原创翻译:龙腾网 http://www.ltaaa.com 翻译:君子冲盈 转载请注明出处
论坛地址:http://www.ltaaa.com/bbs/thread-479577-1-1.html

Why memory scientists fear the rise of fake video We’re underestimating the mind-warping potential of fake video

为什么记忆科学家害怕假视频的出现:我们低估了假视频扭曲思维的潜力

Updated Jul 24, 2018, 7:47am EDT By Brian Resnick



Doctored photos can easily create false memories. What happens when there’s fake video?
Seeing is believing. And because of this fact, we’re screwed.
Due to advances in artificial intelligence, it’s now possible to convincingly map anyone’s face onto the body of another person in a video.

修改过的照片可以轻易地创造虚假记忆,而当视频也是假的时候会发生什么呢?
眼见为实,但因为这个“ 事实”,我们却被骗了。
由于人工智能的进步,现在可以令人信服地将任何人的脸映射到视频中另一个人的身体上。

In April, BuzzFeed published a demonstration featuring the actor and director Jordan Peele. Using FakeApp, the same tool used in the celebrity face-swapping porn, BuzzFeed took an old video of President Obama and swapped in Peele’s mouth as he performed an impression of Obama. It’s a creepily powerful PSA with a forceful message: “This is a dangerous time. Moving forward, we need to be more vigilant with what we trust from the internet.”

今年4月,BuzzFeed 发布了一场以演员兼导演 Jordan Peele 为代表的示范作品,使用FakeApp—— 这款同样用于替换名人脸的色情工具,BuzzFeed拿了一段前总统奥巴马的旧视频,并用Peele的嘴替换了奥巴马,结果令人印象深刻。这是一个令人毛骨悚然的强大工具:“ 这是一个危险的时刻,展望未来,我们需要对我们从互联网上信任的东西更加警惕。”



As Vox’s Aja Romano has explained, this technique is becoming more common in pornography: An actress’s head can be mapped onto a porn actress’s body. These “deepfakes” can be generated with free software, and they’re different from the photoshopping of the past. This is live action — and uncannily real.

正如 Vox 的 Aja Romano 所解释的那样,这种技术在色情领域变得更加普遍:女演员的头可以被映射到色情演员的身体上,这些“deepfakes”可以用免费软件生成,它们不同于过去的PS,这是真人动作——而且惊人的真实。

And the technology is evolving quickly. In the video above, Peele is essentially acting — via face-swapping software — as a puppeteer for Obama’s face. Not too long ago, this type of software was limited to transferring simple facial expressions and mouth movements from an actor to a fake video. Now, as will be put on display at an motion graphics industry conference next month, the software can account for wide-ranging head and eye movements without much obvious distortion.

而且技术正在迅速发展。在上面的视频中,Peele 本质上是通过面部交换软件来扮演角色,用奥巴马的脸进行的一场木偶表演,不久前,这种软件仅限于将简单的面部表情和嘴部动作从演员转移到一个假的视频,现在,这款软件将在下个月的动态图形行业展会上展出,该软件可以解析大范围的头部和眼睛运动,而不会产生明显的扭曲。

Combine fake audio with fake video and it’s not hard to imagine a future where forged videos are maddeningly hard to distinguish from the truth. Or a future where a fake video of a president incites a riot or fells the market. “We’re not so far from the collapse of reality,” as Franklin Foer summed up at the Atlantic.

把假音频和假视频结合起来,不难想象,未来伪造视频难以分辨出真相,或者将来某个总统的假视频会煽动骚乱或破坏市场,“我们离现实的崩溃并不遥远,”富兰克林·弗尔在《大西洋月刊》上总结出了这样的结论。



We don’t have psychological studies directly looking at the ability of AI-faked video to implant false memories. But researchers have been studying the malleability of our memories for decades.
Here’s what they know: The human mind is incredibly susceptible to forming false memories. And that tendency can be kicked into overdrive on the internet, where false ideas spread like viruses among like-minded people. Which means the AI-enhanced forgeries on the horizon will only make planting false memories even easier.

