中國(guó)科技巨擘競(jìng)逐人工智能

As the quest to master artificial intelligence intensifies, China’s tech trio of Baidu, Alibaba and Tencent have a distinct advantage over their Silicon Valley rivals — data.

在各方對(duì)人工智能(AI)的探索日益白熱化之際，中國(guó)科技三巨頭百度(Baidu)、阿里巴巴(Alibaba)和騰訊(Tencent)比硅谷競(jìng)爭(zhēng)對(duì)手擁有一項(xiàng)明顯優(yōu)勢(shì)——數(shù)據(jù)。

As Robin Li, chairman and chief executive of Baidu, says: “Baidu knows you better than you know yourself.”

正如百度董事長(zhǎng)兼首席執(zhí)行官李彥宏(Robin Li)所說(shuō)的那樣：“百度(比你自己)更懂你。”

Baidu, Alibaba and Tencent, have embraced AI with alacrity: setting up specialist labs at home and overseas and hiring top engineers. Much like their US peers such as Google, they are using machine learning to push into newer fields of autonomous driving, medical diagnosis, facial recognition for payments and AI-enabled hardware that can be operated by voice.

百度、阿里巴巴和騰訊均已欣然接受AI：在國(guó)內(nèi)和海外建立專(zhuān)家實(shí)驗(yàn)室，聘請(qǐng)頂級(jí)工程師。和谷歌(Google)等美國(guó)同行一樣，他們利用機(jī)器學(xué)習(xí)進(jìn)入了一些較新的領(lǐng)域，包括自主駕駛、醫(yī)療診斷、用于支付驗(yàn)證的面部識(shí)別，以及可用語(yǔ)音操作的AI功能硬件。

“We regarded AI as one of our top priorities last year and treat it as a strategic focus,” says Mark Ren, Tencent’s chief operating officer last month.

“去年我們把AI視為我們的優(yōu)先項(xiàng)目之一，把它當(dāng)作一個(gè)戰(zhàn)略重點(diǎn)，”騰訊首席運(yùn)營(yíng)官任宇昕(Mark Ren)上月表示。

Alibaba takes a similar view, saying AI permeates “every aspect of our business . . . we don’t see it as a separate business unit.”

阿里巴巴持有類(lèi)似觀(guān)點(diǎn)，稱(chēng)AI滲透進(jìn)了“我們業(yè)務(wù)的方方面面……我們不把它看作獨(dú)立的業(yè)務(wù)部門(mén)。”

The difference now, says Xiaofeng Wang, senior analyst at Forrester, the consultants, is the scale of the data companies have and the speed with which machines are able to process it. “You can easily do personalisation for a group of customers, but if its millions it’s very hard, and it’s about how fast you can provide that,” she says.

咨詢(xún)機(jī)構(gòu)Forrester高級(jí)分析師王小峰(音)表示，如今的差別在于企業(yè)所掌握的數(shù)據(jù)規(guī)模以及機(jī)器處理數(shù)據(jù)的速度。“你可以輕松地對(duì)一群顧客進(jìn)行個(gè)性化分析，但如果要分析數(shù)百萬(wàn)顧客就會(huì)非常難，這里的關(guān)鍵在于你可能多快得出結(jié)果，”他表示。

Alibaba began highlighting its AI-powered improvements in this year’s earnings report. Jeff Zhang, chief technology officer, told analysts earlier this year that the biggest change was the scope of data — which now tops 1,000 PBs, equivalent to about 580bn books — it is able to process.

阿里巴巴在今年的盈利報(bào)表中開(kāi)始介紹AI推動(dòng)的改進(jìn)。首席技術(shù)官?gòu)埥ㄤh(Jeff Zhang)今年早些時(shí)候向分析師們表示，最大的變化是該公司如今能夠處理的數(shù)據(jù)規(guī)模(1000PB)，相當(dāng)于大約5800億冊(cè)書(shū)。

“We are generating and collecting all different kinds of data,” he says. “We want to leverage this to enable personalisation, to power search, security, customer service in all of these areas we are providing support through our data products.”

“我們產(chǎn)生并收集所有不同類(lèi)型的數(shù)據(jù)，”他稱(chēng)，“我們希望利用這些數(shù)據(jù)來(lái)實(shí)現(xiàn)個(gè)性化，通過(guò)我們的數(shù)據(jù)產(chǎn)品在我們提供支持的所有領(lǐng)域推動(dòng)搜索、安全、客戶(hù)服務(wù)。”

Some of that was in evidence on Singles Day, when Alibaba used AI to generate 400m customised banner advertisements in the month leading up to the shopping day. It also used chatbots, to answer 3.5m simple queries a day over the presale period, such as “Where’s my package?”

