21.AI与真相（1720）

Passage Twenty-One

AI and truth 

AI与真相

【1】AI-generated content is raising the premium placed on trust,It is now possible to generate(create) fake but realistic(vivid) content with little more than the click of a mouse.

AI生成的内容可信度越来越高，现在只需点击鼠标就可以生成虚假但逼真的内容。

【2】This can be fun: a TikTok account on which—among other things—an artificial(fake) Tom Cruise wearing a purple robe sings “Tiny Dancer” to (the real) Paris Hilton holding a toy dog has attracted 5.1m followers.

很有意思的是，在一个TikTok账号上，一只穿着紫色长袍的虚拟汤姆·克鲁斯为抱着一只玩具狗的真人帕丽斯·希尔顿演唱歌曲“Tiny Dancer”，该账号吸粉510万，从一众账号中脱颖而出。

【3】It is also a profound change in societies that have long regarded images, video and audio as close to ironclad proof that something is real.

长期以来，人们都将图像、视频和音频视为证明真实性的铁证，而这无疑是一次深刻的社会变革。

【4】Phone scammers now need just ten seconds of audio to mimic(imitate) the voices of loved ones in distress(pain); rogue AI-generated Tom Hankses and Taylor Swifts endorse(support) dodgy products online, and fake videos of politicians are proliferating.

现在，电信诈骗犯只需十秒的音频就可模仿亲信痛苦的声音；AI生成的虚拟汤姆·汉克斯和泰勒·斯威夫特在网上支持假冒伪劣产品，政客的假视频也在网络上漫天飞舞。

【5】The fundamental(essential) problem is an old one. From the printing press to the internet, new technologies have often made it easier to spread untruths(fakes) or impersonate the trustworthy.

基本问题是一个老问题。从印刷术时代再到互联网时代，新技术的诞生往往使得虚假信息的传播和假冒事件变得更轻而易举。

【6】Typically, humans have used shortcuts to sniff out foul play: one too many spelling mistakes suggests an email might be a phishing attack, for example. Most recently, AI-generated images of people have often been betrayed by their strangely rendered hands; fake video and audio can sometimes be out of sync.

通常，人们会使用一些妙招来嗅探其中猫腻。例如，电子邮件中如果出现太多拼写错误说明可能是个钓鱼攻击。举一些更新的例子，AI生成的人物形象会由于手部形象怪异而“露馅”；假视频和音频有时可能音画不同步。

【7】Implausible content now immediately raises suspicion(doubt) among those who know what AI is capable(able) of doing.

对于那些了解AI能力的人来说，不真实的内容会立马引起他们的怀疑。

【8】The trouble is that the fakes are rapidly(quickly) getting harder to spot. AI is improving all the time, as computing power and training data become more abundant(vast).

问题在于，这些假冒内容迅速地变得越来越难以辨别。随着计算能力和训练数据越发充盈，AI也在不断进步。

【9】Could AI-powered fake-detection software, built into web browsers, identify computer-generated content? Sadly not. As we report this week, the arms race between generation and detection favours the forger(faker).

内置在网页浏览器中的AI伪造检测软件能否识别计算机生成的内容？可惜并不能。正如我们本周报道的那样，AI生成和检测之间的角逐竞赛会让伪造者受益。

【10】Eventually AI models will probably be able to produce pixel-perfect counterfeits—digital clones of what a genuine(true) recording of an event would have looked like, had it happened.

最终，AI模型可能能够还原像素完美的赝品，如果事情真的发生了，数字克隆还会还原事件的真实记录。

【11】Even the best detection system would have no crack(bug) to find and no ledge to grasp. models run by regulated companies can be forced to include a watermark, but that would not affect scammers wielding open-source models, which fraudsters can tweak(adjust) and run at home on their laptops.

即使是最好的检测系统也找不到破绽，无法进一步攻破。正规公司运行的模型可以强制添加水印，但这不会影响那些诈骗者使用开源模型，他们可以在家中的笔记本电脑上对这些模型进行调整和运行。

【12】Dystopian possibilities abound. It will be difficult, for example, to avoid a world in which any photograph of a person can be made pornographic by someone using an open-source model in their basement, then used for blackmail—a tactic(method) the FBI has already warned about.

反乌托邦的可能性比比皆是。例如，有人可以在地下室使用开源模型将任何人的照片制作成色情图片，然后进行勒索，尽管联邦调查局已经提示过要注意这种手段，但是这类情况很难避免。

【13】Perhaps anyone will be able to produce a video of a president or prime minister announcing(declaring) a nuclear first strike, momentarily setting the world on edge. Fraudsters impersonating relatives will prosper.

也许任何人都能够制作一段总统或总理宣布首次核打击的视频，瞬间让世界人民坐立不安。冒充别人亲信的诈骗犯也会占据上风。

【14】Yet societies will also adapt to the fakers. People will learn that images, audio or video of something do not prove that it happened, any more than a drawing of it does (the era of open-source intelligence, in which information can be reliably(responsibly) crowdsourced, may be short-lived).

然而，社会也会适应造假者的存在。人们将意识到，某件事的图像、音频或视频并不能证明其切实发生过，就像它的画像一样（在开源情报的时代，信息可以通过可靠的众包方式获取，这样的时代可能会一去不复返）。

【15】Online content will no longer verify itself, so who posted something will become as important as what was posted.

网络上的内容将不再能够验证自身，因此内容发布者将与发布内容本身同样重要。

【16】Assuming trustworthy sources can continue to identify themselves securely—via URLs, email addresses and social-media platforms—reputation(identity) and provenance(source) will become more important than ever.

假设信任源可以继续通过URL、电子邮件地址和社交媒体平台安全地验明自己的身份，那么身份和来源将变得比以往更加重要。

【17】It may sound strange, but this was true for most of history. The era of trusted, mass-produced content was the exception.

这听起来可能很奇怪，但历史上大多如此。值得信赖、大规模生产内容的时代实属凤毛麟角。

【18】The fact that people may soon struggle to spot the invisible hand of AI does not mean the marketplace of ideas is doomed(destined). In time, the fakes that thrive(spread) will mostly be the funny ones.

人们可能很快就难以发现AI的隐形之手，但这并不意味着人类思想市场注定会灭亡。随着时间的推移，流传的冒牌货将大多沦为笑柄。