How deepfakes undermine truth and threaten democracy | Danielle Citron

翻译人员: Nan Yang 校对人员: Yolanda Zhang

[This talk contains mature content]

【此演讲包含成人内容】

Rana Ayyub is a journalist in India

Rana Ayyub 是一位印度记者。

whose work has exposed government corruption

她的工作揭露了政府腐败

and human rights violations.

和人权侵犯。

And over the years,

这些年来，

she's gotten used to vitriol and controversy around her work.

她已经习惯了工作中的残酷和争议。

But none of it could have prepared her for what she faced in April 2018.

但这些都不足以让她准备好来 面对2018年4月份的事情。

She was sitting in a café with a friend when she first saw it:

当时她和一个朋友坐在 咖啡厅，第一次看见了

a two-minute, 20-second video of her engaged in a sex act.

她自己出现在一个 2分20秒的性爱视频里。

And she couldn't believe her eyes.

她不能相信自己的眼睛。

She had never made a sex video.

她从来没有拍摄过这样的视频。

But unfortunately, thousands upon thousands of people

但是很不幸的是，成千上万的人

would believe it was her.

选择相信就是她。

I interviewed Ms. Ayyub about three months ago,

我在大约三个月前 采访了Ayyub女士，

in connection with my book on sexual privacy.

为了我的关于性隐私的书籍。

I'm a law professor, lawyer and civil rights advocate.

我是一个法律教授， 律师和民权倡导者。

So it's incredibly frustrating knowing that right now,

所以我非常沮丧，因为我知道

law could do very little to help her.

现在的法律几乎帮不到她。

And as we talked,

当我们在谈话的时候，

she explained that she should have seen the fake sex video coming.

她解释道她应该更早意识到 虚假性爱视频的到来。

She said, "After all, sex is so often used to demean and to shame women,

她说：‘’毕竟，性爱经常 被用来贬低和羞辱女性，

especially minority women,

特别是少数民族的妇女，

and especially minority women who dare to challenge powerful men,"

尤其是敢于挑战权势 的少数民族妇女。”

as she had in her work.

就像她在工作所做的那样。

The fake sex video went viral in 48 hours.

那个伪造的性爱视频 在48小时之内像病毒一样传播。

All of her online accounts were flooded with screenshots of the video,

她所有的线上账户 都被这些视频截屏淹没，

with graphic rape and death threats

同时还有图片形式的强奸和死亡威胁

and with slurs about her Muslim faith.

和对她穆斯林信仰的诽谤。

Online posts suggested that she was "available" for sex.

网上的帖子说她可以 随意跟其他人进行性行为。

And she was doxed,

并且她已经被“人肉”，

which means that her home address and her cell phone number

也就是说她的家庭住址和手机号

were spread across the internet.

已经在互联网上随处可见。

The video was shared more than 40,000 times.

那个视频已经被分享了超过4万次。

Now, when someone is targeted with this kind of cybermob attack,

如今，这样的网络暴力，

the harm is profound.

其伤害是非常深远的。

Rana Ayyub's life was turned upside down.

Rana Ayyub的生活已经彻底改变。

For weeks, she could hardly eat or speak.

几周时间里， 她几乎不吃饭也不说话。

She stopped writing and closed all of her social media accounts,

她不再写文章， 关闭了所有社交媒体账户，

which is, you know, a tough thing to do when you're a journalist.

这是作为一个记者很难应付的事情。

And she was afraid to go outside her family's home.

她不敢走出家门。

What if the posters made good on their threats?

如果那些发帖的人 真的进行了威胁呢？

The UN Council on Human Rights confirmed that she wasn't being crazy.

联合国人权理事会 确认她没有精神问题。

It issued a public statement saying that they were worried about her safety.

他们发布了一个声明， 表示担心她的人身安全。

What Rana Ayyub faced was a deepfake:

Rana Ayyub面对的是deepfake ： (“深度学习”和“伪造”的混合词)

machine-learning technology

一种机器学习技术，

that manipulates or fabricates audio and video recordings

能够操纵或者伪造视频和音频

to show people doing and saying things

来展示人们做了或者说了

that they never did or said.

他们从来没有做过或说过的事情。

Deepfakes appear authentic and realistic, but they're not;

Deepfake让这些看起来 是真实的，但事实上并不是；

they're total falsehoods.

这些都是虚假的视频。

Although the technology is still developing in its sophistication,

尽管这项技术还在 这种复杂性中发展，

it is widely available.

