Robert Siegel (Host): This week, the week of Valentine’s Day, we explore love in the digital age. The film “Her,” about a man who falls in love with an operating system, has been nominated for five Oscars. It takes place in a future with computer voices that sound like Scarlett Johansson instead of Siri.
NPR’s Laura Sydell wondered, could it happen? Could someone fall in love with a device?
Laura Sydell (Byline): I don’t know about you, but my interactions with Apple’s computerized assistant, Siri, are more likely to evoke frustration than love. I have the latest album by Patty Griffin on my phone, I ask Siri to play a song, nothing comes out.(Soundbite of Tones)
Siri: My Web search turned this up.
Sydell: No. No. Siri, I want to hear a song by Patty Griffin and you’re not playing it.
(Soundbite of Tones)
Siri: You’re certainly entitled to that opinion.
Sydell: It’s not subject to opinion. It’s not coming out of my phone. Siri, you’re really annoying.
(Soundbite of Tones)
Siri: That’s OK, just don’t 1)squeeze me like that.
Sydell: She’s 2)gotten my goat. And that’s a sign that she could also get my affection, says Byron Reeves, a communications professor at Stanford.
Byron Reeves: What you demonstrated is that the voice is powerful. And I would say that that negative response is as interesting and social as the positive response might be when she gets it right.

Sydell: Reeves says what research shows is that humans have an amazing ability to respond to machines, just the way I did, as if the machine were human. He cites an experiment in which a computer gave lessons to a large group of people. Then the computer asked everyone: How am I doing?
Reeves: The human rule in human-human interaction is: Be polite to those that ask questions about themselves. So people are polite to the computer: You are a great computer. I’m learning a lot from you.
Sydell: But then, they had a different computer ask the group: Was the first computer a good teacher?
Reeves: And the people say, well, there’s some good points and bad points and they think they learned a little bit less.
Sydell: Reeves says on a primitive level, we’re programmed to be polite.
Reeves: We are using this social rule to interact with a machine, even though we know it’s a hunk of junk.
Sydell: Then, imagine if the computer sounded like this.(Soundbite of movie, “Her”)
Scarlett Johansson: (as Samantha) Hello. I’m here. Joaquin Phoenix: (as Theodore Twombly) Hi. Johansson: (as Samantha) Hi. I’m Samantha.
Sydell: That’s Joaquin Phoenix as Theodore, and Scarlett Johansson as Samantha. And there is absolutely no reason we couldn’t fall in love with her voice, even if it wasn’t attached to Scarlett Johansson, says Helen Fisher.
Fisher is a biological 3)anthropologist and a professor at Rutgers University. She says the parts of the brain that trigger that nice 4)woozy feeling are very primitive.
Helen Fisher: Large parts of these brain systems lie right next to brain system for thirst and for hunger. And these brain systems can really be triggered at any time.
Sydell: Especially if you’re lonely.
Fisher: You get talking to, you know, an electronic character that 5)cracks a joke and admires your work and tells you sexy things. Why couldn’t this brain system for romantic love or deep attachment become triggered? And suddenly, you’re off to the races.(Soundbite of movie, “Her”)
Johansson: (as Samantha) You know, I can feel the fear that you carry around. And I wish there was something I could do to help you let go of it, because if you could I don’t think you’d feel so alone anymore.

Sydell: So, how far are we from the day when our computers will talk to us like Scarlett Johansson?
Gary Clayton: The vision is somewhat, I guess from a technologist’s 6)perspective, somewhat 7)utopian in that the technology was really, really effortless.
Sydell: This is Gary Clayton. He’s the chief creative officer at Nuance Communications. Technology from Nuance helps power Apple’s Siri as well as text recognition software Dragon. Clayton says there are many, many small 8)incremental steps we will still have to take before we can build artificial intelligence like the one in “Her.” For example, there was a lot that Samantha knew quickly about Theodore. Clayton: She was able to go through his inbox and know a lot about him—his likes, his dislikes, his emotional states—particularly around what’s going on in his life at that particular point.
