细思极恐的AI换脸我们该如何应对？.docx

《细思极恐的AI换脸我们该如何应对？.docx》由会员分享，可在线阅读，更多相关《细思极恐的AI换脸我们该如何应对？.docx（6页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、细思极恐的AI换脸，我们该如何应对？Freddy HuCSE9课程导读近日，欧盟公布新规定，要求科技公司加强打击虚假信息，尤其是针 对“深度伪造（Deepfake）和虚假账号的行为，否那么将处以巨额罚 款。其中的“深度伪造”技术，在国内互联网也已经不是什么新鲜事， 之前在抖音上爆火的“中国马斯克“，最近上了热搜的“微笑特效”，都 运用了“深度伪造”技术。这项技术会引发什么问题？我们应该如何看 待这一新现象？ 一起来听今天的讲解。本文封面图片来自ing.dk,如有版权问题，请联系删除。本篇课程首发于2022年6月29日英文原文Deepfake Technology Is Now a Threat

2、 to Everyone. What Do We Do? “深度伪造”技术威胁着每个人。我们该怎么办？By Kartik HosanagarLast October; MIT Prof. Sinan Aral warned his Twitter followers that he had discovered a video of himself that he hadnt recorded endorsing an investment funds stock-trading algorithm. In reality, it wasnt Prof. Aral in the video,

3、but an artificial-intelligence creation in his likeness, or what is known as a highly persuasive deepfake.去年10月，麻省理工学院教授希南阿拉尔在推特上提醒他的粉丝们 说，他发现了一段自己并没有参与录制的视频，他在其中推销一个投 资基金的股票交易算法。事实上，视频里的并不是阿拉尔教授本人， 而是一个照着他的样子，用人工智能创造的人，也叫作“深度伪造”，足以以假乱真。It is striking that scammers targeted Prof. Aral, considering h

4、e is a leading expert on the study of misinformation online. It also suggests that deepfake technology is now at an inflection point: Thanks to a number of free deepfake apps that are just a Google search away anyone can become a victim of such a scam.阿拉尔教授本身就是研究网上虚假信息的顶级专家，诈骗者竟以他为 目标，真是非同寻常。这也意味着，“

5、深度伪造”技术已经来到了一 个转折点：只要谷歌搜索一下，就能找到许多免费的“深度伪造”应 用程序，因此任何人都可能成为此类骗局的受害者。The term deepfake has come to mean the use of AI to create synthetic media (images, audio, video) in which someone appears to be doing or saying what in reality they haven!t done or said.目前，“深度伪造” 一词指的是使用人工智能来创立合成媒体(图 像、音频、视频)，其中的人会

6、做他在现实中没做过的事，说他在现 实中没说过的话。One such solution is to detect deepfakes via machine-learning methods. While these detectors can be successful in the short term, people looking to evade such systems will likely just respond with better technology, creating a continuing and expensive cat-and-mouse game. 一种解

7、决方法是通过机器学习的方法来检测“深度伪造短期内， 检测可以奏效。但想要设法避开此类系统的人，很可能会以更强的技 术进行应对。这么一来，就会衍生出一场永无止境、耗资巨大的猫鼠 游戏。A better approach with a longer time horizon is media provenance or authentication systems to verify the origins of images and videos.从更长的时间维度来看，更好的方法是通过“媒体出处记录或验证系 统”来核实图像和视频的来源。While legislation eventually m

8、ay offer protection against deepfakes, I believe the market could be quickerprovided we, as consumers and citizens, care.为了防范“深度伪造”，也许最终会有立法层面的措施。但我相信，只要作为消费者和公民的我们关心这个问题，市场便能更快地做出响 应。生词好句.deepfake英dizpfeik美dirpfeikn.深度伪造拓展：深度伪造（Deepfake）是英文deep learning（深度学习）和fake （伪造）的混合词，即利用深度学习算法，实现音视频的模拟和伪 造。1

9、.endorse英mdDis美mdairsV.（本文）（在广告中对某种产品）宣传，吹捧；（公开）支持拓展：He endorsed the presidents new policy.他公开支持总统的新政策。They paid a lot of money to the pop star to endorse the new product.他们给这个明星付了很多钱，让这个明星在广告中推销他们的新产 品O3.investment fund投资基金4.1ikeness英laiknas美laiknasn.（外表的）相似拓展：His likeness to that man is astonishin

