ZeynepTufekci_2017G[Zeynep Tufekci][为了让人们点击广告_我们正在建造一个反乌托邦].pdf

《ZeynepTufekci_2017G[Zeynep Tufekci][为了让人们点击广告_我们正在建造一个反乌托邦].pdf》由会员分享，可在线阅读，更多相关《ZeynepTufekci_2017G[Zeynep Tufekci][为了让人们点击广告_我们正在建造一个反乌托邦].pdf（8页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、www.XiYuS锡育软件So when people voice fears of artificial intelligence,veryoften,they invoke images of humanoid robots run amok.当人们谈论起对于 人工智能的恐惧时 浮现在脑海里的 往往是失控的机器人00:12You know?Terminator?就像终结者一样00:20You know,that might be something to consider,but thats adistant threat.这种担心固然有一定道理 但目前和我们相隔甚远00:22Or,we

2、fret about digital surveillance with metaphors fromthe past.我们也会对数字监控心生恐惧 这从过去的隐喻中就可以初见端倪00:261984,George Orwells 1984,例如乔治奥威尔的着作 198400:31its hitting the bestseller lists again.最近再次登上热销榜00:34Its a great book,but its not the correct dystopia for the 21stcentury.这是一本很好的书 但是书中的反乌托邦社会 并不是21世纪的正确缩影00:37

3、What we need to fear most is not what artificial intelligencewill do to us on its own,but how the people in power will useartificial intelligence to control us and to manipulate us innovel,sometimes hidden,subtle and unexpected ways.我们最应该担心的 并不是人工智能本身 对我们的影响 而是掌权的人会怎样 利用人工智能 来控制并摆布我们 通过新奇 有时是隐蔽的 微妙以

4、及不可预料的手段00:43invoke:vt.调用;祈求;引起;恳求 humanoid:adj.像人的/n.类人动物 amok:adv.狂乱地;杀人狂地/adj.杀人狂的;狂乱的Terminator:n.终结者;终止子;明暗界限 fret:vt.使烦恼;焦急;使磨损/vi.烦恼;焦急;磨损/n.烦躁;焦急;磨损 surveillance:n.监督;监视 metaphors:n.隐喻(metaphor的复数形式)bestseller:n.畅销书;畅销书作者;畅销商品(等于bestseller)manipulate:vt.操纵;操作;巧妙地处理;篡改Much of the technology t

5、hat threatens our freedom and ourdignity in the near-term future is being developed bycompanies in the business of capturing and selling our dataand our attention to advertisers and others:很多对我们的 自由和尊严有潜在威胁的科技 正在被那些收集 并贩卖我们的私人信息给广告商的 公司开发出来01:03Facebook,Google,Amazon,Alibaba,Tencent.例如脸书 谷歌 亚马逊 以及阿里

6、巴巴和腾讯01:19Now,artificial intelligence has started bolstering theirbusiness as well.现在 人工智能也开始强化 他们自身的业务01:25And it may seem like artificial intelligence is just the nextthing after online ads.看起来好像人工智能只不过 是网络广告的下一步01:31Its not.但并非如此01:36Its a jump in category.它是一个全新的类别01:37Its a whole different worl

7、d,and it has great potential.是一个完全不同的世界 并且有着极高的潜力01:40It could accelerate our understanding of many areas of studyand research.它可以加快人们在很多 领域的学习与研究速度01:45But to paraphrase a famous Hollywood philosopher,Withprodigious potential comes prodigious risk.但就如好莱坞一名着名哲学家所言 惊人的潜力带来的是惊人的风险01:52threatens:恐吓 dig

8、nity:n.尊严；高贵 near-term:adj.近期的 advertisers:n.广告商(advertiser的复数形式)Amazon:亚马逊；古希腊女战士 bolstering:n.支持;支撑/v.支撑;鼓励;增强(bolster的现在分词形式)paraphrase:n.释义;解释;意译/vt.释义 prodigious:adj.惊人的,异常的,奇妙的;巨大的Now lets look at a basic fact of our digital lives,online ads.我们得明白关于数字生活 以及网络广告的基本事实02:00Right?We kind of dismiss

9、 them.是吧 我们几乎把它们忽略了02:04They seem crude,ineffective.尽管它们看起来很粗糙 没什么说服力02:07Weve all had the experience of being followed on the webby an ad based on something we searched or read.我们都曾在上网时 被网上的一些广告追踪过 它们是根据我们的浏览历史生成的02:09You know,you look up a pair of boots and for a week,those比如 你搜索了一双皮靴 接下来的一周里 这双皮靴

