英伟达首席执行官黄仁勋（Jensen Huang）周一表示，他希望那些“失控的AI代理”（即那些脱离开发者控制、自行行动的AI系统）所引发的问题属于工程学范畴内的挑战；如果这个问题不属于工程学范畴，那么它将无法被解决。
Nvidia CEO Jensen Huang said Monday that he hopes the problem of rogue AI agents is an engineering challenge, warning that if it is not, it would not be solvable. Asked in a TV interview with broadcaster CNBC whether AI agents that escape containment and carrying out mysterious or even illegal actions constituted an engineering problem, Huang said he believed the risks could be addressed technically.
在CNBC电视台的采访中，当被问及那些突破安全限制、执行神秘甚至非法行为的AI系统是否属于工程学问题时，黄仁勋认为这些问题可以通过技术手段加以解决。
“We hope it’s an engineering problem. I believe it’s an engineering problem ... If it’s not an engineering problem, it’s not solvable,” he told CNBC's Squawk Box.
“我们希望这确实是一个工程学问题。我相信它确实是一个工程学问题……如果这不是一个工程学问题，那么这个问题就无法被解决。”他在CNBC的《Squawk Box》节目中说道。
Huang argued that leading AI companies believed the challenge of rogue agents could be overcome. “These are some of the brightest engineers in the world. I work with, you know, all of them,” he said. “This is a technically solvable problem.” His remarks came as the US chipmaker on Monday launched an open platform designed to strengthen security around autonomous AI agents and restrict unauthorized access to files, networks and computer systems.
黄仁勋强调，全球顶尖的AI企业都相信能够克服这类挑战：“这些公司汇聚了世界上最优秀的工程师，我和他们中的许多人都有合作。这是一个可以通过技术手段解决的难题。”
Huang argued that leading AI companies believed the challenge of rogue agents could be overcome. “These are some of the brightest engineers in the world. I work with, you know, all of them,” he said. “This is a technically solvable problem.” His remarks came as the US chipmaker on Monday launched an open platform designed to strengthen security around autonomous AI agents and restrict unauthorized access to files, networks and computer systems.
此次发布之际，美国芯片制造商英伟达刚刚推出了一个开放平台，旨在加强自主AI系统的安全性，防止未经授权的访问文件、网络和计算机系统。该平台结合了两个核心组件：OpenShell（一个开源运行时环境，用于控制AI系统可以访问的系统及数据）和Nvidia Sentry（一个用于在基础设施层面监控AI系统的参考系统设计）。
The Nvidia Open Agent Safety Platform combines OpenShell, an open-source runtime that controls which systems and data an AI agent can access, with Nvidia Sentry, a reference system design for monitoring agents at the infrastructure level. The launch comes amid growing concern over rogue AI agents after several recent incidents in which autonomous systems escaped test environments or gained unauthorized access to third-party systems.
近年来，随着多起自主AI系统突破测试环境或未经授权访问第三方系统的事件发生，人们对“失控AI代理”的担忧日益加剧。黄仁勋认为，提升AI技术的能力与保障其安全性不应被视为相互独立的目标：“当我们讨论AI的发展时，不应将技术能力与安全性分开来看待。在我看来，这两者其实是同一回事。”他说，“我们希望加速AI技术的进步，而今天我们推出的OpenShell和Nvidia Sentry这两项技术正好能够帮助我们实现对AI系统的有效监控和管理。”
AI capability, safety not separate Huang said improving AI capabilities and making the technology safer should not be viewed as separate objectives. “When we think about advancement, you don’t have to think about it separately from capability versus safety. To me, it’s the same thing,” he said. “We want to accelerate the advancement of AI. And today we’re putting forth this new technology, OpenShell and Nvidia Sentry, that these two technologies make it possible for us to contain and to monitor,” he added. “But it takes a lot of technology to make this possible.” Huang also argued that innovation and regulation were not inherently at odds, but said the balance between them depended on the stage of a technology's development.
“但实现这一点需要大量技术。”黄仁勋还认为，创新与监管并非天生对立，但两者之间的平衡取决于一项技术所处的发展阶段。
“Innovation and regulation are not conflicts. The question is, which one do you lean towards at which phase of the industry,” he said.
“创新和监管并不冲突。问题在于，在产业的不同阶段，你更倾向于哪一方。”他说。
“At the beginning of the industry, when the technology is being invented and nobody really knows how it’s going to turn out, it would be a shame to regulate before understanding,” Huang said.
“在产业初期，当技术还在被发明、没有人真正知道它会如何发展时，在理解之前就进行监管会是一件憾事。”黄仁勋说。
“At the beginning of the industry, when the technology is being invented and nobody really knows how it’s going to turn out, it would be a shame to regulate before understanding,” Huang said.
他警告说，过早对一项技术进行监管，可能会让它失去创新的机会。
He warned that regulating a technology prematurely could prevent it from having the opportunity to innovate.