一位司機正開著車子在高速公路上行駛，突然她想到，我這周的周末要干什么來著？于是她通過語音助手功能要求車載系統(tǒng)查看一下她的日程安排。幾秒鐘后，汽車喇叭里來傳來聲音，告訴她系統(tǒng)已經(jīng)自動回復了一個朋友的生日宴會邀請，并且建議她帶上一件生日禮物——一件底特律雄獅隊的球衣。

于是，這名司機要求數(shù)字助手下單訂購這件球衣。不到兩分鐘，下單已經(jīng)成功了。與此同時，車內(nèi)的這段對話已經(jīng)被發(fā)送到了一個遠程數(shù)據(jù)中心，用于針對NFL比賽的廣告分析——即便這位司機并不喜歡橄欖球。

語音識別技術并不是什么新生事物，只不過以前的語音識別技術很不完善，識別的正確率也不高。但近年來，這項技術得到了突飛猛進的發(fā)展，并且迅速與汽車技術進行融合，使駕駛員能夠使用自然語言發(fā)布更加廣泛的指令。有了人工智能的加持，新的“個人助理”功能可以執(zhí)行從日程管理到在線網(wǎng)購的一系列功能，在這個過程中，司機的眼睛可以始終看著道路，只用語言下達指令。

與智能家居設備同步后，語音識別軟件還可以遠程控制恒溫器、燈和門鎖。雖然亞馬遜Alexa、蘋果Siri、谷歌助手等語音識別軟件為有車一族帶來了許多便利，但由此帶來的隱私問題和個人數(shù)據(jù)采集問題也是十分值得關注的。

位于華盛頓的消費者權益保護機構電子隱私咨訊中心（Electronic Privacy Information Center）的律師克里斯汀·班南認為：“消費者有充分的理由擔心車載語音識別系統(tǒng)的隱私問題。有人說消費者是歡迎這項新技術的，這說明他們并不擔心隱私問題。這是一種詭辯?！?/p>

IHS Markit公司的數(shù)據(jù)顯示，到明年，搭載有語音識別系統(tǒng)的汽車預計將達到1.2億輛，是2018年的4000萬輛的3倍。目前，大多數(shù)車載系統(tǒng)都是基于智能手機的，不過亞馬遜和谷歌也在開發(fā)不需要與手機綁定的車載模塊。

汽車行業(yè)研究機構AutoPacific的副總裁埃德·金指出：“現(xiàn)在，就連最便宜的車型上也能使用蘋果的CarPlay和安卓的Android Auto?？梢韵胂螅迥曛畠?nèi)，大多數(shù)新車都將搭載某種形式的語音識別功能?！?/p>

Inching down the freeway, a driver realizes she’s forgotten her weekend plans and asks her digital copilot to check her calendar. Within seconds, a voice emanates from the car’s speakers telling the driver that she has RSVP’d to a friend’s birthday party and then suggesting a gift—a Detroit Lions jersey.

The entire process ends less than two minutes after the driver asks the digital assistant to order the jersey. Meanwhile, the dialogue inside the car is sent to a distant data center, ready to be mined for ads pitching an NFL game, even though the driver dislikes football.

Once clunky and buggy, voice-recognition technology is improving and quickly spreading to the dashboard, allowing drivers to issue a wider range of commands using natural speech. Designed to keep drivers’ eyes on the road, “personal assistants” powered by artificial intelligence can perform a variety of functions, from managing calendars to placing Amazon Prime orders.

When synced with smart-home gadgets, the software can also turn up the thermostat, turn off the lights, and lock the doors from miles away. But if the benefits of bringing Amazon Alexa, Apple’s Siri, or Google Assistant into the car seem endless, so do concerns about privacy and how the information collected will be used.

“Consumers have good reasons to be concerned about the use of voice-recognition systems in cars,” says Christine Bannan, an attorney for the Electronic Privacy Information Center, a consumer advocacy group in Washington, D.C. “There is a false narrative that because consumers embrace new technology, that means they are indifferent about privacy.”

An estimated 120 million cars will be equipped with voice-recognition systems next year, up from 40 million in 2018, according to IHS Markit. Currently, most in-car systems are smartphone-based, but Amazon and Google are also developing embedded devices that can work in cars without being tethered to a phone.

“Apple CarPlay and Android Auto are available in even the cheapest models now,” says Ed Kim, a vice president at automotive research firm AutoPacific. “I would imagine that within five years, the vast majority of new vehicles will have some form of voice recognition.”

這也意味著，當附近的商店打折時，司機通過語音或屏幕上的提示，可以迅速獲知相關促銷信息。再舉個例子，當司機想喝咖啡時，他可以先通過語音助手下單，這樣等他將車子開到星巴克時，他的大杯拿鐵已經(jīng)做好了。

車載語音系統(tǒng)的普及，也給很多制造企業(yè)、營銷機構和其他企業(yè)帶來了利好。一旦人工智能技術能夠更好地理解各種口音和方言、俗語，汽車就可以對駕駛員的行為產(chǎn)生更深入的見解，并將這些信息與其他企業(yè)進行共享。

語音識別技術使汽車能夠輕易收集你的購物習慣和你常去的地方的信息，但同時這些信息也可能被用來制作有針對性的視頻或音頻廣告，或是銷售訂閱服務，甚至是被泄露給第三方企業(yè)。對于亞馬遜、谷歌和蘋果這種大公司來說，他們通過汽車收集的用戶數(shù)據(jù)越多，就有更多的機會將這些數(shù)據(jù)賣給其他客戶。