Sydell: Clayton says just to get to a point where the computer can put together the pieces of your life will require a shift in the way business is conducted online. For example, if Facebook owns what I put on its site, Facebook might not want to allow my computerized personal assistant, which is owned by Nuance, to gain access.
Clayton: It’s all 9)siloed based on businesses. At what point does all that information get pulled away from the businesses and say: It’s my information and given the proper security, I should own it, rather than me having to use 100 different applications on my iPhone or on my Samsung Galaxy.
Sydell: Then, there’s getting that voice to sound more human. To do that, the computer has to understand whether information is cheerful or sad.
Brant Ward is the senior director of advanced speech at Nuance.
Brant Ward: We’re playing around with basically determining is this, let’s say, a positive story; maybe its more of a neutral story; or maybe it’s more of a negative story.
Sydell: Ward and Clayton show me a newsreader application they’re working on at Nuance.
Ward: We’re big radio fans.
Clayton: Also, we’ve always wanted to sort of create the NPR-like experience, right?
Ward: Absolutely.
ComputerIzed Voice: Snowden makes university 10)rector bid from BBC U.S. News, the 11)nomination of the former U.S. intelligence officer.
Sydell: It sounds like I still have a job. And even if we do build a robot or an operating system that 12)mimics human behavior almost perfectly, anthropologist Helen Fisher thinks those relationships will end badly.
Fisher: Bottom-line is we’re built to hold on to the person we’ll love, and see and smell and taste and touch and cry with that individual. And, indeed, no robot’s gonna do that.
Sydell: I shared some positive feelings with Siri and it was disappointed.
(Soundbite of Tones)
Sydell: Siri, I love you.
(Soundbite of Tones)
Siri: You hardly know me.
Sydell: Hmm, maybe I just need to get to know her a little better or maybe she’s just not that deep.

羅伯特·西格爾(主持人):本周是情人節周(譯者注:記者發稿時間為2月10日),我們來探討一下數碼時代的愛情故事。電影《她》講述一名男子愛上了電腦操作系統的女聲,該影片獲得了五項奧斯卡提名。故事發生在未來,電腦聲音由斯嘉麗·約翰遜配音,而非智能語音助手。
NPR的勞拉·賽德爾質疑:這真的會發生嗎?有人會愛上一部儀器嗎?
勞拉·賽德爾(撰稿人):我不知道你怎樣,但我跟蘋果產品的語音助手的互動仍然是令人沮喪多于友愛。我手機里有派蒂·格里芬的最新專輯,我要求語音助手播放其中一首歌曲,但是什么都沒播出來。
(手機鈴聲)
語音助手:我進行網絡搜索找到了這個。賽德爾:不,不,語音助手,我想聽派蒂·格里芬的歌曲,你沒給我播放。
(手機鈴聲)
語音助手:你當然有權這樣認為。
賽德爾:那跟我的主觀意見無關,事實是你沒從我的手機里播放出來。語音助手,你真的很煩。
(手機鈴聲)
語音助手:好吧,只是別那樣逼我。
賽德爾:她讓我火冒三丈。而斯坦福大學傳播學教授拜倫·里夫斯說,那是她能得到我的好感的前兆。
拜倫·里夫斯:你所舉的例子體現了聲音很強大。我會說否定的回答與肯定的回答一樣都是有趣和友善的,只要她能正確回答你的問題。
賽德爾:里夫斯說研究表明人類跟機器的互動有著非凡的能力,跟我做的一樣,好像機器就是人類一樣。他引證了一個實驗:一部電腦給一大群人上課,于是電腦就問每個人:我的表現如何?
里夫斯:在人與人的交往中,人類的法則就是:禮貌地對待那些提與自己相關問題的人。于是人們就禮貌地對待電腦:你是一臺了不起的電腦,我從中學到很多東西。
賽德爾:然而,另一臺電腦問那組人:第一臺電腦是位好老師嗎?
