AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. 中文译文 这是我的第一条帖子,我分享一封英伟达参与联署的公开信,阐释开放模型为何至关重要。人工智能将重塑各行各业,赋能每家企业,并由世界各国共同建设。开放模型能够强化安全与网络防御,加速创新与技术普及,实现技术主权。世界既需要顶尖闭源模型,也需要顶尖开放模型。 中文完整译文(严谨直译) 开放权重模型,即任何人都可下载、审查、修改并在自有基础设施上运行权重的人工智能模型,对维持美国在人工智能领域的领导地位至关重要。 当前华盛顿正在展开一场辩论: 是否应当限制开放权重模型。许多人担忧开放权重可能遭到滥用,这类顾虑值得严肃对待。但限制开放权重将会损害美国的创新能力、市场竞争、网络安全与数字主权,反而让海外竞争对手获得优势。 历史提供清晰参照:上世纪 80 年代开源软件的兴起。早期开源先驱打破了一种观念 —— 软件进步只能依靠严格管控的专有代码。他们搭建起透明生态,全球开发者能够学习、修改、改进共享技术。如今,开源软件支撑互联网绝大部分基础设施,成为各大科技巨头的底层底座,在美国创造数百万就业岗位。 开放权重模型将同样的发展逻辑带到人工智能领域。 初创企业、高校、公共机构与工业主体,无需从零训练前沿大模型,就能够基于顶尖 AI 开展研发。这扩大经济参与机会、降低行业准入门槛、充分激发竞争,创新不再局限于少数资金雄厚的实验室。 评判美国的 AI 领导力,不应以能否掌控单一前沿模型作为标尺,而要看美国能否搭建一套渗透各行各业、稳健开放的人工智能生态。开放权重模型正是这套生态的基石。 批评者提出的风险客观存在: 模型对外发布后,开放权重可以被第三方修改,脱离原始开发者管控。但全面禁止并非正确解决方案。限制措施无法消除风险,只会促使风险向不透明的闭源体系转移,同时倒逼相关产业流向海外。仅仅依靠闭源模型,并不能天然保障安全。闭源系统同样面临黑客入侵、滥用、运行故障等问题;将先进 AI 能力集中在少数封闭体系中,反而制造单点重大风险。 开放体系能够强化安全水平。当成千上万研究者可以审查模型权重,漏洞能够更快被发现与修复。开放权重同时推动数字主权建设:各国与企业能够依托自有硬件运行 AI,不必受制于境外服务商。 政策制定者应当制定针对性防范滥用的保障机制,而非大范围封禁开放权重。监管规则需要区分合法对模型进行适配改造与非法窃取知识产权两类行为。监管机构应当鼓励分层安全机制、信息透明原则与自愿安全标准落地。 想要充分释放人工智能价值,世界既需要顶尖闭源模型,也需要顶尖开放权重模型。限制开放权重,会削弱美国竞争优势;拥抱开放权重,才能持续维系美国创新活力、强化网络安全,守住全球人工智能竞赛中的领先地位。 Open Weights and American AI Leadership《开放权重与美国人工智能领导力》 Open weights models — AI models whose weights anyone can download, inspect, modify, and run on their own infrastructure — are essential to sustaining American leadership in artificial intelligence. Today, a debate is underway in Washington about whether to restrict open weights models. Many fear open weights could enable misuse. These concerns deserve serious consideration. But restricting open weights would undermine U.S. innovation, competition, cybersecurity, and national sovereignty. It would hand an advantage to competitors overseas. History offers a clear parallel: the rise of open-source software in the 1980s. Early open-source pioneers challenged the idea that software progress depended solely on tightly controlled proprietary code. They built transparent ecosystems where developers worldwide could learn, modify, and improve shared technology. Today, open-source software powers most of the internet, underpins every major tech company, and created millions of jobs across the United States. Open weights models bring the same dynamic to artificial intelligence. They let startups, universities, public agencies, and industrial operators build on state-of-the-art AI without training frontier models from scratch. This expands economic participation, lowers barriers to entry, and fuels competition. Innovation is no longer confined to a small set of well-resourced labs. American AI leadership should not be measured by control over a single frontier model. It should be measured by our ability to build a robust, open AI ecosystem deployed across every industry. Open weights models are the foundation of that ecosystem. Critics rightly note risks: once released, open weights can be modified and used outside the original developer’s control. But prohibition is the wrong remedy. Restrictions will not eliminate risk; they will shift risk toward opaque, closed systems and push development offshore. Reliance only on closed models does not guarantee safety. Closed models can be hacked, misused, or fail. Concentrating advanced AI capability in a small number of closed systems creates single points of failure. Open systems strengthen security. When thousands of researchers can inspect model weights, vulnerabilities are found and fixed faster. Open weights also advance digital sovereignty: nations and companies can run AI on their own hardware, independent of foreign providers. Policymakers should pursue targeted safeguards against misuse rather than broad bans on open weights. Rules should distinguish between legitimate adaptation of models and unlawful theft of intellectual property. Regulators should encourage layered security practices, transparency, and voluntary safety standards. The world will need both frontier closed models and frontier open models to maximize AI benefits. Restricting open weights weakens America’s competitive edge. Embracing open weights will sustain U.S. innovation, strengthen cybersecurity, and preserve American leadership in the global AI race. END

查看原文