但已经广泛被使用。

Now, the most recent attention to deepfakes arose,

大家最近对deepfakes最关注的，

as so many things do online,

就像很多网上的事情一样，

with pornography.

是色情媒介。

In early 2018,

在2018年早期，

someone posted a tool on Reddit

有人在Reddit论坛上发布了一个工具，

to allow users to insert faces into porn videos.

可以让用户在色情视频中插入人脸。

And what followed was a cascade of fake porn videos

这促成了一系列伪造的色情视频，

featuring people's favorite female celebrities.

使用了人们最喜爱的女星的面貌。

And today, you can go on YouTube and pull up countless tutorials

现在，你在YoutTube上能搜索到

with step-by-step instructions

不计其数步骤详细的教学视频

on how to make a deepfake on your desktop application.

教你如何在电脑上制作这种视频。

And soon we may be even able to make them on our cell phones.

很快，我们也许也可以 在我们的手机上制作。

Now, it's the interaction of some of our most basic human frailties

此时，正是我们最基本的 人性弱点和网络工具的

and network tools

相互作用

that can turn deepfakes into weapons.

使得deepfakes变成了武器。

So let me explain.

让我解释一下。

As human beings, we have a visceral reaction to audio and video.

作为人类，我们对音频 和视频有直观的反应。

We believe they're true,

我们相信反应是真实的，

on the notion that of course you can believe

理论上当然你应该相信

what your eyes and ears are telling you.

你的眼睛和耳朵告诉你的事情。

And it's that mechanism

并且正是那种机制

that might undermine our shared sense of reality.

削弱了我们共有的对现实的感觉。

Although we believe deepfakes to be true, they're not.

尽管我们相信deepfakes 是真的，但它不是。

And we're attracted to the salacious, the provocative.

我们被那些淫秽和挑衅吸引了。

We tend to believe and to share information

我们倾向于去相信和共享

that's negative and novel.

消极和新奇的信息。

And researchers have found that online hoaxes spread 10 times faster

研究者们发现网上的骗局比真实故事

than accurate stories.

的传播速度快10倍。

Now, we're also drawn to information

我们也容易被迎合

that aligns with our viewpoints.

我们观点的信息吸引。

Psychologists call that tendency "confirmation bias."

心理学家称这种倾向为 “验证性偏见”。

And social media platforms supercharge that tendency,

同时社交媒体平台 极大鼓励了这种倾向，

by allowing us to instantly and widely share information

允许我们即刻并广泛地分享

that accords with our viewpoints.

符合我们自我观点的信息。

Now, deepfakes have the potential to cause grave individual and societal harm.

Deepfakes能够对个人 和社会造成严重的潜在危害。

So, imagine a deepfake

想象一个deepfake视频

that shows American soldiers in Afganistan burning a Koran.

显示美国士兵在阿富汗焚烧可兰经。

You can imagine that that deepfake would provoke violence

你可以想象这个视频会挑起

against those soldiers.

针对这些士兵的暴力行为。

And what if the very next day

然后如果第二天

there's another deepfake that drops,

又有另一个deepfake视频出现，

that shows a well-known imam based in London

展示在伦敦的一位有名的 伊玛目（伊斯兰教领袖的称号）

praising the attack on those soldiers?

在赞美对士兵的袭击行为呢？

We might see violence and civil unrest,

我们可能会看见暴力和内乱，

not only in Afganistan and the United Kingdom,

不仅出现在阿富汗和英国，

but across the globe.

而是全世界。

And you might say to me,

你可能会对我说，

"Come on, Danielle, that's far-fetched."

“拜托， Danielle，你说的太牵强了。 ”

But it's not.

但并不是。

We've seen falsehoods spread

我们已经看见了许多虚假信息

on WhatsApp and other online message services

在WhatsApp和其它 在线聊天服务里传播，

lead to violence against ethnic minorities.

导致了对少数民族的暴力。

And that was just text --

而那还仅仅是文字——

imagine if it were video.

如果是视频呢？

Now, deepfakes have the potential to corrode the trust that we have

Deepfakes有可能会腐蚀 我们在民主机构里

in democratic institutions.

的所拥有的信任。

So, imagine the night before an election.

想象一下在选举的前一晚，

There's a deepfake showing one of the major party candidates

Deepfake展示了 其中一个主要党派的候选人

gravely sick.

生了重病。

The deepfake could tip the election

它可以颠覆这个选举

and shake our sense that elections are legitimate.

并动摇我们认为选举合法的想法。

Imagine if the night before an initial public offering

接下来，想象在一个跨国银行的

of a major global bank,

首次公开募股的前夜，

there was a deepfake showing the bank's CEO

deepfake显示银行的CEO

drunkenly spouting conspiracy theories.

