不能小看機器人作家

LET me hazard a guess that you think a real person has written what you’re reading. Maybe you’re right. Maybe not. Perhaps you should ask me to confirm it the way your computer does when it demands that you type those letters and numbers crammed like abstract art into that annoying little box.

讓我來猜猜看，你認爲你所閱讀的內容是由一個真實存在的人寫的。你可能是對的，也可能是錯的。或許你應該讓我確認這種說法，就像你的電腦要求你將抽象藝術般的字母和數字輸入那個令人厭煩的小盒子一樣。

Because, these days, a shocking amount of what we’re reading is created not by humans, but by computer algorithms. We probably should have suspected that the information assaulting us 24/7 couldn’t all have been created by people bent over their laptops.

因爲，目前有相當多的閱讀內容不是由人類編寫的，而是由計算機算法完成的。我們可能應該會猜想，每天24小時向我們襲來的信息可能不完全是由人類俯在筆記本電腦前編寫的。

It’s understandable. The multitude of digital avenues now available to us demand content with an appetite that human effort can no longer satisfy. This demand, paired with ever more sophisticated technology, is spawning an industry of “automated narrative generation.”

這是可以理解的。人類的努力已經無法滿足我們現在能夠使用的各種數字渠道對內容的需求。這種需求，再加上更加成熟的技術，滋生了一個“文本自動生成”產業。

Companies in this business aim to relieve humans from the burden of the writing process by using algorithms and natural language generators to create written content. Feed their platforms some data — financial earnings statistics, let’s say — and poof! In seconds, out comes a narrative that tells whatever story needs to be told.

該領域中的公司旨在利用算法和自然語言生成器編寫內容，使人類擺脫寫作過程中的負擔。將一些數據——比如金融收益數據——輸入它們的平臺，然後“嗖”的一聲！幾秒鐘之內就會產生一些內容，提供人們需要的各種報道。

These robo-writers don’t just regurgitate data, either; they create human-sounding stories in whatever voice — from staid to sassy — befits the intended audience. Or different audiences. They’re that smart. And when you read the output, you’d never guess the writer doesn’t have a heartbeat.

這些機器人寫手並不只是重複數據；它們以適合目標受衆的風格——從古板到活潑——寫出看起來像是人類編寫的報道。它們非常聰明。當你閱讀這些報道時，你絕不會猜到這個作者沒有心跳。

Consider the opening sentences of these two sports pieces:

看看這兩篇體育報道的開篇語句。

“Things looked bleak for the Angels when they trailed by two runs in the ninth inning, but Los Angeles recovered thanks to a key single from Vladimir Guerrero to pull out a 7-6 victory over the Boston Red Sox at Fenway Park on Sunday.”

“週日，天使隊(Angels)在第九局中落後兩分時，情況看起來不妙，但憑藉弗拉迪米爾·葛雷諾(Vladimir Guerrero)贏得的關鍵一分，洛杉磯天使隊挽回敗局，在芬威球場(Fenway Park)以七比六的比分擊敗波士頓紅襪隊(Boston Red Sox)。”

“The University of Michigan baseball team used a four-run fifth inning to salvage the final game in its three-game weekend series with Iowa, winning 7-5 on Saturday afternoon (April 24) at the Wilpon Baseball Complex, home of historic Ray Fisher Stadium.”

“週六下午（4月24日），密歇根大學(University of Michigan)棒球隊在威爾彭棒球場(Wilpon Baseball Complex)——具有歷史意義的雷·費舍爾體育場(Ray Fisher Stadium)的所在地，通過贏得四分的第五局比賽，扭轉局勢，最終以七比五的比分贏得了與愛荷華棒球隊在週末舉行的三場比賽中的最後一場。”

If you can’t tell which was written by a human, you’re not alone. According to a study conducted by Christer Clerwall of Karlstad University in Sweden and published in Journalism Practice, when presented with sports stories not unlike these, study respondents couldn’t tell the difference. (Machine first, human second, in our example, by the way.)

