中國挑戰(zhàn)美國人工智能主導(dǎo)地位

Artificial intelligence has had more than its fair share of ups and downs over the past 60 years, but one constant feature of the field has been the dominance of the US. Significant contributions to AI have certainly arisen elsewhere but, until recently, every AI system that made worldwide news originated in the US.

過去60年人工智能經(jīng)歷了不尋常的跌宕起伏，但該領(lǐng)域的一個不變特征是，美國一直占據(jù)主導(dǎo)地位。其他地方當(dāng)然也對人工智能做出了重大貢獻，但直到不久前，每個引起全球轟動的人工智能系統(tǒng)全都源于美國。

DeepBlue, which defeated chess grandmaster Garry Kasparov, was an IBM system, as was Watson, which defeated champion Jeopardy players in 2011. The robot Stanley, which demonstrated the feasibility of driverless cars in 2005, was developed at Stanford University in the heart of Silicon Valley. And if you dig a little deeper, the reasons for the dominance of the US become clear: Darpa, the US military research funding agency, is acknowledged in many of the key research papers in the AI canon.

擊敗國際象棋特級大師加里•卡斯帕羅夫(Garry Kasparov)的“深藍”(DeepBlue)是IBM研發(fā)的系統(tǒng)，在2011年《危險邊緣》(Jeopardy)節(jié)目中獲得冠軍的“沃森”(Watson)也是一樣。在2005年證明了無人駕駛汽車可行性的機器人斯坦利(Stanley)是居于硅谷中心的斯坦福大學(xué)(Stanford University)開發(fā)的。如果你再深究一下，美國占據(jù)主導(dǎo)地位的原因就變得很清楚了：關(guān)于人工智能標(biāo)準(zhǔn)的許多重要研究論文都鳴謝了美國軍事研究資助機構(gòu)——美國國防高級研究計劃局(DARPA)。

But today American hegemony in AI is being seriously challenged for the first time. One of the most remarkable features of the current AI boom is the sudden and very visible presence of China as a global force.

但如今美國在人工智能方面的霸權(quán)首次受到嚴(yán)峻挑戰(zhàn)。當(dāng)前人工智能熱潮的一個最顯著特點是，中國突然崛起為耀眼的全球人工智能強國。

One crude but nevertheless useful way of measuring national scientific muscle is to look at how a country performs in the leading scientific publication venues. Historically, one of the key scientific conferences in AI is the annual meeting of the Association for the Advancement of AI. This was first held in 1980 and, within a few years, the conference was receiving some 5,000 delegates. The 1980 conference was dominated by the US: there was not a single paper written by researchers from a Chinese institute, and only a modest European presence.

衡量一國科學(xué)力量的一種粗糙而又有用的方法是，觀察一國在重要科學(xué)出版界的表現(xiàn)。從歷史上看，人工智能領(lǐng)域的一個重要科學(xué)會議是人工智能促進協(xié)會(Association for the Advancement of AI)的年會。它在1980年首次舉行，幾年內(nèi)就有了約5000名代表參會。1980年的會議由美國主導(dǎo)：沒有一份中國機構(gòu)的研究人員撰寫的論文，歐洲也只有為數(shù)不多的論文。

This is not too surprising, of course: in its early days the conference was a US-centric event, and China was a very different place back then.

當(dāng)然，這并不太令人感到意外：會議在早期是以美國為中心的，而那時的中國與現(xiàn)在截然不同。

Go forward 18 years to the 1998 conference, and America still dominated, but with a substantial non-US presence, particularly from Europe. But there was just one paper from China — more specifically from Hong Kong, which had reverted to Chinese rule just one year before.

快進18年來到1998年的大會，美國仍然占據(jù)主導(dǎo)地位，但出現(xiàn)了許多美國以外地方的研究論文，尤其是來自歐洲的。但只有一篇論文來自中國——更具體地說來自一年前回歸中國的香港。

Today, the picture is startlingly different. At the 2018 conference, held in New Orleans in February, China submitted 25 per cent more papers than the US (1,242 to 934). More tellingly, it was just three papers behind the US in acceptances.