因为喝醉宣扬阴谋论。

The deepfake could tank the IPO,

这个视频可以降低IPO的估值，

and worse, shake our sense that financial markets are stable.

更糟的是，动摇我们认为 金融市场稳定的感觉。

So deepfakes can exploit and magnify the deep distrust that we already have

所以，deepfakes可以利用 并放大我们对政客，商业领袖

in politicians, business leaders and other influential leaders.

和其他有影响力的领导人 已有的深深的不信任性。

They find an audience primed to believe them.

它们找到了愿意相信它们的观众。

And the pursuit of truth is on the line as well.

对真相的追寻也岌岌可危。

Technologists expect that with advances in AI,

技术人员认为， 未来随着人工智能的进步，

soon it may be difficult if not impossible

想要分辨出真视频和假视频

to tell the difference between a real video and a fake one.

可能将会变得非常困难。

So how can the truth emerge in a deepfake-ridden marketplace of ideas?

那么真相如何才能在一个deepfake 驱使的思想市场中产生？

Will we just proceed along the path of least resistance

我们仅仅沿着阻力最小的路前进

and believe what we want to believe,

并相信我们想去相信的，

truth be damned?

而真相已经被诅咒了吗？

And not only might we believe the fakery,

而且不仅是我们相信虚假，

we might start disbelieving the truth.

可能是我们也开始不相信真相。

We've already seen people invoke the phenomenon of deepfakes

我们已经见过人们 利用deepfakes的现象

to cast doubt on real evidence of their wrongdoing.

来质疑能证明他们 错误行为的真实证据。

We've heard politicians say of audio of their disturbing comments,

我们也听过政客回应 他们令人不安的评论，

"Come on, that's fake news.

“拜托，那是假新闻。

You can't believe what your eyes and ears are telling you."

你不能相信你的眼睛 和耳朵告诉你的事情。”

And it's that risk

这是一种风险，

that professor Robert Chesney and I call the "liar's dividend":

Robert Chesney教授和我 把它叫做“撒谎者的福利”：

the risk that liars will invoke deepfakes

撒谎者利用deepfakes来

to escape accountability for their wrongdoing.

逃避他们为错误行为负责的风险。

So we've got our work cut out for us, there's no doubt about it.

所以我们知道我们的工作 非常艰难，毋庸置疑。

And we're going to need a proactive solution

我们需要一个 主动积极的解决方案，

from tech companies, from lawmakers,

来自于科技公司，国会议员，

law enforcers and the media.

执法者和媒体。

And we're going to need a healthy dose of societal resilience.

并且我们将需要健康的社会复原力。

So now, we're right now engaged in a very public conversation

当下我们正在与科技公司

about the responsibility of tech companies.

进行关于社会责任的公开谈话。

And my advice to social media platforms

我对社交媒体平台的建议

has been to change their terms of service and community guidelines

已经影响了他们的 服务条款和社区准则，

to ban deepfakes that cause harm.

来禁止deepfakes造成危害。

That determination, that's going to require human judgment,

那种决心需要人类的判断，

and it's expensive.

并且代价很大。

But we need human beings

但是我们需要人类

to look at the content and context of a deepfake

去看看deepfake的内容和背景，

to figure out if it is a harmful impersonation

弄清楚它是否是有害的伪装，

or instead, if it's valuable satire, art or education.

或者相反，是否是有价值的 讽刺，艺术或教育。

So now, what about the law?

接下来，法律呢？

Law is our educator.

法律是我们的教育家。

It teaches us about what's harmful and what's wrong.

它教育我们什么是有害的， 什么是错误的。

And it shapes behavior it deters by punishing perpetrators

法律通过惩罚肇事作案者 和保护对受害者的补救

and securing remedies for victims.

来规范行为。

Right now, law is not up to the challenge of deepfakes.

现在，法律还无法应对 deepfakes带来的挑战。

Across the globe,

在全球范围内，

we lack well-tailored laws

我们缺少精心制定的法律，

that would be designed to tackle digital impersonations

解决数字伪造问题的法律，

that invade sexual privacy,

针对会侵犯性隐私，

that damage reputations

毁坏名誉

and that cause emotional distress.

并造成情绪困扰的事件的法律。

What happened to Rana Ayyub is increasingly commonplace.

发生在Rana Ayyub身上的 事情越来越普遍。

Yet, when she went to law enforcement in Delhi,

但是，当她去德里 的执法部门举报时，

she was told nothing could be done.