我们没有直接地研究人工智能伪造视频植入假记忆的能力,但是研究者们几十年来一直在研究我们记忆中的可塑性。
以下是他们所熟知的:人类的大脑非常容易形成错误的记忆。 这种趋势可能会在互联网上被激活,虚假的想法像病毒一样在志同道合的人之间传播, 这意味着,人工智能增强的赝品只会让植入错误记忆变得更加容易。

How a faked photo rewrites our memory

一张假照片如何重写我们的记忆



Do you remember this image? It’s of former President Barack Obama shaking hands with Mahmoud Ahmadinejad, the former president of Iran.
I hope you do not remember this photo, because it never happened. It’s photoshopped.

你还记得这张照片吗?这是一张前总统奥巴马与伊朗前总统马哈茂德·内贾德握手的照片。
我希望你不记得这张照片——因为它从来没有发生过,这张照片是PS的。

Yet in 2010, when Slate asked about 1,000 of its readers whether they remembered seeing the photo, around 21 percent said yes. Another 25 percent said they remembered that the event happened but couldn’t recall specifically seeing the photo.

然而在2010年,当 Slate 杂志询问大约1000名读者是否记得这张照片时,大约21% 的读者给出了肯定答案,另外25% 的人说他们记得事情发生过,但是不记得是否看过这张照片。

Slate also gave its readers some space to write out their feelings about the photo. “I thought Obama was correct, snubbing Ahmadinejad so blatantly would have been a mistake,” one reader wrote, recalling a feeling that certainly never occurred to him.

Slate 杂志也给了读者一些空间来表达他们对这张照片的感觉。 一位读者写道:" 我认为奥巴马是正确的,如此公然地怠慢艾哈迈迪内贾德是一个错误。",一位读者写道,这让他回忆起一种他从未有过的感觉。

Inspired by Loftus’s work, Slate’s William Saletan had created a quiz that presented readers with a mix of photos from real events plus one of five randomly selected fake events.

受到洛夫特斯作品的启发,Slate 杂志的 William Saletan 发起了一个测试,向读者展示了真实事件的混合照片以及五个随机选择的虚假事件中的一个。

One of the fabricated photos showed Sen. Joseph Lieberman voting “guilty” during President Bill Clinton’s impeachment proceedings (Lieberman voted “not guilty”). Another showed President Bush on vacation with baseball player Roger Clemens during the Hurricane Katrina disaster. Bush spent a chunk of time after Katrina hit at his ranch in Crawford, Texas — but not with any baseball players.

其中一张伪造的照片显示,参议员约瑟夫利伯曼在克林顿总统弹劾程序中投票“有罪”( 实际上利伯曼投的是“无罪”票) 。另一张则是布什总统在卡特里娜飓风灾难期间与棒球运动员罗杰·克莱门斯一起度假的虚假照片,在遭受卡特里娜飓风袭击后布什在德克萨斯州克劳福德农场呆了很长时间——但并没有任何棒球运动员一起。

Each faked photo provoked fake memories in at least 15 percent of the respondents. And overall, 50 percent of participants reported they believed the event depicted in the false photo happened. Slate’s project drew the attention of Steven Frenda, a Cal State Los Angeles psychologist who often collaborates with Loftus. Frenda and colleagues took Slate’s data and analyzed it further.

每张伪造的照片都会在至少15% 的受访者中引发虚假记忆。 总的来说,50% 的参与者表示他们相信虚假照片中描述的事件发生过, Slate杂志的项目引起了加州洛杉矶心理学家史蒂文 · 弗兰达的注意,他经常与洛夫特斯合作,弗兰达和他的同事们取得了 Slate 的数据,并进行了进一步的分析。

Here’s what they found: People were more likely to say they remember a faked photo when it fit with their political worldview.

以下是他们的发现:当一张假照片与他们的政治世界观相符时,人们更倾向于记住一张假照片。



“Conservatives were more inclined to believe that Obama had greeted this hostile foreign leader,” Frenda says. And liberals were more likely to recall Bush cavorting with Clemons. “When people encounter events that are complimentary with their political beliefs or preferences, they’re more likely to mistakenly believe those events really happened,” he says.

