克里斯多弗•斯坦納《自動化：算法統(tǒng)治世界》書評：計算機(jī)算法正掌控世界，左右人們生活的方方面面，大到投資決策，小到出行路線的選擇。算法帶來了效率，但同時也蘊藏著風(fēng)險。騎士資本的交易算法出錯就曾在一夜之間蒸發(fā)了數(shù)億美元。算法統(tǒng)治世界，因此它也能摧毀世界。
早上你醒來后馬上就查看Gmail郵箱，其中郵件的排列順序是由谷歌專門開發(fā)的算法決定的；午餐時分，你登錄Deadspin網(wǎng)站閱讀體育類八卦消息，該網(wǎng)站又使用另一套專門算法來推廣最積極的評論者；吃完晚飯，你從一份精選電影節(jié)目單中選一部觀看——友好的Netflix算法早已根據(jù)你過去選看電影的口味選出了你可能喜歡的新片。其他一些算法則幫助我們決定沿哪條路出行、聽什么音樂乃至股票會以什么方式波動。
簡而言之，算法正在接管我們的生活。所謂算法，只是一套軟件代碼而已，運作方式類似于決策樹，即考慮多種變量，然后產(chǎn)生結(jié)論或推薦。“機(jī)器人程序”（bot）通常是一系列算法的集合。如果不像克里斯多弗•斯坦納那樣退一步看，人們很難意識到算法近幾年來的發(fā)展多么迅猛、影響多么深遠(yuǎn)，也無法認(rèn)識到它對現(xiàn)代文明的影響。斯坦納的著作《自動化：算法統(tǒng)治世界》（Automate This: How Algorithms Came to Rule Our World）闡明了這些問題。
斯坦納原是《福布斯》（Forbes ）撰稿人，三年前他決定闡釋交易算法（其最簡單的形式是，根據(jù)操作者輸入的數(shù)據(jù)決定何時買/賣）對華爾街的巨大影響。那時，極為復(fù)雜的交易機(jī)器人程序已經(jīng)變革了金融市場。隨后斯坦納又決定將他的報道拓展到金融領(lǐng)域之外，結(jié)果就寫成了這樣一本無所不包的著作，描繪機(jī)器開始統(tǒng)治包括婚戀交友服務(wù)、音樂、醫(yī)藥在內(nèi)的各行各業(yè)的情形。
| |
| In the morning you wake up and check your Gmail, which is sorted for you courtesy of a proprietary Google (GOOG) bot. At lunchtime you read sports gossip on Deadspin, which deploys another set of proprietary algorithms to promote the most prolific commenters. After dinner you pick a movie from a menu of choices that a friendly Netflix (NFLX) algorithm queued up for you based on its record of your cinematic taste. Other algorithms help determine which streets we drive, the music we hear, and which way stocks move.
In short, algorithms are taking over our lives. An algorithm is simply a piece of software code that operates like a decision tree, considering multiple variables and then spitting out a decision or recommendation. (A bot is typically a collection of algorithms.) Without taking a step back, as Christopher Steiner does in Automate This: How Algorithms Came to Rule Our World, it's hard to appreciate how fast and far algorithms have come in recent years, and what the consequences are for modern culture.
Steiner, a former Forbes writer, set out three years ago to explain how trading algorithms (which, in their simplest form, make buy/sell decisions based on various data inputs) had overtaken Wall Street. By that time, wildly complex trading bots had transformed financial markets. Steiner then decided to expand his reporting outside finance. The result is an encompassing tale of how industries as diverse as dating services, music and medicine all came to be ruled by machines.
|
有家叫Savage Beast的公司很能說明問題。該公司創(chuàng)立于上世紀(jì)90年代，其它的運作方式是付費請數(shù)百名音樂人聽歌，然后依據(jù)400項音樂特質(zhì)（包括節(jié)奏、音調(diào)及眾多其他要素）對其進(jìn)行歸類。Savage Beast試圖向Tower Records及百思買（Best Buy）之類音樂相關(guān)產(chǎn)品零售商銷售其音樂推薦服務(wù)，但銷路慘淡。該公司差點就沒能活過2000年的網(wǎng)絡(luò)股泡沫破滅潮，2005年時已經(jīng)奄奄一息。此后，它轉(zhuǎn)用算法、而不是非真正的音樂人來生成音樂推薦信息，并搖身一變，改名為Pandora。，2011年，這家該公司上市，市值高達(dá)30億美元。
eLoyalty是另一家發(fā)展歷程體現(xiàn)了算法威力的公司。這家客戶管理咨詢公司從事的業(yè)務(wù)平庸無奇——給呼叫中心提供建議。eLoyalty的算法能通過掃描一個擁有約200萬說話方式的數(shù)據(jù)庫，界定來電者的個性。，如此，銷售代表或服務(wù)專員能立即了解來電客戶較為情緒化還是相對理性，并采取相應(yīng)的銷售或服務(wù)技巧。沃達(dá)豐（Vodaphone）簽約使用了eLoyalty的服務(wù)，此后，其接線員就可有針對性的提供服務(wù)。，比如，面對情緒化的顧客時，需要用小道消息套近乎，才能讓他們對升級服務(wù)感興趣；而面對更善于分析的客戶時，只需談?wù)劮⻊?wù)的價值定位就行了。應(yīng)用eLoyalty服務(wù)后，沃達(dá)豐的升級服務(wù)率提升了8,600%。
盡管斯坦納援引了大量案例，但他似乎并不是這個新奇世界的優(yōu)秀向?qū)АＳ捎谠摃茨軐Ｗ⒂谌A爾街或醫(yī)藥界之類的某一個特定領(lǐng)域，講清楚算法到底是如何顛覆其原有模式的。，它只好覆蓋太多不同行業(yè)，結(jié)果是有些材料顯得陳腐。比如書中有一章講述音樂品味的自動化，當(dāng)中連上世紀(jì)90年代和21世紀(jì)初的報紙上的報道都摘錄了。同樣，美國國家航空航天局（NASA）開發(fā)個性檢測系統(tǒng)以便利太空任務(wù)宇航員團(tuán)隊的遴選還是上世紀(jì)七八十年代的事情。
我真心希望能愛上這本書，因為這個充滿互聯(lián)網(wǎng)機(jī)器人程序的新世界既令人擔(dān)憂，又引人入勝。我們這個世界的運作，越來越取決于華爾街、Facebook、谷歌（Google）和亞馬遜（Amazon）如何部署其算法。可是，盡管斯坦納撰寫了很多例子，講述機(jī)器人程序?qū)ξ覀兩畹挠绊懀摃奶刭|(zhì)和敘述方式仍然無法使人愛不釋手。相反，它讀起來就像一篇寫的太長的讀書報告。