她被告知他们无能为力。

And the sad truth is that the same would be true

并且令人悲伤的事实是， 在美国和欧洲

in the United States and in Europe.

可能是同样的情况。

So we have a legal vacuum that needs to be filled.

所以我们需要填补这一法律空白。

My colleague Dr. Mary Anne Franks and I are working with US lawmakers

我和同事Mary Anne Franks博士 正在和美国国会议员一起

to devise legislation that would ban harmful digital impersonations

制定可以禁止有危害的数字伪造，

that are tantamount to identity theft.

相当于身份盗窃的法律。

And we've seen similar moves

同时我们看见了

in Iceland, the UK and Australia.

在冰岛，英国和澳大利亚 有类似的过程在进行。

But of course, that's just a small piece of the regulatory puzzle.

但是当然，这只是 整个监管难题的一小部分。

Now, I know law is not a cure-all. Right?

我知道法律并不是灵丹妙药。

It's a blunt instrument.

它是一个武器。

And we've got to use it wisely.

我们需要智慧地使用它。

It also has some practical impediments.

法律也有一些实际的障碍。

You can't leverage law against people you can't identify and find.

你不能利用法律对抗 你无法找到和定位的人。

And if a perpetrator lives outside the country

如果一个作案者居住在

where a victim lives,

不同于受害者的国家，

then you may not be able to insist

那么你可能无法强制作案者

that the perpetrator come into local courts

来到当地法庭

to face justice.

来接受审判。

And so we're going to need a coordinated international response.

因此我们需要一项协调一致的， 联合全球的响应策略。

Education has to be part of our response as well.

教育也是我们响应策略中的一部分。

Law enforcers are not going to enforce laws

执法者不会去强制执行

they don't know about

他们不知道的法律，

and proffer problems they don't understand.

也不会提出他们不明白的问题。

In my research on cyberstalking,

在我关于网络追踪的研究中，

I found that law enforcement lacked the training

我发现执法者缺少

to understand the laws available to them

了解可供使用的法律

and the problem of online abuse.

和网络暴力问题的培训。

And so often they told victims,

所以他们经常告诉受害者，

"Just turn your computer off. Ignore it. It'll go away."

“把电脑关掉就行了， 不要管，自然就没了。”

And we saw that in Rana Ayyub's case.

我们在Rana Ayyub的案例中 也发现了这点。

She was told, "Come on, you're making such a big deal about this.

她被告知： “拜托，你太小题大做了。

It's boys being boys."

他们只是不懂事的孩子。”

And so we need to pair new legislation with efforts at training.

所以我们需要展开针对新法律的培训。

And education has to be aimed on the media as well.

媒体也同样需要接受相关的培训。

Journalists need educating about the phenomenon of deepfakes

记者们需要接受关于 deepfakes现象的培训，

so they don't amplify and spread them.

这样它们就不会被放大和传播。

And this is the part where we're all involved.

这正是我们所有人 需要参与其中的部分。

Each and every one of us needs educating.

我们每个人都需要这方面的教育。

We click, we share, we like, and we don't even think about it.

我们点击，我们分享，我们点赞， 但是我们没有认真去思考。

We need to do better.

我们需要做得更好。

We need far better radar for fakery.

我们需要针对虚假事件 更敏锐的雷达。

So as we're working through these solutions,

所以在我们研究解决方案的时候，

there's going to be a lot of suffering to go around.

会有很多痛苦和折磨围绕着我们。

Rana Ayyub is still wrestling with the fallout.

Rana Ayyub仍在和后果搏斗。

She still doesn't feel free to express herself on- and offline.

她还是无法自在地 在线上和线下表达自己。

And as she told me,

就像她告诉我的，

she still feels like there are thousands of eyes on her naked body,

她仍然感觉有上千双眼睛 在看着她的裸体，

even though, intellectually, she knows it wasn't her body.

尽管理智上她知道 那并不是她的身体。

And she has frequent panic attacks,

而且她经常遭受惊吓，

especially when someone she doesn't know tries to take her picture.

特别是当有陌生人试图去偷拍她。

"What if they're going to make another deepfake?" she thinks to herself.

“如果他们想拍来做另一个 deepfake呢？”她不禁想。

And so for the sake of individuals like Rana Ayyub

所以为了像Rana Ayyub一样的个人，

and the sake of our democracy,

为了我们的民主，

we need to do something right now.

我们现在就要有所作为。

Thank you.

谢谢大家。

(Applause)

（掌声）