“保守派更倾向于相信奥巴马曾对这位敌对的外国领导人表示欢迎,”弗兰达说,自由派更有可能记住布什与棒球运动员克莱门斯,他说:“当人们遇到那些与他们政治信仰或偏好相符合的事件时,他们更可能错误地相信那些事件真的发生过。”

Think about what this means. Doctored photos can change the way we remember history. And not just our memories for facts, but possibly even our recollections of what we saw with our own eyes. It means that bad actors may be able to prey on our political biases to change our understanding of world events. And remember, these were Slate readers, who are presumably people who like to keep up with current events.

想想这意味着什么,篡改照片不仅可以改变我们对事实的记忆,也可以改变我们记忆历史的方式,而且甚至可能是我们对我们亲眼所见事实的记忆,这意味着不怀好意者可能会利用我们的政治偏见来改变我们对世界的理解,记住,这些都还只是 Slate 杂志的读者,他们还是那些喜欢跟上时事潮流的人。

Even if people were lying on the survey in order to seem intelligent, the photos could still implant itself in memory. “When people lie or stretch the truth, sometimes that itself distorts their memory,” Frenda says. “So they will sometimes incorporate their lies or exaggerations into what it is to what they think they genuinely remember.”

即使人们为了看起来聪明而在调查中撒谎,但照片仍然可以植入记忆中。“ 当人们说谎或或者夸大事实时,有时候这本身就扭曲了他们的记忆,”Frenda说“ 因此,他们有时会把谎言或夸张的事实融入他们认为自己真正记住的东西上。”

Frenda and his colleagues published this conclusion in 2013 in the Journal of Experimental Social Psychology. And this topic — false memories for political events — hasn’t been directly followed up in the academic literature (though a 2010 paper did find similar results). But this finding checks out with what we know about political psychology: We’re more likely to remember facts that support our teams.

Frenda 和他的同事在2013年的《实验社会心理学》杂志上发表了这个结论。而这个话题——政治事件的虚假记忆——在学术文献中没有直接跟进( 尽管2010年的一篇论文确实找到了类似的结果 ),但是这个发现与我们对政治心理学的了解一致:我们更有可能记住那些支持我们价值观的事实。

“Back in 2013,” Frenda says, “the potential for misusing [or] digitally doctoring images, and now videos,” was not that hot of a topic. Now, “there are a lot of things to worry about.” And there are reasons to believe that the false videos will be an even more powerful way to change how we remember the past.

“ 回顾2013年,”Frenda说,“ 是滥用或通过数字化手段篡改图像,现在是通过篡改视频“,这并不新鲜,现在,“ 有很多事情要担心。”而且有理由相信,虚假视频将会是一种改变我们对过去的记忆的更有效的方式。

Why we form false memories, and how

为什么我们会形成错误的记忆,以及如何形成



We think memory works like a film reel. It doesn’t. Getty Images/EyeEm

图:我们认为记忆就像电影胶片一样,其实不是

Human memory does not function like a videotape or a digital recording. When we remember, we don’t wind back our minds to a moment in time and relive that exact moment.
Instead, memory is constructed.

人类记忆功能不像录像带或数字录音那样,当我们记住某件东西时,我们不会把我们的想法及时地回到一个时刻,重新体验那个确切的时刻。
相反,记忆是构建出来的。

This is a tricky thing to understand. When we call on a memory, we have to piece it back together from disparate pieces of information that exist in our minds. Some of what ends up in our recollection is the truth. But there’s a laziness to our recollections. In reconstructing our memories, our brains often grab the easiest bit of information to recall. And information that we’ve learned since the event will be added to fill in memory gaps.

这是一件很难理解的事情。 当我们唤起记忆的时候,我们必须把它们从存在于我们脑海中的不同的信息片段中拼凑回来。我们记忆中的一些东西是真相,但是我们的记忆有些懒惰,在重建记忆的过程中,我们的大脑常常抓住最容易回忆的信息,我们从某次活动中学到的信息将被添加以填补记忆空白。

Our memory is not like a videotape. A more helpful metaphor is that it’s like a video editor working on a millisecond deadline. In its mad rush, the editor splices in bits of truth with whatever filler is handy. And what’s handy is often our biases, or new information altogether.