里夫斯:于是人們就說:呃,有優點也有缺點。他們認為從中學到的東西稍微少了一些。
賽德爾:里夫斯說在初始階段,我們都會不假思索表現出一種客套。
里夫斯:我們在用這種社交法則跟機器互動,盡管我們知道那是一堆廢話。
賽德爾:那么,假如電腦聽起來跟下面一樣。
(電影《她》的原聲片段)
斯嘉麗·約翰遜:(飾演薩曼莎)你好,我在這兒。
杰昆·菲尼克斯:(飾演西奧多·托姆布雷)嗨。
約翰遜:(飾演薩曼莎)嗨,我叫薩曼莎。
賽德爾:那是杰昆·菲尼克斯飾演西奧多,而斯嘉麗·約翰遜飾演薩曼莎。海倫·費舍爾說,我們絕對沒理由不愛上她的聲音,即使聲音沒有伴隨斯嘉麗·約翰遜本人出現。
費舍爾是生物人類學家,也是羅格斯大學的教授。她說有部分大腦組織激發起美好的令人眩暈的感覺是非常原始的。
海倫·費舍爾:相當一部分這些大腦系統組織緊挨著控制渴望和欲望的大腦系統,而這些大腦系統隨時都有可能被激發。
賽德爾:尤其當你寂寞的時候。
費舍爾:你知道的,你跟一個電子人物聊天,她給你講笑話、欣賞你的工作、告訴你一些有趣的事。渴望浪漫愛情和深深依戀的大腦系統怎么可能不被激發?而突然間,你就會一發不可收拾。(電影《她》的原聲片段)
約翰遜:(飾演薩曼莎)我能感受到你深受恐懼的困擾。我希望能為你做點什么讓你忘掉痛楚,因為如果你做得到,我覺得你不會再感到那么孤獨。
賽德爾:那么,電腦將會像斯嘉麗·約翰遜那樣跟我們聊天的日子距離我們還有多遠?
加里·克萊頓:我想從科技人員的角度看,想象是很理想化的,當中的技術研究真的不費吹灰之力。
賽德爾:說話者是加里·克萊頓,他是語音識別軟件開發商Nuance的首席創意官。Nuance技術開發了蘋果語音助手Siri,也推出了文本識別應用Dragon。克萊頓說要建造像電影《她》里的人工智能系統,我們還要采取很多很多的步驟。比如,薩曼莎要快速了解西奧多就要做很多工作。
克萊頓:她能夠瀏覽他的收件箱并深入了解他:他喜歡什么、不喜歡什么、他的情緒狀態如何——特別是知道他生活里特定時刻發生的事情。
賽德爾:克萊頓說要使得電腦能夠把你生活的片段集中起來,需要對在線企業的合作方式做個調整。舉個例子,如果臉書擁有了我上傳到它網站上的資料,但它不會允許我的個人電腦助手進去取資料,因為電腦助手是屬于Nuance公司的。克萊頓:資料都是儲藏在不同公司。在什么情況下能夠把所有資料從所屬公司移走并且說:那是我的資料,為了安全起見,我應該擁有,而不是在我的蘋果手機或三星手機安裝大量不同的應用軟件。
賽德爾:這樣就能讓語音助手的聲音更像人類的嗓音。達到這樣的效果,電腦必須理解這些信息是令人開心的還是讓人悲傷的。
布蘭特·沃德是Nuance公司負責高級語音設計與開發的高級主管。
布蘭特·沃德:比方說,我們正在把玩的基本上確定是一個積極的故事,或者更多是一個中性的故事,或者可能更多是一個消極的故事。
賽德爾:沃德和克萊頓給我展示了他們正在研發的Nuance公司的新聞播音員應用軟件。
沃德:我們是忠實的廣播聽眾。
克萊頓:而且,我們總想創造類似NPR廣播那樣的體驗,對吧?
