Should Internet Firms Pay for the Data Users Currently Give Away?網絡公司應為用戶產出的數據付費嗎？

劉莉

And， as a new paper proposes， should the data-providers unionise？ 還有一篇新論文提議，數據提供者是否應當成立工會？

You have multiple jobs， whether you know it or not. Most begin first thing in the morning， when you pick up your phone and begin generating the data that make up Silicon Valley’s most important resource. That， at least， is how we ought to think about the role of data-creation in the economy， according to a fascinating new economics paper. We are all digital labourers， helping make possible the fortunes generated by firms like Google and Facebook， the authors argue. If the economy is to function properly in the future—and if a crisis of technological unemployment is to be avoided—we must take account of this， and change the relationship between big internet companies and their users.

Artificial intelligence （AI） is getting better all the time， and stands poised to transform a host of industries， say the authors （Imanol Arrieta Ibarra and Diego Jiménez Hernández， of Stanford University， Leonard Goff， of Columbia University， and Jaron Lanier and Glen Weyl， of Microsoft）. But， in order to learn to drive a car or recognise a face， the algorithms that make clever machines tick must usually be trained on massive amounts of data. Internet firms gather these data from users every time they click on a Google search result， say， or issue a command to Alexa. They also hoover up valuable data from users through the use of tools like reCAPTCHA， which ask visitors to solve problems that are easy for humans but hard for AIs， such as deciphering text from books that machines are unable to parse. That does not just screen out malicious bots， but also helps digitise books. People “pay” for useful free services by providing firms with the data they crave.

These data become part of the firms’ capital， and， as such， a fearsome source of competitive advantage. Would-be startups that might challenge internet giants cannot train their AIs without access to the data only those giants possess. Their best hope is often to be acquired by those very same titans， adding to the problem of uncompetitive markets.

That， for now， AI’s contributions to productivity growth are small， the authors say， is partly because of the free-data model， which limits the quality of data gathered. Firms trying to develop useful applications for AI must hope that the data they have are sufficient， or come up with ways to coax users into providing them with better information at no cost. For example， they must pester random people—like those blur-deciphering visitors to websites—into labelling data， and hope that in their annoyance and haste they do not make mistakes.

Even so， as AI improves， the amount of work made vulnerable to displacement by technology grows， and ever more of the value generated in the economy accrues to profitable firms rather than workers. As the authors point out， the share of GDP paid out to workers in wages and salaries—once thought to be relatively stable—has already been declining over the past few decades.

To tackle these problems， they have a radical proposal. Rather than being regarded as capital， data should be treated as labour—and， more specifically， regarded as the property of those who generate such information， unless they agree to provide it to firms in exchange for payment. In such a world， user data might be sold multiple times， to multiple firms， reducing the extent to which data sets serve as barriers to entry. Payments to users for their data would help spread the wealth generated by AI. Firms could also potentially generate better data by paying. Rather than guess what a person is up to as they wander around a shopping centre， for example， firms could ask individuals to share information on which shops were visited and which items were viewed， in exchange for payment. Perhaps most ambitiously， the authors muse that data labour could come to be seen as useful work， conferring the same sort of dignity as paid employment： a desirable side-effect in a possible future of mass automation.

The authors’ ideas need fleshing out; their paper， thought-provoking though it is， runs to only five pages. Parts of the envisioned scheme seem impractical. Would people really be interested in taking the time to describe their morning routine or office habits without a substantial monetary inducement （and would their data be valuable enough for firms to pay a substantial amount）？ Might not such systems attract data mercenaries， spamming firms with useless junk data simply to make a quick buck？

Nothing to use but your brains

Still， the paper contains essential insights which should frame discussion of data’s role in the economy. One concerns the imbalance of power in the market for data. That stems partly from concentration among big internet firms. But it is also because， though data may be extremely valuable in aggregate， an individual’s personal data typically are not. For one Facebook user to threaten to deprive Facebook of his data is no threat at all. So effective negotiation with internet firms might require collective action： and the formation， perhaps， of a “data-labour union”.

This might have drawbacks. A union might demand too much in compensation for data， for example， impairing the development of useful AIs. It might make all user data freely available and extract compensation by demanding a share of firms’ profits; that would rule out the pay-for-data labour model the authors see as vital to improving data quality. Still， a data union holds potential as a way of solidifying worker power at a time when conventional unions struggle to remain relevant.

Most important， the authors’ proposal puts front and centre the collective nature of value in an AI world. Each person becomes something like an oil well， pumping out the fuel that makes the digital economy run. Both fairness and efficiency demand that the distribution of income generated by that fuel should be shared more evenly， according to our contributions. The tricky part is working out how.