我们的记忆不像录像带,一个更恰当的比喻是,它就像一个在毫秒级范围内运作的视频编辑器,在它高速而匆忙的运转中,编辑器用任何方便的填充物将真相拼接成小块,而我们的偏见,或者说新的信息,通常会被用上。



False memories don’t just fudge the fine details. It’s also possible to falsely remember whole events.
That happened in the Slate survey. But researchers have also found that it can happen in studies where the false information is from a trusted source. We trust family, and faked letters from relatives can lead people to believe they were once lost in a mall as a child or that they committed a crime when they were younger. And we trust photos: Experiments that use altered photos to implant memories tend to be more effective than those that do not.

错误的记忆并不仅仅是捏造细节,也有可能错误地记忆整件事。
这在 Slate 会调查中发生了,但是研究人员还发现,这种情况也可能发生在那些虚假信息来自可信的来源的研究中。我们信任家人,而伪造信件的亲戚可以让人们相信他们曾经在商场里迷路过,或者他们年轻时犯过罪——我们相信照片:使用修改过的照片植入记忆的实验往往比那些不使用的更有效。

Here’s what’s really frustrating: We’re confident in our misrememberings. “Once we’ve updated memories, we don’t remember that we’ve done that,” Linda Levine, a University of California, Irvine psychologist who studies memory and emotion. “We have the illusion that we remember things as they happened.”

真正令人沮丧的是: 我们对自己的错误记忆很有信心。 " 一旦我们更新了记忆,我们就不会记得我们曾经做过这样的事情," 加州大学欧文分校研究记忆和情感的心理学家琳达 · 莱文说, " 我们有一种错觉,认为我们记得发生过的事情。"

We’re biased for a sense of continuity, despite the fact that our memories are changing all the time. The video editor in our brains seals the seams up quietly.

尽管我们的记忆一直在改变,但我们对连续性有偏见,我们大脑中的视频编辑器悄悄地把缝合处封了起来。

That’s true for even the most intense events in memory. In the days after 9/11, psychological researchers interviewed thousands of people across the country, asking them simple questions like, “Where were you when it happened?” and then followed up over 10 years. Even as the participants’ memories for the day shifted, as the details they remembered changed, they still remained extremely confident they were remembering the truth.

即使是记忆中最激烈的事件也是如此。911之后的日子里,心理研究者采访了全国数千人,问他们一些简单的问题,比如:“ 911发生的时候你在哪里?”然后追踪询问了10多年,即使参与者们的记忆发生了改变,他们记忆中的细节也发生了变化,但他们仍然非常自信的认为自己记住了真相。

Today, the scientific debate on this topic is around how prent false memory formation is (critics charge that reports of fully implanting a false childhood memory are overinflated), and how to more precisely define memory.

今天,关于这个话题的科学辩论围绕着虚假记忆形成的普通程度( 批评者指责完全植入错误童年记忆的报告被夸大了 ),以及如何更精确地定义记忆。

And researchers propose a few reasons why evolution has favored us to have such malleable memories. One is that it allows us to be creative. We can imagine what the future might look like by merging memories with new information. Another is that saves us effort. We don’t need to remember every detail in our lives.
A third hypothesis highlights the danger of false memories in these polarized times: They help us form a cohesive sense of reality with our groups.

研究者们提出了一些理由来解释为什么进化使我们有了这种可塑性的记忆,其一是它让我们有创造力,通过将记忆与新信息融合,我们可以想象未来会是什么样的,另一个,这可以节省我们的精力,我们不需要记住我们生活中的每一个细节。
第三个假说则突出了虚假记忆在极化时期的危险性:它们帮助我们形成了一个我们与群体之间有凝聚力的现实意识。

Bill Hirst studies collective memory at the New School. His work explains how the process of talking to one another causes us to selectively remember, forget, and create a false collective narrative.
“All of these [memory] distortions allow us to craft a shared representation of the past,” Hirst says. “And because you and I both have a shared representation of the past, it leads to a common view that reinforces a collective identity.”