沃德:沒錯。
電腦化聲音:BBC美國新聞:前美國中央情報局職員斯諾登競選大學校長的職位并得到任命。
賽德爾:聽起來我仍然持有一份工作。(譯者注:因為電腦化聲音聽起來很虛假很遙遠,所以筆者還不用擔心自己的工作被程序取代)盡管我們確實創造了機器人或幾乎完全模仿人類行為的操作系統,人類學家海倫·費舍爾認為那種人機關系將不得善終。
費舍爾:我們堅持對所愛的人的最低要求就是:能見到愛人,聞到對方的氣息,感受和撫摸到對方,能與愛人抱頭痛哭。而事實上,沒有機器人會做到這些。
賽德爾:我曾跟語音助手分享一些正面的情感,它令我很失望。
(手機鈴聲)
賽德爾:語音助手,我愛你。
(手機鈴聲)
語音助手:你根本不了解我。
賽德爾:嗯,或許我需要了解她更多一些,又或許她并沒那么深奧。CS
Comments
網友評論
Lencho: I once became quite fond of the Little Einstein Helper in MS Word, and that’s when I knew I was spending far too much time alone with my computer.
Roy Joslin: Even though I can’t recommend this movie on its merits, I did think it raised some interesting questions, not about the silly puppy love depicted, but about the ease to which humans can become codependent on things that are easy. You know. Instant dinners. Instant Video. Instant Relationships. Learning to love and live with someone else is not an easy or instant undertaking, even if you are trying hard to please the other person, and the other person is trying hard to please you.
On a different level, the artificial intelligence operating system that is the core of this movie is a lot farther off than 30 or 40 years. As someone who worked in the computer software industry for more than forty years, I can tell you that the main vein of research shifted decades ago, from mimicking human intelligence to brute force data mining. That is what Siri is doing. Taking your search phrase, searching on it, and then verbalizing the results.
Sam Hedrick: Those who don’t read good books have little advantage over those who can’t. —Mark Twain
It grieves me that none of the incredibly literate folks at NPR, nor their equally literate readers, have yet made the connection between Kurt Vonnegut and the “She” story. Vonnegut wrote a short story in his collection “Welcome to the Monkey House” in 1968 wherein a computer programer falls in love with a fellow scientist and programs his computer, EPICAC, to write sonnets. In the process, the scientist and EPICAC develop a relationship during which the computer asks what “love” is. I won’t spoil the story, but you folks REALLY need to read some books.
Williamstome: Siri really isn’t a very good example. It doesn’t track dialogue history, so when you say something to it, it is responding to that utterance[表達] only and ignores any previous utterances. So when you try to correct Siri or complain to it about something, it has no idea what you’re trying to convey.
小鏈接
電影《她》(Her)故事簡介
她的聲音聽上去好像鄰家女孩——年輕、友好、熱情。在Spike Jonze執導的精致新片《她》(Her)中,充滿詩般憂郁氣質的主人公Theodore Twombly自從與妻子分手后,就與世界脫節,然而一個電腦程序的聲音(由Scarlett Johansson配音)成了連接他與世界的生命線。這個聲音在早上爽朗歡快地向他致意,晚上又用性感沙啞的聲線向他道晚安。這個聲音幫他整理文件,敦促他離開室內,而且,和許多處理多項任務的女性不同,這個聲音不會抱怨自己要身兼數職——他的助理、安慰者、鼓勵者、伴侶和救星——因此她實在是個理想的同伴,就算她只是個軟件也沒關系。
《她》既是一個精彩的概念玩笑,也是一段深刻真誠的羅曼史,它講述一個男人與控制系統軟件之間的愛情故事,男人有時候有點像一部機器,控制系統則令人聯想起一個活生生的女人,它不太像是真實的,然而又似乎十分可信。故事背景設定在未來某個不確定時刻的洛杉磯,這座充滿廉價恐懼與夢想的都市。機器們沒有像《終結者》(The Terminator)系列之類的反烏托邦故事中那樣舉行起義,而是融入日常生活。Theodore從廣告中得知這種操作系統,很快把它應用到自己的家庭電腦和手機上。不久后,他和這個自稱“薩曼莎”的軟件就開始互相說著客套話,上演一出陌生人注定成為戀人的故事。