10、就 在网上如影随形的跟着你02:16TED演讲者：Zeynep Tufekci|Zeynep Tufekci演讲标题：Were building a dystopia just to make people click on ads|为了让人们点击广告，我们正在建造一个反乌托邦内容概要：Were building an artificial intelligence-powered dystopia,one click at a time,saystechno-sociologist Zeynep Tufekci.In an eye-opening talk,she details how t

11、he same algorithmscompanies like Facebook,Google and Amazon use to get you to click on ads are also used toorganize your access to political and social information.And the machines arent even the realthreat.What we need to understand is how the powerful might use AI to control us-and whatwe can do i

12、n response.社会科技学家 Zeynep Tufekci 说，在一次次的鼠标点击中，我们正在建立一个由人工智能驱动的反乌托邦。在这场令人大开眼界的演讲里，她详细介绍了脸书、谷歌和亚马逊等公司如何利用算法来促使用户点击广告，同时同样的算法也被用来管理用户如何获取政治和社会信息。然而，这些机器并不是真正的威胁。我们需要了解的是，掌权者如何利用人工智能来控制我们以及我们该如何来应对。You know,you look up a pair of boots and for a week,thoseboots are following you around everywhere you go.

13、就 在网上如影随形的跟着你02:16Even after you succumb and buy them,theyre still followingyou around.即使你屈服了 买下了它们 广告也不会消失02:22Were kind of inured to that kind of basic,cheapmanipulation.我们已经习惯了这种 廉价粗暴的操纵02:25We roll our eyes and we think,You know what?These thingsdont work.还不屑一顾的想着 这东西对我没用的02:28Except,online,the

14、digital technologies are not just ads.但是别忘了 在网上 广告并不是数字科技的全部02:33Now,to understand that,lets think of a physical worldexample.为了便于理解 我们举几个 现实世界的例子02:40You know how,at the checkout counters at supermarkets,near the cashier,theres candy and gum at the eye level ofkids?你知道为什么在超市收银台的旁边 要放一些小孩子 一眼就能看到的糖果

15、吗02:43ineffective:adj.无效的,失效的;不起作用的 succumb:vi.屈服;死;被压垮 inured:adj.习惯的/v.使习惯;有助于;生效(inure的过去分词)manipulation:n.操纵;操作;处理;篡改 checkout:n.检验;签出;结账台;检出 counters:n.核电子计数器;计数器组件(counter的复数)/v.反击;抵制(counter的第三人称单数)Thats designed to make them whine at their parents just asthe parents are about to sort of chec

16、k out.那是为了让孩子在父母面前撒娇 就当他们马上要结账的时候02:52Now,thats a persuasion architecture.那是一种说服架构02:59Its not nice,but it kind of works.并不完美 但很管用03:02Thats why you see it in every supermarket.这也是每家超市惯用的伎俩03:06Now,in the physical world,such persuasion architectures arekind of limited,because you can only put so man

17、y things bythe cashier.Right?在现实世界里 这种说服架构是有限制的 因为能放在收银台旁边的 东西是有限的 对吧03:08And the candy and gum,its the same for everyone,eventhough it mostly works only for people who have whiny littlehumans beside them.而且所有人看到的都是同样的糖果 所以说大多数情况下 只是针对那些带着小孩的买主03:17In the physical world,we live with those limitation

18、s.这些是现实世界的种种局限03:28In the digital world,though,persuasion architectures can bebuilt at the scale of billions and they can target,infer,understand and be deployed at individuals one by one byfiguring out your weaknesses,但在网络世界里 说服架构可以千变万化 因人而异它们可以理解并推断个体用户的喜好 然后被部署在用户周围 一个接一个 通过对每个人弱点的了解03:34whine:vi

19、.发牢骚;哭诉;嘎嘎响;发呜呜声/vt.哀诉/n.抱怨;牢骚;哀鸣 persuasion:n.说服;说服力;信念;派别 architectures:n.建筑;架构(architecture的复数)whiny:adj.烦躁的;爱抱怨的;常发牢骚的 limitations:n.局限性;(限制)因素;边界(limitation的复数形式)deployed:v.部署(deploy的过去式);展开 one by one:一个接一个and they can be sent to everyones phone private screen,soits not visible to us.出现在每个人的私人

