请叫我“愚蠢的诺查丹玛斯”。两天前，我参加了朋友帕布罗·托雷的播客节目，讨论我最近关于人工智能的报道。我采访过的许多杰出网络安全专家都确信，迅速发展的AI系统、这些系统通过开源模型的扩散，以及世界软件基础设施摇摇欲坠的现状，将共同导致某种如滚滚浪潮般的混乱。我向帕布罗转述了这些专家的观点：“将会有各种东西失效，扰乱你的日常生活——而这还是你所能指望的最好结果。”
Call me Dumb Nostradamus. Two days ago, I was on my friend Pablo Torre’s podcast to discuss some of my recent reporting on artificial intelligence. A lot of accomplished cybersecurity people I’ve spoken with are convinced that the combination of rapidly advancing AI systems, the diffusion of those systems via open-source models, and the rickety state of the world’s software will lead to something like rolling waves of chaos. Paraphrasing those experts, I told Pablo, “There’s going to be stuff that doesn’t work that disrupts your daily lives—and that’s the best you can hope for.”
而我今晚航班安全降落、乘客和达美航空机组人员都异常冷静的几分钟前后，美国所有国内航班竟被勒令停飞，这恰恰说明：预测混乱和为混乱做好准备是两项截然不同的技能。
That I learned all domestic flights in the United States were being grounded minutes after my own flight this evening landed—safely, and with the country’s chillest passengers and Delta crew—just goes to show that predicting chaos and being prepared for it are two very different skills.
话说回来，我完全不知道今天是什么原因。交通部长肖恩·达菲归咎于通信问题干扰了空管。这可能是一次经典的软件升级乌龙；极罕见的全境停飞指令，恰逢美国联邦航空局（FAA）近期推行的SMART（Strategic Management of Airspace Routes and Trajectories，空域航路与轨迹战略管理）系统——一个耗资8.75亿美元的AI增强型空管管理系统。该系统制造商“空域智能”公司否认其技术有过错。当然，停飞也恰逢昨天的微软Outlook大面积故障。两天内出现这么多网络疑点。
To be clear, I have no idea what caused this. Transportation Secretary Sean Duffy blamed a communications issue for interfering with air traffic control. It could be a classic software-upgrade face-palm; the extremely rare ground stop coincided with the FAA’s recent rollout of SMART (Strategic Management of Airspace Routes and Trajectories), an $875 million AI-enhanced air-traffic-management system. The system’s maker, Air Space Intelligence, denies that its technology was at fault. Of course, the stop also coincided with yesterday’s Microsoft Outlook outage. That’s a lot of cyber smoke in two days.
即使今天有个平淡无奇的解释，我交谈的那些专家也相当确信：未来还会有大量类似的日子到来，主要有两个原因。
Even if today has a boring explanation, the experts I’ve been talking with are pretty sure that plenty of days like this are coming, for two main reasons.
首先，现在任何人都可能成为坏人。从技术层面讲，切断一个国家的电网早就已成可能。过去阻止这种事发生的，是维持冷战不升级为热战的那套逻辑。唯有政府具备制造此类混乱的能力，而政府深知，关掉另一个国家的灯火，招来的回应必是导弹。威慑力作用于那些有所可失、且有明确地址可供还击的对象。AI 的麻烦之一在于，它可能把国家级的破坏能力交到既无所可失、也无固定地址的人手中。
First, anyone can be a bad guy now. Switching off a country’s power grid has technically been possible for a long time. What stopped anyone was the logic that kept the Cold War cold. The only people capable of that kind of chaos were governments, and governments know that turning off another country’s lights is exactly the sort of thing that gets answered with missiles. Deterrence works on parties who have something to lose and an address where you can send a reply. The trouble with AI (um, one of the troubles) is that it may hand state-grade capability to people who have neither.
你无法威慑一个破坏者。你不知道他是谁，他也不在乎，他此刻可能正坐在电竞椅上吃麦片。过去拿着球棒砸邮筒的孩子，现在能搞到算力，关掉一个小镇的冰箱——不是为了钱，也不是为了什么大义，只是心想“为什么不呢？”。历史上，这正是许多人类恶作剧的动机。
You can’t deter a vandal. You don’t know who he is, and he doesn’t care, and he may currently be eating cereal in a gaming chair. The kids who used to grab a bat and swing at mailboxes can now grab some compute and turn off a town’s refrigerators—not for money or a cause, but because why not? Historically, that’s been the motive behind a great deal of human mischief.
其次，我们对抗此类混乱的最佳手段是“切除”。我交谈过的安全专家没人预言会出现末日场景。但他们相当确信，未来 18 个月将充满麻烦，我们依赖的服务将频繁中断。原因在于，当网络防御者发现来不及快速修补的漏洞时，理性的做法就是“切除”：关停系统，静观其变。大型机构早已在这样做，比你想象的更安静、更频繁。
Second, our best defense against this kind of chaos is amputation. None of the security people I’ve talked with has predicted an apocalypse. But they’re pretty convinced that the next 18 months are going to be a hassle, with regular disruptions to services that we rely on. That’s because when cyber defenders find a hole they can’t patch fast enough, the rational move is to amputate: Switch the system off and wait. Big institutions already do this, more quietly and more often than you’d think.
无论今天导致全境航班停飞、以“安全理由”暂停所有起飞、直到查清状况的原因是什么，这都不是系统失灵的信号。越来越多地，这 就是 系统。我刚刚从 22C 座位上亲眼目睹了这一切。
Whatever caused today’s ground stop, halting every departure in the country “for reasons of safety” until someone works out what’s going on isn’t a sign that the system is failing. More and more, it the system. I’ve just watched it from 22C.
你可以采取一些措施来保护自己，但这些都属于“使用牙线”那一类：手机提示更新时就安装更新，用通行密钥代替短信验证码，备一点现金和水。人多力量大，有一定安全性。Gmail 和你的大银行是显而易见的攻击目标，但它们也雇佣了大批非常专业、薪酬极高的人员，他们的全部工作就是寻找麻烦。万一它们倒了，嘿，我们都会面临大得多的问题，届时全世界读你的情书、翻看你的信用卡账单，将不再是最棘手的事。
You can do some things to protect yourself, but they fall into the realm of flossing: Install the update when your phone nags you, use passkeys instead of text-message codes, keep a little cash and water around. There’s some safety in numbers. Gmail and your giant bank are obvious targets, but they also employ armies of very serious, very well-paid people whose whole job is looking for trouble. And if they go down, hey, we’ll all have problems big enough that the world reading your love letters and poring over your credit-card statement will be the least of them.
难以适应的将是这种迷失感。我们大多数人有幸生活在一个和平、繁荣、最重要的是稳定的世界里。我们入睡时无需担心明天红绿灯是否还能正常工作。现在，如果我的消息来源可靠，我们应该做好准备，醒来时身处“软件版贝鲁特”：这里没有什么是时刻坏着的，但任何东西都可能随时坏掉，你最需要的技能就是不感到惊讶。
The disorientation will be harder to adjust to. Most of us have been privileged to live in a world of peace, prosperity, and, above all, stability. We’ve gone to bed without having to wonder whether the traffic lights will work tomorrow. Now, if my sources are correct, we should plan to wake up in Software Beirut, where nothing is broken all the time, but anything might break at any time, and the main skill you need is not being surprised.