如果你無法分辨哪一篇是由人類寫的，那你不是唯一一個。瑞典卡爾斯塔得大學(Karlstad University)的克里斯特·克萊瓦爾(Christer Clerwall)開展了一項研究，並在《新聞實踐》(Journalism Practice)上發表了相關論文。研究顯示，當看到類似的體育報道時，調查對象無法辨別其中的區別。（順便說一下，在我們提供的例子中，第一篇是機器寫的，第二篇是人寫的。）

Algorithms and natural language generators have been around for a while, but they’re getting better and faster as the demand for them spurs investment and innovation. The sheer volume and complexity of the Big Data we generate, too much for mere mortals to tackle, calls for artificial rather than human intelligence to derive meaning from it all.

算法和自然語言生成器已經存在了一段時間，但隨着對它們的需求刺激了投資和創新，它們變得越來越好，越來越快。我們產生海量的大數據(Big Data)，而且很複雜，凡人難以處理，需要人工智能，而不是人類智能，來從中獲取有意的信息。

Set loose on the mother lode — especially stats-rich domains like finance, sports and merchandising — the new software platforms apply advanced metrics to identify patterns, trends and data anomalies. They then rapidly craft the explanatory narrative, stepping in as robo-journalists to replace humans.

將之應用於大量資源，特別是在金融、體育和銷售規劃等數據繁多的領域，這種新的軟件平臺就會應用先進的度量標準，去確認模式、趨勢和反常數據。然後，它們會迅速產生解釋性文本，成爲代替人類的機器人記者。

The Associated Press uses Automated Insights’ Wordsmith platform to create more than 3,000 financial reports per quarter. It published a story on Apple’s latest record-busting earnings within minutes of their release. Forbes uses Narrative Science’s Quill platform for similar efforts and refers to the firm as a partner.

美聯社(The Associated Press)每季度利用自動化洞察力公司(Automated Insights)的Wordsmith平臺撰寫3000多篇金融報道。他們在蘋果(Apple)公司公佈最新創紀錄收益幾分鐘之後，就發表了一篇報道。福布斯(Forbes)利用敘述科學公司(Narrative Science)的Quill平臺撰寫類似報道，並稱該公司是他們的合作伙伴。

Then we have Quakebot, the algorithm The Los Angeles Times uses to analyze geological data. It was the “author” of the first news report of the 4.7 magnitude earthquake that hit Southern California last year, published on the newspaper’s website just moments after the event. The newspaper also uses algorithms to enhance its homicide reporting.

然後又出現了Quakebot，《洛杉磯時報》(The Los Angeles Times)利用這種算法分析地質數據。它是第一篇有關南加利福尼亞州去年發生的4.7級地震的新聞報道的“作者”。地震發生後，該報立即在其網站了發表了這篇報道。該報還利用算法加強命案報道。

But we should be forgiven a sense of unease. These software processes, which are, after all, a black box to us, might skew to some predicated norm, or contain biases that we can’t possibly discern. Not to mention that we may be missing out on the insights a curious and fertile human mind could impart when considering the same information.

如果我們對此感到一絲不安，這也是可以理解的。這些軟件程序畢竟對我們來說是一個黑盒子，它們可能偏向於一些特定的基準，或包含我們可能無法辨別的傾向性。更不用說，我們可能會錯失一個好奇的、具有創造力的人類在思考相同的信息時所能產生的那種洞見。

The mantra around all of this carries the usual liberation theme: Robo-journalism will free humans to do more reporting and less data processing.

這一切所表達的呼聲，包含着常見的解放主題——機器新聞將會解放人類，使人類能夠更多地進行報道，減少數據處理工作。

That would be nice, but Kristian Hammond, Narrative Science’s co-founder, estimates that 90 percent of news could be algorithmically generated by the mid-2020s, much of it without human intervention. If this projection is anywhere near accurate, we’re on a slippery slope.

這不失爲一件美事。但是，據敘述科學聯合創始人克里斯蒂安·哈蒙德(Kristian Hammond)估計，到本世紀20年代中期，將有90%的新聞由計算機算法生成，其中大多都無需人工干預。倘若這個預測接近事實，那麼我們就會處在一個滑坡之上。

It’s mainly robo-journalism now, but it doesn’t stop there. As software stealthily replaces us as communicators, algorithmic content is rapidly permeating the nooks and crannies of our culture, from government affairs to fantasy football to reviews of your next pair of shoes.