比尔 · 赫斯特在纽约新学院研究集体记忆,他的工作解释了与他人交谈的过程如何导致我们有选择地记住、忘记和创造一个错误的集体叙述。
赫斯特说:“所有这些[记忆]扭曲使我们能够创造出共同的过去。”,“因为你和我都有过去的共同表现,它导致了一个共同的观点,从而强化了集体身份。”

Social media is fertile ground for us to acquire truly bizarre false memories. Consider the case of the Shazaam truthers.
社交媒体是我们获得真正怪异的虚假记忆的沃土。想想电影《Shazaam》这一例子。

We also have the power to influence each other’s memories and construct new versions of reality as we go.
In 2016, the New Statesman magazine published a feature about a community of Reddit users who are fans of a movie from the 1990s starring the comedian Sinbad (remember him?). The movie is called Shazaam, and Sinbad plays a bumbling genie who adventures with two small kids.

我们也有力量去影响彼此的记忆,并构建新的现实版本。
2016年,《新政治家》杂志发表了一篇关于 Reddit 用户群体的专题,这些用户是90年代由喜剧演员辛巴达( 还记得他吗? ) 主演的电影的粉丝,这部电影叫做《Shazaam》 ,辛巴达扮演了一个笨手笨脚的精灵,和两个小孩子一起冒险。

【译注:关于这个案例和集体虚假记忆,大致是说另一部名叫《Kazaam》的电影被误传为《Shazaam》,实际上根本不存在一部叫《Shazaam》的电影,大家可以参见这篇文章,有比较详细的解释】
【集体虚假记忆:曼德拉效应的背后是什么? http://www.artsbj.com/show-17-540415-1.html 】

The piece, however, was not the typical essay on ’90s pop culture nostalgia. It turns out that the movie Shazaam never existed, and yet many of the people New Statesmen writer Amelia Tait spoke to could not be convinced otherwise. And they were genuinely upset the movie didn’t exist. “It feels like a part of my childhood has now been stolen from me,” one believer told her. And these people insist they’re not simply misremembering the movie Kazaam, which starred Shaquille O’Neal in the role of a genie. (This collective misremembering is sometimes called the “Mandela Effect,” named after the odd phenomenon of people swearing they heard Nelson Mandela died in the 1980s. He died in 2013.)

然而,这篇文章并不是“90年代流行文化怀旧”中的典型文章,事实证明电影《Shazaam》从来没有存在过,但《新政治家》杂志的作者阿米莉亚 · 塔伊特却无法说服很多人,他们真的很失望这部电影根本就不存在,一位信徒告诉她说:“这感觉好像我童年的一部分已经从我身上偷走了。”而这些人坚称,他们并不是简单地记错了电影《Kazaam》,这部电影则是奥尼尔扮演精灵的角色。( 这种集体的自我主义有时被称为“曼德拉效应”,得名于人们发誓他们曾听闻了了纳尔逊 · 曼德拉于1980年代去世的奇怪现象,(实际上)他是2013去世的。)

Here’s one possible reason these redditors believed in the movie so intensely: the Reddit forum itself. Social media brought together like-minded people. “Any type of corroboration is going to strengthen your memory for something, whether it’s true or not, and your confidence in your memory,” Frenda says. These Shazaam truthers were all riffing off one another, giving each other subtle suggestions to form a false narrative that exists in their minds.
It’s a clear of example of how, in an ever-polarizing world, we’re increasingly telling different stories from our neighbors because we’re plugged into separate feeds of information.

这些 Reddit 用户如此强烈地相信这部电影存在的一个可能原因:Reddit 论坛本身,社交媒体聚集了志趣相投的人。
“任何形式的确证都会增强你对某事物的记忆,不管它是真的还是假的,也会增强你对自己记忆的信心”弗伦达说。这些相信 Shazaam 存在的人都在互相欺骗,彼此提供微妙的建议,以形成存在于他们头脑中的虚假叙述。
这是一个很明显的例子,说明在这个日益两极分化的世界里,我们是如何越来越多地讲述从邻居那里听来的不同的故事,因为我们与不同的信息源相连。

Fake video is even scarier than doctored photos

假视频比伪造的照片更可怕



Let’s put this all together. False memories fester when they make sense to our political worldview, when it’s familiar and repeated ad nauseam, when we trust the source of the information, and when this information is corroborated, shared, and discussed by like-minded people.