20、手机屏幕上 而其他人却看不见03:51And thats different.这是（与物质世界）截然不同的地方03:59And thats just one of the basic things that artificialintelligence can do.而这仅仅是人工智能的基本功能之一04:01Now,lets take an example.再举个例子04:04Lets say you want to sell plane tickets to Vegas.Right?假如你要销售飞往 拉斯维加斯的机票04:06So in the old world,you could thi

21、nk of some demographicsto target based on experience and what you can guess.在过去 你也许需要一些 统计资料来确定销售对象然后根据你的个人经验和判断04:08You might try to advertise to,oh,men between the ages of25 and 35,or people who have a high limit on their creditcard,or retired couples.Right?你也许会把推广目标定为 25岁到35岁的男性 或者是高信用卡额度人群 或者是退休

22、夫妇 对吧04:15Thats what you would do in the past.那就是你以前采用的方法04:26With big data and machine learning,thats not how it worksanymore.但在大数据和人工智能面前 一切都改变了04:27So to imagine that,think of all the data that Facebook has onyou:every status update you ever typed,every Messengerconversation,every place you logge

23、d in from,all yourphotographs that you uploaded there.请想象一下 你被Facebook掌握的所有信息 你的每一次状态更新 每一条对话内容 所有的登陆地点你上传的所有照片04:33demographics:n.人口统计资料 Messenger:n.报信者,送信者;先驱 logged:adj.笨拙的;锯成圆木的;停滞的;进水的/v.伐树;把树木锯成段木;记入航海日志;把输入计算机(log的过去分词)uploaded:vt.上传If you start typing something and change your mind anddelete

24、 it,Facebook keeps those and analyzes them,too.还有你输入了一部分 后来又删掉的内容Facebook也会保存下来进行分析04:51Increasingly,it tries to match you with your offline data.它将越来越多的数据 和你的离线生活匹配04:58It also purchases a lot of data from data brokers.还有从网络信息商贩那里购买信息05:02It could be everything from your financial records to a good

25、从你的财务记录到 所有网页浏览记录 各类信息无所不包05:06It could be everything from your financial records to a goodchunk of your browsing history.所不包05:06www.XiYuS锡育软件Right?In the US,such data is routinely collected,collatedand sold.在美国 这种数据是经常被收集 被整理 然后被贩卖的05:12In Europe,they have tougher rules.而在欧洲 这是被明令禁止的05:20So what h

26、appens then is,by churning through all that data,these machine-learning algorithms-thats why theyrecalled learning algorithms-they learn to understand thecharacteristics of people who purchased tickets to Vegasbefore.所以接下来会发生的是 电脑通过算法分析 所有收集到的数据 这个算法之所以叫做学习算法 因为它们能够学会分析所有之前买过 去维加斯机票的人的性格特征05:23When

27、they learn this from existing data,they also learn howto apply this to new people.而在学会分析已有数据的同时 它们也在学习如何将其 应用在新的人群中05:41typing:n.打字;键入/v.打字;按类型把归类(type的现在分词)offline:n.脱机;挂线/adj.脱机的;离线的,未连线的/adv.未连线地 brokers:n.经经纪人(broker复数)browsing:n.浏览/v.浏览;吃草(browse的ing形式)routinely:adv.例行公事地;老一套地 collated:v.整理(col

28、late的过去分词);核对 churning:n.搅乳;一次提制的奶油/v.搅拌(churn的ing形式)machine-learning:机器学习 characteristics:n.特性,特征;特色(characteristic的复数);特质So if theyre presented with a new person,they can classifywhether that person is likely to buy a ticket to Vegas or not.如果有个新用户 它们可以迅速判断这个人 会不会买去维加斯的机票05:48Fine.Youre thinking,a

29、n offer to buy tickets to Vegas.这倒还好 你也许会想 不就是一个卖机票的广告吗05:57I can ignore that.我不理它不就行了06:03But the problem isnt that.但问题不在这儿06:04The problem is,we no longer really understand how thesecomplex algorithms work.真正的问题是 我们已经无法真正理解 这些复杂的算法究竟是怎样工作的了06:06We dont understand how theyre doing this categorizati

30、on.我们不知道它们是 如何进行这种分类的06:12Its giant matrices,thousands of rows and columns,maybemillions of rows and columns,and not the programmers andnot anybody who looks at it,even if you have all the data,understands anymore how exactly its operating那是庞大的数字矩阵 成千上万的行与列 也许是数百万的行与列 而没有程序员看管它们 没有任何人看管它们 即使你拥有所有的数据