《自動化：算法統(tǒng)治世界》出版的時機(jī)（8月30日上市）既可說幸運，也可說不幸。很多美國人仍在熱議騎士資本（Knight Capital）造成的混亂，該。這家公司的交易算法出錯，一夜之間就造成了幾億美元的損失。可是，騎士資本事件所提出的問題，該書并未回答。騎士資本的算法問題只是影響了幾只股票而已，可要是醫(yī)療保健行業(yè)最終也部署機(jī)器人程序來給我們開藥方，那系統(tǒng)會不會還出故障？很少有人深入探討過算法普及的缺陷，而斯坦納也放過了這個話題。
機(jī)器人程序一旦進(jìn)駐，就不會撤走。不管人們將其應(yīng)用到哪個領(lǐng)域，算法都能帶來效率、巧妙與速度。可與絕大多數(shù)其他突破性創(chuàng)新一樣，它們已開始體驗到成長中的陣痛。既然算法已經(jīng)統(tǒng)治世界，那緊接著就應(yīng)該擔(dān)心它們的缺陷是否會毀掉世界。
譯者:小宇
| |
| There's Savage Beast, a 1990s startup that paid hundreds of musicians to listen to songs and classify them according to some 400 musical attributes, including rhythm, tonality, and much more. Savage Beast tried without luck to sell its music recommendation service to music retailers like Tower Records and Best Buy (BBY). The company barely survived the 2000 dotcom bust and was on life support by 2005, when it started to produce music recommendations using algorithms instead of live musicians. Along the way, Savage Beast changed its name to Pandora (P). In 2011 it went public with a $3 billion valuation.
ELoyalty is another company whose story shows the power of algorithms. The customer management consultant deals in the stodgy business of advising call centers. ELoyalty's algorithms scan a database of about two million speech patterns to classify callers by personality. As a result, sales and service reps can instantly tell if a customer is more emotional or more thought-driven, and tailor their pitches accordingly. Vodaphone (VOD) signed on to eLoyalty's program, and afterward its operators knew if they were talking to an emotional customer who needed chummy gossip to get interested in upgrades, as opposed to more analytic clients who only wanted to hear about the value proposition. After adopting eLoyalty, Vodaphone's sales upgrades increased by 8,600%.
Despite his wealth of case material, Steiner turns out to be an uncertain guide to this newfangled world. Because the book lacks a narrow focus on how algos are upending, say, Wall Street or the medical field, it tries to cover too many industries. As a result, some of the material feels stale. A chapter on the automation of musical taste, for instance, includes stories told in newspapers in the 1990s and early 2000s. Similarly, NASA's personality-detecting system, which helped the space program pick teams of astronauts, was developed in the 1970s and 1980s.
I really wanted to fall in love with this book, for the new world of bots is at once alarming and engrossing. Increasingly, our world is being shaped by how Wall Street, Facebook (FB), Google, and Amazon (AMZN) deploy their algorithms. But while Steiner has written an exhaustive account of the bots powering our lives, the book lacks the characters and narrative to be a page-turner. Instead it feels like a book report that ran long.
The timing of Automate This (available Aug. 30) is both lucky and unlucky. Half of America is still talking about the fiasco at Knight Capital, where trading algorithms went haywire and caused the firm to lose several hundred million dollars overnight. Yet the Knight Capital story raises questions the book doesn't answer. Knight's algo issues only affected a few stocks. But if the health care industry eventually deploys bots to prescribe our medicines, for example, can we expect similar glitches? There's a downside to this story that's rarely been explored, and Steiner lets it pass.
Once bots move in, they don't move out. Algorithms have brought efficiency, craftiness, and speed to nearly everything that humans have tasked them with. But as with most breakthrough innovations, they have experienced growing pains. Now that algorithms rule the world, the next story will be how their shortcomings might destroy it.
|
|