11月11日的“光棍節(jié)”展現(xiàn)了AI的一部分威力，阿里巴巴在這個(gè)購(gòu)物節(jié)之前的一個(gè)月利用AI制作了4億個(gè)定制橫幅廣告。在預(yù)售期，該公司還利用聊天機(jī)器人每天回答350萬(wàn)個(gè)簡(jiǎn)單詢(xún)問(wèn)，比如“我的包裹在哪?”

All of this was possible previously, through the use of microphones and cameras, but not on the massive scale and at the split-second speeds afforded by AI technologies.

這一切在過(guò)去也可以通過(guò)麥克風(fēng)和攝像頭實(shí)現(xiàn)，但是達(dá)不到如此龐大的規(guī)模以及AI技術(shù)帶來(lái)的高速度。

Edouard de Mezerac, head of data and analytics for Asia-Pacific at Oliver Wyman, the consultancy, highlights another change under way on Taobao, Alibaba’s third-party ecommerce platform. It allows users to search by image using deep learning, a technique designed to emulate the way a human brain works, to find a matching or similar item. This enables shoppers, for example, to source a dress worn by a celebrity. “That does not exist on Amazon today,” he says.

奧緯咨詢(xún)(Oliver Wyman)亞太區(qū)數(shù)據(jù)和分析主管愛(ài)德華•德梅澤拉克(Edouard de Mezerac)介紹了阿里巴巴的第三方電商平臺(tái)淘寶網(wǎng)(Taobao)正在經(jīng)歷的另一個(gè)變化。它允許用戶(hù)通過(guò)圖片進(jìn)行搜索，利用深度學(xué)習(xí)(一種旨在模擬人類(lèi)大腦運(yùn)作方式的技術(shù))找到相同或類(lèi)似的商品。比如，這可以讓購(gòu)物者搜索某個(gè)名人穿的裙子。“現(xiàn)在這種技術(shù)在亞馬遜(Amazon)的平臺(tái)上還不存在，”他表示。

It is still early days for the industry but applications such as these hint at how the country could take the global lead, say analysts, especially as companies ramp up how they use the technology.

分析師們表示，該行業(yè)在利用AI方面仍處于起步階段，但是這類(lèi)應(yīng)用暗示了中國(guó)如何可能居于全球領(lǐng)先地位，特別是隨著企業(yè)加快利用相關(guān)技術(shù)。

Anand Swaminathan, a senior partner at consultants McKinsey, highlights two further factors that set China apart from Silicon Valley. “Their consumer testing ground is bigger than anywhere in the world,” he says. “Here they are testing with 1bn-plus people so for that reason the US is automatically handicapped. And the pace and scale of investment is fundamentally higher.”

咨詢(xún)公司麥肯錫(McKinsey)高級(jí)合伙人阿南德•斯瓦米納坦(Anand Swaminathan)列舉了讓中國(guó)跑贏(yíng)硅谷的另外兩個(gè)因素。“他們的消費(fèi)者測(cè)試場(chǎng)比世界上其他任何地方都大，”他表示，“他們?cè)谶@里能夠借助10億以上人口進(jìn)行測(cè)試，因此美國(guó)在這方面先天不足。而且中國(guó)的投資速度和規(guī)模從根本上高于美國(guó)。”

The Chinese also generate far more data that is far more accessible. While Apple and Google have sought to fortify user data against governments, most notably last year when it thwarted the FBI’s attempt to unlock the San Bernardino shooter’s iPhone. Chinese companies are less coy.

中國(guó)還產(chǎn)生在規(guī)模和易于獲取程度上遠(yuǎn)超其他地區(qū)的數(shù)據(jù)。蘋(píng)果(Apple)和谷歌都試圖頂住政府的壓力來(lái)保護(hù)用戶(hù)數(shù)據(jù)，最突出的例子是去年蘋(píng)果拒絕幫助美國(guó)聯(lián)邦調(diào)查局(FBI)解鎖圣貝納迪諾(San Bernardino)槍擊者的iPhone。中國(guó)企業(yè)沒(méi)那么不配合。

Delegates at this month’s Fortune Global Conference in Guangzhou enjoyed unusual access to Google, Facebook and Twitter, services that are normally blocked in China. “I spoke to a number of folks here who said, ‘How nice of them to provide free WiFi so I can access everything’,” says Mr Swaminathan. “But that means [the companies] can access everything of yours.”

本月在廣州參加《財(cái)富》全球論壇的代表們，不同尋常地得以訪(fǎng)問(wèn)谷歌、Facebook和Twitter這些正常情況下在中國(guó)被屏蔽的服務(wù)。“我在這里聽(tīng)到很多人說(shuō)，‘他們提供免費(fèi)WiFi多好啊，這樣我能訪(fǎng)問(wèn)一切服務(wù)’，”斯瓦米納坦表示，“但是這意味著(這些公司)可以獲得你所有的信息。”

A hint of what could be coming from China’s tech giants is how they might use AI in communications and healthcare.