目前，機器新聞已經佔據主導，但它並未就此止步。隨着軟件悄悄取代我們成爲傳播者，從政府事務到夢幻足球，再到對你下一雙鞋子的評價，算法生成的內容也在迅速向我們文化中的各個角落和縫隙滲透。

Automated Insights states that its software created one billion stories last year, many with no human intervention; its home page, as well as Narrative Science’s, displays logos of customers all of us would recognize: Samsung, Comcast, The A.P., and Yahoo. What are the chances that you haven’t consumed such content without realizing it?

自動化洞察力公司指出，其軟件去年一共創作了10億個報道，許多都沒有人工干預；它和敘述科學公司的主頁上，展示着我們耳熟能詳的客戶標誌：三星(Samsung)、康卡斯特(Comcast)、美聯社、和雅虎(Yahoo)。所以你極有可能在沒有意識的情況下消費了這種內容。

Books are robo-written, too. Consider the works of Philip M. Parker, a management science professor at the French business school Insead: His patented algorithmic system has generated more than a million books, more than 100,000 of which are available on Amazon. Give him a technical or arcane subject and his system will mine data and write a book or report, mimicking the thought process, he says, of a person who might write on the topic. Et voilà, “The Official Patient’s Sourcebook on Acne Rosacea.”

機器人還在寫書。來看看法國的歐洲工商管理學院(Insead)管理科學教授菲利普·M·帕克(Philip M. Parker)的作品：他的專利算法系統已經生成了超過100萬本圖書，其中有10萬多本在亞馬遜上銷售。他說，給他一個技術性或晦澀難懂的話題，他的系統就能模仿可能就此題目進行寫作的人的思維過程，挖掘數據，撰寫一本書或一篇報告。比如，《紅斑痤瘡患者官方資料》(The Official Patient’s Sourcebook on Acne Rosacea)。

Narrative Science claims it can create “a narrative that is indistinguishable from a human-written one,” and Automated Insights says it specializes in writing “just like a human would,” but that’s precisely what gives me pause. The phrase is becoming a de facto parenthetical — not just for content creation, but where most technology is concerned.

敘述科學聲稱它可以創作“與出自人類的作品分毫不差的文本”。自動化洞察力則稱它的專長是“像一個人一樣”寫作，但這正是讓我擔憂的地方。這種說法事實上已經成爲一段插入語——不只是對內容創作，而且對於大多數科技都是如此。

Our phones can speak to us (just as a human would). Our home appliances can take commands (just as a human would). Our cars will be able to drive themselves (just as a human would). What does “human” even mean?

我們的手機可以（像一個人一樣）和我們說話。我們的家用電器能夠（像一個人一樣）接受指令。我們的汽車將能（像一個人一樣）自行駕駛。那麼，“人”究竟是什麼意思？

With technology, the next evolutionary step always seems logical. That’s the danger. As it seduces us again and again, we relinquish a little part of ourselves. We rarely step back to reflect on whether, ultimately, we’re giving up more than we’re getting.

在科技的幫助下，下一個革命性的進展似乎總顯得順理成章。這就是危險所在。鑑於它反覆引誘我們，我們就會放棄一小部分自己。我們很少會後退一步，反思我們最後放棄的東西是否比得到的更多。

Then again, who has time to think about that when there’s so much information to absorb every day? After all, we’re only human.

再者，當每天都有這麼多信息需要吸收的時候，誰還有時間去思考這那個問題？畢竟，我們只是人類。

Related: Interactive Quiz: Did a Human or a Computer Write This? A shocking amount of what we’re reading is created not by humans, but by computer algorithms. Can you tell the difference? Take the quiz.

相關內容：互動問答：這是人還是計算機寫的？現在我們讀到的內容中，由計算機算法而非人類編寫的比例相當之高。你能區分嗎？來試試。