不論你知道與否，其實你正身兼數職。大多數人早晨就開工了——你拿起手機開始產生數據，構成了硅谷最重要的信息來源。一篇引人入勝的經濟學新論文提出，我們至少應當從這個角度去思考數據創造在經濟學當中的角色。作者們認為，我們所有人都是數字勞工，為谷歌、臉書之類的公司制造財富。要想讓未來的經濟正常運轉，要想避免技術帶來的失業危機，我們就必須考慮到這一點，改變大型互聯網公司與其用戶的關系。

人工智能（AI）日新月異，時刻準備著讓一系列行業轉型換代，論文的作者們（來自斯坦福大學的伊馬諾爾·阿列塔·伊瓦拉與迭戈·希門尼斯·埃爾南德斯，來自哥倫比亞大學的倫納德·戈夫，來自微軟公司的雅龍·拉尼爾與格倫·韋爾）表示。不過，為了學習汽車駕駛和人臉識別，智慧機器所用的算法通常需要先在海量數據中訓練運行。互聯網公司的數據，來源于用戶對谷歌搜索的每一次點擊、對亞馬遜語音助手Alexa發出的每一條指令。他們還會使用reCAPTCHA之類的工具，從用戶身上抓取有價值的數據——該工具要求訪客去解決對人類很容易但AI卻難以勝任的問題，例如對書中的文本進行句法分析。這樣做不僅能篩除惡意自動程序，還能將紙本圖書電子化。人們向互聯網公司提供他們渴求的數據，從而為免費又好用的服務“買單”。

這些數據不但成為了互聯網公司的資本，更可以帶來驚人的競爭優勢。躍躍欲試的創業公司也許會向互聯網巨頭發起挑戰，但卻必須借助巨頭手中的數據才能訓練自家AI。他們最好的結局往往是被巨頭同行收購，讓競爭本就不夠充分的市場雪上加霜。

論文作者們認為，目前AI對生產力增長的貢獻不大，部分原因在于免費數據模式限制了數據采集的質量。若要開發實用的AI應用，互聯網公司必須寄希望于充足的數據，或者想辦法誘導用戶無償向其提供更優質的信息。例如，他們必須纏著隨機人群去給數據貼標簽，比如那些要識別模糊驗證碼的訪客，而且還要希望他們在煩擾和匆忙中不出錯。

即便如此，隨著AI的改進，越來越多的工作會因技術進步而被取代，所產生的經濟價值也會更多地落入贏利公司而非工人手中。作者們指出，薪水支出所占的GDP份額曾被認為是相對穩定的，但過去幾十年間卻每況愈下。

為了應對這些問題，他們提出了一種激進的方案。數據不應該被當作資本看待，而應當作為勞動成果——具體來講就是信息產生者的財產，除非他們同意向公司提供數據以換取報酬。如此一來，用戶數據可能會多次兜售給多家公司，從而降低數據作為準入門檻的高度。向提供數據的用戶支付報酬，有利于將AI制造的財富分配開來，也讓互聯網公司有望獲得更好的數據。舉個例子，與其猜測商場里的顧客想要什么，不如請求人們分享自己的信息以換取報酬，告訴互聯網公司他們到訪了什么店鋪、瀏覽了哪些物品。那些論文作者們最大膽的想法也許是，數據勞動可能會漸漸被視作一項有用的工作，像帶薪職務一樣賦予人們尊嚴——未來興許會出現的大規模自動化便帶有這種令人期待的副作用。

這些作者的想法雖然發人深省，但只有區區五頁篇幅，還需詳加闡述。他們設想的這個體系里，有些部分似乎不切實際。如果沒有可觀的酬金，人們是否真有興趣花時間描述自己每天早上的起居或辦公室里的習慣（他們的數據又是否真那么寶貴，值得互聯網公司大掏腰包）？這些體系會不會引來一眾數據雇傭兵，為了掙快錢而拿沒用的垃圾數據敷衍交差？

除了大腦別無可用

當然，這篇論文仍然具有一些重要洞見，給探討數據在經濟活動中扮演的角色擬訂了框架。其中一個角色，便牽涉數據市場中權力的失衡。大型互聯網公司的集中性是一方面，還有一個原因則是，盡管數據總體的價值極高，個體提供的單一數據一般卻無足輕重。就算某位用戶拒不提供他的數據，也不會對臉書構成任何威脅。因此，要與互聯網公司進行有效磋商，可能需要采取集體行動：也許，還需要成立一個“數據工會”。

這樣做也許有其弊端。比如說，工會也許會開出過高的數據價碼，令實用AI的開發受阻。工會也許會要求互聯網公司以利潤分成來換取免費使用所有數據的權利。這就與論文作者們主張的數據付費勞動模型背道而馳了。他們認為該模型對提高數據質量至關重要。不過，在傳統工會慘淡經營之際，數據工會作為鞏固工人權力的的一種方式還是有前景的。

最重要的是，作者們的提議將AI世界中價值的集體性本質放到了聚光燈下。每個人變成了像油井一樣的東西，從中可抽出數字經濟賴以運行的燃料來。不論是出于公平還是效率的要求，那種燃料產生的收入都應當按勞分配。至于如何實現，則是難點所在。