目前，各大汽車廠商都在加速收割語音識別技術的紅利。有些廠商放棄了亞馬遜、蘋果和谷歌的技術，轉而研發(fā)自己的車載語音識別系統(tǒng)。比如今年，梅賽德斯奔馳和寶馬都將推出自己的獨家產(chǎn)品?？萍脊九c汽車廠商圍繞車主數(shù)據(jù)的爭奪必將愈演愈烈。

埃德·金表示：“未來幾年，用戶數(shù)據(jù)及其可能產(chǎn)生的收益，將成為雙方爭奪的戰(zhàn)場。”

雖然語音識別系統(tǒng)一般都需要一個“喚醒詞”才能被喚醒，但這個過程也并非是萬無一失的。在全球用戶使用語音識別系統(tǒng)的過程中，也曾出現(xiàn)過許多各種各樣的問題和狀況，有些狀況比較搞笑，而有些狀況則只能用恐怖來形容。最令人擔憂的是，在某些案例中，系統(tǒng)不小心將人機對話發(fā)送給了車主手機通訊錄中的聯(lián)系人。在個別案例中，系統(tǒng)意外下單購買了大量商品。亞馬遜已經(jīng)采取了一定措施，以避免這些意外狀況的發(fā)生——比如允許用戶針對語音識別程序設置網(wǎng)購密碼，在接打電話時先要求用戶確認等。

目前，很多汽車廠商都表示，他們在使用用戶數(shù)據(jù)進行營銷，或將用戶數(shù)據(jù)與第三方共享時，會首先征得用戶的許可。比如沃爾沃在一份聲明中稱，該公司的技術符合歐盟關于個人數(shù)據(jù)收集的相關法律，并且允許讓用戶來決定其個人數(shù)據(jù)的分享方式。

亞馬遜公司的一位發(fā)言人表示，亞馬遜分享的只是“匿名化和匯總后的特性數(shù)據(jù)，以幫助汽車制造商改善客戶體驗。”亞馬遜不會向汽車廠商或開發(fā)者提供任何個人身份信息。

寶馬公司表示，該公司雖然會共享它收集的一些數(shù)據(jù)，但不會直接通過它賺錢。寶馬負責數(shù)字產(chǎn)品的高級副總裁迪特爾·梅表示：“假設某個人正在車上聽一首音樂，而我們發(fā)現(xiàn)最近恰好要開一場大型音樂會，我們可能就會把這個信息轉給銷售部門，使用戶有機會買到一張?zhí)貎r票。”

雖然政府和企業(yè)已經(jīng)開始著手解決車載語音識別系統(tǒng)的信息安全問題了，但目前尚不清楚誰將掌握這些收集到的數(shù)據(jù)的控制權。

市場研究機構Strategy Analytics的汽車業(yè)務副總監(jiān)羅杰·蘭托特認為：“語音識別是汽車行業(yè)的最新亮點，擁有最好的語音識別技術的廠商必將勝出。在這個贏者通吃的世界，贏家的回報將是相當奢侈的。”（財富中文網(wǎng)）

本文原載于2019年2月刊的《財富》雜志。

譯者：樸成奎

That means more drivers will have access to technology that alerts them to a sale at a nearby store—via an audio cue or an on-screen prompt—or ensures that their grande latte is ready when they swing by their local Starbucks.

But manufacturers, marketers, and other businesses may gain even more. Once artificial intelligence becomes better at understanding and interpreting speech across a range of dialects, accents, and colloquialisms, cars will be able to glean even deeper insight into drivers’ behavior and share that information with other companies.

While it may be convenient for cars to know your shopping habits and favorite places, that information can also be used to create targeted audio or visual ads, sell subscription services, or tip off third-party businesses. The more data that Amazon, Google, and Apple collect by placing their devices inside cars, the more opportunities they have to serve—or sell to—customers.

For now, automakers are gearing up to reap the rewards. Some manufacturers are forgoing technology from Amazon, Apple, and Google to install their own. Proprietary systems from Mercedes-Benz and BMW that are slated to debut this year will let manufacturers compete with the tech giants for driver data.

“In the coming years, the battleground will be over user data and the revenue made possible,” Kim says.

Although the devices are supposed to listen only when a “wake” phrase like “Hey, Alexa” is used, the process is not foolproof. Anecdotes from users worldwide range from amusing to horrifying. The most worrisome are incidents in which the devices accidentally record and send conversations to contacts in users’ phones, or inadvertently make large purchases. Amazon has begun taking measures to prevent unintended commands, including allowing customers to set a PIN for voice activated purchases and asking for confirmation before placing a call.

Currently automakers say they get cus?tomer permission before they use the individual data they collect for marketing or share it with third parties. Volvo said in a statement that its technology “tak ” and complies with a European Union law that lets residents control how t leheir personal data is shared.

An Amazon spokesman says that the company merely shares “anonymized, aggregated performance data to help automakers improve the customer experience” and that it doesn’t provide personally identifiable information to car companies or developers.

BMW shares the data it collects but says it doesn’t make money from it directly. “Let’s say the person is listening to certain music, and we know there’s a big concert,” says Dieter May, senior vice president of digital products for BMW. “Then we would probably give that to our salespeople to make an offer for a special ticket.”

But even as governments and corporations begin to address security questions, it’s unclear who will control the data that is collected.

“It’s the latest shiny thing in the car, and the one with the best speech recognition will win,” says Roger Lanctot, associate director for automotive at Strategy Analytics, a market research and consulting firm. “In our winner-take-all world, that win will be very lucrative.”

This article originally appeared in the February 2019 issue of Fortune.