让我们把这些综合起来,当虚假记忆对我们的政治世界观有意义时,当我们信任信息的来源时,当我们相信信息的来源,当这些信息得到证实、分享,并由志趣相投的人讨论时,错误的记忆就会进一步恶化。

Where else do all these things happen? Social media. Fake stories tend to move more quickly to people on these platforms than the truth, fueled by surprise and bias.
And so what happens when doctored audio and video get added to the mix? (Shudders.)

这些事情还会发生在哪里呢?社交媒体。假新闻往往比真相更快地向这些平台上的人贴近,充满了惊奇和偏见。
那么,当篡改过的音频和视频混合添加在一起会发生什么呢?( 颤抖 )

Most of the studies on false memories were conducted with faked photos or written documents. Since it’s so new, no one has studied the effect of deepfake video yet, but Levine and Loftus suspect it will be even more compelling. “Here that person is, in the midst of that situation, saying and doing these things,” Levine said. If the video is seamless, that’s going to be convincing. And “it’s legitimately worrisome,” she says.
Loftus agrees: “Having the [faked] video, and the richness of it, is just going to exacerbate the false memory potential,” she says.

大部分关于虚假记忆的研究都是基于伪造的照片或者书面文件进行的,由于它是如此的时新,没有人研究过 deepfake 视频的影响,但 Levine 和 Loftus 怀疑它将更加引人注目, Levine 说:“在这种情况下,某个人(看上去)就是这样说和这样做的。”如果视频是无缝的,那么这将是令人信服的,她说:“这是情况确实令人担忧。”
洛夫特斯同意这个说法:" 伪造的视频以及它的丰富性,只会加剧错误记忆的可能性。"

The “collapse of reality” isn’t predestined. But there’s good reason to feel uneasy.

“现实的崩溃”并不是注定的,但我们有充分理由感到不安

We do not yet know how to inoculate people against forming false memories. In fact, it doesn’t seem like any human is immune to forming them. And all around us, the seeds of false reminiscence are being sewn.

我们还不知道如何使人们免于形成错误的记忆。 事实上,似乎没有任何人能免疫它们,而在我们周围,错误记忆的种子正在萌芽。

A December 2017 poll found 44 percent of Republicans believed President Trump repealed Obamacare. That might be because the legislation was confusing to follow as it ping-ponged through Congress. Or because there’s some truth in that parts of the health care law have been weakened. Or because Republicans increasingly don’t trust the media. Whatever the reason, know this: The repeal didn’t happen.

2017年12月的一项民意调查显示,44% 的共和党人认为特朗普总统废除了奥巴马医改,这可能是因为这项立法令人困惑,因为它通过了国会,通过了“平权法案”, 或者是因为医疗保健法的某些部分被削弱了,这是事实,或者是因为共和党人越来越不相信媒体,不管是什么原因,请记住:废除并没有发生。



It’s possible fake video and audio will not become sophisticated enough to have an impact. Perhaps people can learn to spot it. Or maybe someone can design algorithms to squash it. Perhaps the tech giants will become vigilant in banning these forgeries. Or maybe we’ll grow so accustomed to lies that we won’t believe anything — a concept technologist Aviv Ovadya described as “reality apathy” in an interview with BuzzFeed.

假的视频和音频可能不会变得足够复杂,不会产生影响,也许人们可以学会发现它,或者有人可以设计一些算法来压制它,也许科技巨头们会警惕地禁止这些伪造品,或者我们会变得习惯于谎言,以至于我们不会相信任何事情ーー概念技术专家 Aviv Ovadya 在接受 BuzzFeed 采访时称之为" 现实冷漠 "。

We can hope that as much as our memories of the past diverge, there will always be historians getting the permanent record down straight.
But I worry that as our country and our media fracture politically, the divergence in our memories and understanding of facts is also going to grow worse. In 40 years, will fewer of us agree on the facts of history?

我们可以希望,就像我们对过去的记忆有分歧一样,历史学家总能把永久的记录直接写下来。
我担心,随着我们的国家和媒体在政治上的分裂,我们的记忆和对事实的理解上的分歧也会变得更糟,40年后,我们会更少地认同历史事实吗?