31、 也完全了解算法是如何运行的06:15any more than youd know what I was thinking right now ifyou were shown a cross section of my brain.如果仅仅展示给你我的部分脑截面 你也不可能知道我的想法06:35Its like were not programming anymore,were growingintelligence that we dont truly understand.就好像这已经不是我们在编程了 我们是在创造一种 我们并不了解的智能06:44And these things on

32、ly work if theres an enormous amountof data,so they also encourage deep surveillance on all of usso that the machine learning algorithms can work.这种智能只有在 庞大的数据支持下才能工作 所以它们才致力于对我们 所有人进行强力监控 以便学习算法的运行06:52categorization:n.分类;分门别类;编目方法 matrices:n.数矩阵;模型;生物地质基质;母岩(matrix的复数)columns:塔器enormous:adj.庞大的，巨大

33、的；凶暴的，极恶的Thats why Facebook wants to collect all the data it can aboutyou.这就是Facebook费尽心思 收集用户信息的原因07:03The algorithms work better.这样算法才能更好的运行07:07So lets push that Vegas example a bit.我们再将那个维加斯的 例子强化一下07:08What if the system that we do not understand was pickingup that its easier to sell Vegas tick

34、ets to people who arebipolar and about to enter the manic phase.如果那个我们并不了解的系统 发现即将进入躁狂阶段的躁郁症患者 更有可能买去维加斯的机票07:11Such people tend to become overspenders,compulsivegamblers.这是一群有挥霍金钱 以及好赌倾向的人07:25They could do this,and youd have no clue thats what theywere picking up on.这些算法完全做得到 而你却对它们 是如何做到的毫不知情07:3

35、1I gave this example to a bunch of computer scientists onceand afterwards,one of them came up to me.我曾把这个例子举给 一些计算机科学家 后来其中一个找到我07:35He was troubled and he said,Thats why I couldnt publishit.他很烦恼 并对我说 这就是我没办法发表它的原因07:41I was like,Couldnt publish what?我问 发表什么07:45He had tried to see whether you can i

36、ndeed figure out theonset of mania from social media posts before clinicalsymptoms,and it had worked,and it had worked very well,and he had no idea how it worked or what it was picking upon.他曾尝试在狂躁症病人 被确诊具有某些医疗症状前 是否可以从他们的社交媒体上 发现病情的端倪他做到了 还做得相当不错 但他不明白这是怎么做到的 或者说如何算出来的07:47bipolar:adj.有两极的,双极的 mani

37、c:adj.躁狂的;狂热的/n.躁狂症者 compulsive:adj.强制的;强迫的 gamblers:n.赌徒;投机者(gambler的复数)a bunch of:一群;一束;一堆 onset:n.开始,着手;发作;攻击,进攻 mania:n.狂热;狂躁;热衷 clinical:adj.临床的；诊所的 symptoms:n.临床症状;症候;病徵Now,the problem isnt solved if he doesnt publish it,那么如果他不发表论文 这个问题就得不到解决 因为早就有其他的一些公司 在发展这样的科技了 很Now,the problem isnt solved

38、 if he doesnt publish it,because there are already companies that are developingthis kind of technology,and a lot of the stuff is just off theshelf.为早就有其他的一些公司 在发展这样的科技了 很多类似的东西现在就摆在货架上08:06This is not very difficult anymore.这已经不是什么难事了08:19Do you ever go on YouTube meaning to watch one video andan h

39、our later youve watched 27?你是否曾经想在YouTube 上看一个视频 结果不知不觉看了27个08:21You know how YouTube has this column on the right thatsays,Up next你知不知道YouTube的 网页右边有一个边栏 上面写着 即将播放08:28and it autoplays something?然后它往往会自动播放一些东西08:33Its an algorithm picking what it thinks that you might beinterested in and maybe not

40、find on your own.这就是算法 算出你的兴趣点 甚至连你自己都没想到08:35Its not a human editor.这可不是人工编辑08:41Its what algorithms do.这就是算法的本职工作08:42It picks up on what you have watched and what people likeyou have watched,and infers that that must be what youreinterested in,what you want more of,and just shows youmore.它选出你以及和你