至于中國(guó)科技巨頭未來(lái)可能推出什么新穎產(chǎn)品和服務(wù)，從他們?cè)谕ㄐ藕歪t(yī)療行業(yè)可能如何利用AI可見(jiàn)一斑。

Alibaba is taking the voice-activated personal assistant concept a step further than Apple’s Siri or Amazon’s Alexa, with a plan to deploy ticket dispensing machines in subway stations that can cancel out background noise and focus purely on the person ordering the tickets.

阿里巴巴目前采用的語(yǔ)音激活個(gè)人助手概念比蘋(píng)果的Siri和亞馬遜的Alexa領(lǐng)先一步，該公司計(jì)劃在地鐵站推出可以過(guò)濾掉背景雜音、只聽(tīng)訂票者說(shuō)話(huà)的售票機(jī)。

The technology itself — essentially combining microphones and cameras to identify the speaker — has “a long history in research”, says Zhijie Yan, who heads up the speech interaction research at Alibaba.

阿里巴巴語(yǔ)音交互研究主管鄢志杰表示，該技術(shù)本身——本質(zhì)上是結(jié)合麥克風(fēng)和攝像頭來(lái)識(shí)別說(shuō)話(huà)人——“的研發(fā)歷史很長(zhǎng)”。

What is new, he says, is its application into a real product, helped by improving technology, which means personal assistants can move to public arenas: airports, train stations, restaurants and reception desks.

他表示，新穎之處在于，在不斷改進(jìn)的技術(shù)的幫助下，把它應(yīng)用于實(shí)際產(chǎn)品，這意味著個(gè)人助手可能進(jìn)入公共場(chǎng)所：機(jī)場(chǎng)、火車(chē)站、餐廳和前臺(tái)。

Smarter AI and bigger data caches apply to medical applications too. Tencent is among those looking to corral the technology into detecting lung cancer at an early stage.

智能程度更高的AI和更龐大的數(shù)據(jù)規(guī)模也適用于醫(yī)療應(yīng)用場(chǎng)合。騰訊是打算把該技術(shù)應(yīng)用于檢測(cè)早期肺癌的公司之一。

Helping on that score are cross-border collaborations, led by university researchers, such as a project enabling researchers in Hong Kong to access American databases so computers can learn to recognise the features of malignant nodules.

推動(dòng)這方面進(jìn)展的是由高校研究員牽頭的跨境合作，比如某個(gè)項(xiàng)目讓香港的研究員能夠訪(fǎng)問(wèn)美國(guó)數(shù)據(jù)庫(kù)，以便讓計(jì)算機(jī)可以學(xué)習(xí)惡性結(jié)節(jié)的特征。

David Lam, assistant professor and specialist in respiratory medicine at the University of Hong Kong, says it will be five years before early results from projects materialise.

香港大學(xué)(University of Hong Kong)助理教授、呼吸內(nèi)科專(zhuān)家大衛(wèi)•林(David Lam)表示，項(xiàng)目得出初步結(jié)果將需要5年時(shí)間。

One reason is that data sharing is complicated by the fact different ethnicities have different characteristics. Lung cancer affords one example: in the US, UK and much of the world, the main cause is smoking. That is not the case in Hong Kong, Taiwan and South China, where there is also far more prevalence among women.

其中一個(gè)原因是，由于不同民族有不同的特征，數(shù)據(jù)共享比較復(fù)雜。肺癌就是這方面的一個(gè)例子：在美國(guó)、英國(guó)以及世界大部分地區(qū)，肺癌的主要病因是吸煙。但在香港、臺(tái)灣以及中國(guó)南方不是這樣，這些地區(qū)女性的發(fā)病率高得多。

Other early stage efforts that have come in for criticism over concerns that they have little real-world application include programmes that beat world champions of chess and Go.

一些早期努力因?yàn)槿狈?shí)用性而招致批評(píng)，包括打敗圍棋和國(guó)際象棋世界冠軍的程序。

But Si Chen, senior strategy manager at Tencent’s AI lab, says the requisite perception and decision-making learning to beat those games will resonate.

但是騰訊AI實(shí)驗(yàn)室的高級(jí)戰(zhàn)略經(jīng)理陳思(音)表示，贏(yíng)得這些游戲所必需的感知和決策學(xué)習(xí)將具有借鑒價(jià)值。

“In the real world that is the same as self-driving,” she says. “Based on your sensors or what you can see, you have an understanding of the environment and from that be able to make decisions and take an action, which could be to turn the steering wheel left or brake.”

“在現(xiàn)實(shí)世界中，這就跟自動(dòng)駕駛一樣，”她表示，“基于你的傳感器或者你能看見(jiàn)的情況，你對(duì)環(huán)境有一個(gè)了解，據(jù)此可以作決定和采取行動(dòng)，比如向左打方向盤(pán)或是剎車(chē)。”