10、g.他和那个人长得很像，太令人惊讶了。5.or what is known as以.名称而被人所知的6.striking英straikiij美straikiij adj.惊人的,（让人）惊讶的 7.scammer英skaems美skaemsrn.诈骗犯，骗子拓展：scam v./n,诈骗scam somebody欺骗，诈骗某人8.1eading英liidiij美liidiijadj.顶级的9.misinformation英mismf31mei/n美.mismfsrmeifnn.假消息lO.inflection point拐点拓展：inflectionUK /mflekfn/ US /mflek

11、fn/n.（尤指词尾的）屈折变化.thanks to归因于.11 .evade英fveid美fveidv.躲避，回避13.provenance英provnsns美praivnansn.起源，出处拓展：something is of uncertain provenance 是来历 不明的14.authentication英Dientfkeijh美ai.Oentskeifnn.认证拓展：authenticUK /DiOentik/ US /a:0entik/ adj.真品的，真的外刊原文Deepfake Technology Is Now a Threat to Everyone. What D

12、o We Do?By Kartik HosanagarThe Wall Street JournalUpdated Dec. 7, 2021In October; MIT Prof. Sinan Aral warned his Twitter followers that he had discovered a video of himself that he hadnt recorded endorsing an investment funds stock-trading algorithm. In reality it wasnt Prof. Aral in the video, but

13、 an artificial-intelligence creation in his likeness, or what is known as a highly persuasive deepfake.”It is striking that scammers targeted Prof. Aral, considering he is a leading expert on the study of misinformation online. It also suggests that deepfake technology is now at an inflection point:

14、 Thanks to a number of free deepfake apps that are just a Google search away, anyone can become a victim of such a scam.The term deepfake has its origins in pornography, but it has come to mean the use of AI to create synthetic media (images, audio, video in which someone appears to be doing or sayi

15、ng what in reality they havent done or said. The technology isnt always misused. Cadbury for example, joined with Bollywood celebrity Shahrukh Khan on a marketing campaign for small businesses in India hit by Covid-19. Business owners uploaded details of their stores, and Cadbury used deepfake techn

16、ology to create the effect of Mr. Khan promoting them in tailored TV ads. (The campaign was transparent about its fakery).But positive use cases are likely to be overshadowed in coming years by the technologys potential role in financial fraud, identity theft and worse一from the savaging of reputatio

17、ns to the stoking of civil and political unrestCurrent laws targeting fraudulent impersonation werent designed for a world with deepfake technology, and efforts at the federal level to update these laws have faltered so far. One stumbling block is the need to also protect parodies and other free spe

18、ech.Another big challenge is that in an online world where people can anonymously upload content, it can be difficult to find the individuals behind deepfakes. Some researchers have proposed putting the onus on website platforms such as Facebook and YouTube by making their protections in relation to

19、 user-generated content conditional on their taking reasonable steps to police their own platforms.Broad adoption of these kinds of laws could create meaningful deterrents一eventually. But the technology is moving so fast that lawmakers will likely always lag behind. That is why I believe we are goin

20、g to have to rely on technology to protect us from a problem it helped create.One such solution is to detect deepfakes via machine-learning methods. For instance, while deepfakes appear highly realistic, the technology isnt yet capable of generating natural eye blinking in the impersonated individua

21、ls. As such, machine-learning algorithms have been trained to detect deepfakes using eye-blinking patterns. While these detectors can be successful in the short term, people looking to evade such systems will likely just respond with better technology, creating a continuing and expensive cat-and-mou

22、se game.A better approach with a longer time horizon is media provenance or authentication systems to verify the origins of images and videos. Microsoft, for instance, has developed a prototype of a system called AMP (Authentication of Media via Provenance) that enables mediacontent creators to crea

23、te and assign a certificate of authenticity to their content. Under such a system, every time you watch a video of, say the U.S. president, the technology would help your browser or mediaviewing software verify the source of the video (for example, a news network or the White House), The process cou

24、ld be delivered as simply as through an icon-much like the current browser padlock icon that indicates any information you send to that particular website is protected from third-party tampering en route. To be effective in practice, such systems would have to be widely adopted by all content creato

25、rs, which will take time.While legislation eventually may offer protection against deepfakes, I believe the market could be quicker一provided we/ as consumers and citizens, care.Reprinted by permission of The Wall Street Journal, Copyright 2022 Dow Jones & Company, Inc. All Rights Reserved Worldwide.Original Date of Publication: Dec. 7, 2021