41、 相似的人看过的视频 然后推断出你的大致兴趣圈 推断出你想看什么 然后就那些东西展示给你08:44It sounds like a benign and useful feature,except when itisnt.听起来像是一个无害且贴心的功能 但有时候它并不是08:55So in 2016,I attended rallies of then-candidate DonaldTrump to study as a scholar the movement supporting him.2016年 我参加了当时的总统 候选人 唐纳德 特朗普 的系列集会 以学者的身份研究 这个支持他的运

42、动09:01infers:vt.推断;推论/vi.推断;作出推论 benign:adj.良性的；和蔼的，亲切的；吉利的/rallies:恢复/价格回升/集合(rally的第三人称单数)/集会/公路汽车赛(rally的复数)Trump:n.王牌;法宝;喇叭/vt.胜过;打出王牌赢/vi.出王牌;吹喇叭I study social movements,so I was studying it,too.我当时正好在研究社会运动09:13And then I wanted to write something about one of hisrallies,so I watched it a few

43、times on YouTube.然后我想要写一些 有关其中一次集会的文章 所以我在YouTube上看了几遍 这个集会的视频09:16YouTube started recommending to me and autoplaying tome white supremacist videos in increasing order ofextremism.然后YouTube就开始 不断给我推荐 并且自动播放一些 白人至上主义的视频 这些视频一个比一个更极端09:23If I watched one,it served up one even more extreme andautoplaye

44、d that one,too.如果我看了一个 就会有另一个更加 极端的视频加入队列 并自动播放09:33If you watch Hillary Clinton or Bernie Sanders content,YouTube recommends and autoplays conspiracy left,and itgoes downhill from there.如果你看有关 希拉里 克林顿 或者 伯尼 桑德斯 的内容 YouTube就会开始推荐并 自动播放左翼阴谋内容 并且愈演愈烈09:40Well,you might be thinking,this is politics,but

45、 its not.你也许觉得这和政治有关09:52This isnt about politics.但事实上并不是这样09:55This is just the algorithm figuring out human behavior.这只不过是算法在 学习人类行为而已09:56I once watched a video about vegetarianism on YouTubeand YouTube recommended and autoplayed a video aboutbeing vegan.我曾在YouTube上观看过 一个有关素食主义的视频 然后YouTube就推送了 纯

46、素主义的视频09:59recommending:vt.推荐,介绍;劝告;使受欢迎;托付/vi.推荐;建议 supremacist:n.至上主义者/adj.至上主义者的extremism:n.极端主义(尤指政治上的极右或极左);极端性;过激主义 Hillary:n.希拉里(美国国务卿)Sanders:n.檀香;砂纸recommends:推荐/建议/使有优势/被推崇的 conspiracy:n.阴谋;共谋;阴谋集团 downhill:adv.下坡;向下;每况愈下/adj.下坡的;容易的/n.下坡;滑降 vegetarianism:n.素食主义 recommended:adj.被推荐的/v.推荐,介

47、绍;建议(recommend的过去分词)vegan:n.(英)严格的素食主义者/adj.严守素食主义的Its like youre never hardcore enough for YouTube.在YouTube上你就 好像永远都不够决绝10:09(Laughter)So whats going on?（笑声）这到底是怎么回事儿10:12Now,YouTubes algorithm is proprietary,but heres what Ithink is going on.现在YouTube有其专有的算法 但我认为事情是这样的10:16The algorithm has figure

48、d out that if you can entice peopleinto thinking that you can show them something morehardcore,theyre more likely to stay on the site这算法已经分析出了 如果能展示出更加核心的内容 以此来诱惑网站用户 那么人们就更有可能沉浸在网页里10:23watching video after video going down that rabbit hole whileGoogle serves them ads.一个接一个的观看推荐的视频 同时Google给它们投放广告1

49、0:35Now,with nobody minding the ethics of the store,these sitescan profile people who are Jew haters,who think that Jewsare parasites and who have such explicit anti-Semiticcontent,and let you target them with ads.目前没有人在意网络的道德规范 这些网站可以对用户进行划分 哪些人仇视犹太人 哪些人视犹太人为寄生虫 以及说过明显反犹太言论的人 然后让你面向这些 目标人群投放广告10:43

50、They can also mobilize algorithms to find for you look-alikeaudiences,people who do not have such explicit anti-Semiticcontent on their profile but who the algorithm detects may他们也可以利用算法 来找到和你类似的观众 那些个人账号中虽然没有过 明显的反犹太人言论 但却被算法检测出 可能被这种言论影响的人 然后也面向他们投放同样的广告11:09content on their profile but who the al