如今情況已經(jīng)截然不同。在今年2月份在新奧爾良舉行的2018年大會上，中國提交的論文數(shù)量比美國多出25%(前者是1242份，后者是934份)。更具說服力的是，收錄的中國論文數(shù)只比美國少了3篇。

It is hard to avoid the conclusion that China is now in serious competition with the US for dominance in AI. No European nation remotely competes at these volumes and, even taken as a whole, Europe is not really in competition for gold or silver.

人們很難不得出一個結(jié)論，即中國現(xiàn)在正處于與美國爭奪人工智能領(lǐng)域主導(dǎo)地位的激烈競爭之中。沒有一個歐洲國家能在這種規(guī)模上參與競爭，即使作為整體，歐洲實際上也不具備爭奪金牌或者銀牌的實力。

So why is China suddenly so prominent? In a word: scale. The machine-learning techniques behind the current AI boom are seriously data hungry. To recognise human faces, translate languages and pilot driverless cars requires huge quantities of “training data” — the fuel for machine learning algorithms that we generate every time we go online or use our smartphones.

那么為什么中國突然如此突出? 一句話：規(guī)模。當(dāng)前人工智能熱潮背后的機器學(xué)習(xí)技術(shù)對數(shù)據(jù)極其依賴。識別人臉、翻譯語言和試驗無人駕駛汽車需要大量的“訓(xùn)練數(shù)據(jù)”——這是我們每次上網(wǎng)或使用智能手機時產(chǎn)生的供機器學(xué)習(xí)算法使用的燃料。

With a population in a single market larger than the US and Europe combined, Chinese companies have a natural advantage in terms of access to data. While they might not be familiar to ordinary consumers in the west, Chinese tech companies like Tencent, Baidu, Alibaba and JD.com are global giants in terms of the number of users and market capitalisation. And they are all investing in AI on a dizzying scale. Ask a British teenager if they know WeChat, Tencent’s social media app, and you will be met by blank looks (I know because I’ve tried); but in China, the app has nearly a billion users.

由于中國的人口數(shù)量比美國和歐洲的總和還多，因此中國企業(yè)在獲取數(shù)據(jù)方面具有天然的優(yōu)勢。雖然西方普通消費者可能不熟悉騰訊(Tencent)、百度(Baidu)、阿里巴巴(Alibaba)和京東(JD.com)等中國科技公司，但從用戶數(shù)量和市值方面來說，這些公司都是全球巨頭。它們?nèi)家泽@人的規(guī)模大舉投資于人工智能。如果你問一個英國少年，他們是否知道騰訊的社交媒體應(yīng)用微信(WeChat)，他們會一臉茫然(我之所以知道是因為我這么問過)，但在中國，微信擁有近10億用戶。

One face of China’s AI revolution is Andrew Ng. British born to parents from Hong Kong, Mr Ng was director of the AI lab at Stanford, one of the great historical centres for US AI research. He made his name developing AI software to control helicopters, and won the Computers and Thought Award, the key research award for young AI scientists.

中國人工智能革命的代表人物之一是吳恩達(Andrew Ng)。吳恩達出生于英國，父母是香港人，他是斯坦福大學(xué)人工智能實驗室的主任，這是美國人工智能研究領(lǐng)域歷史悠久的幾大中心之一。他因開發(fā)人工智能軟件來控制直升機而聞名，并贏得了獎勵年輕人工智能科學(xué)家的重要研究獎——計算機與思想獎(Computers and Thought Award)。

He went on to work for Google, starting the Google brain project, and then became chief scientist for Baidu. Last year, he left Baidu to pursue other ventures. Brilliant, charismatic, and above all, remarkably energetic, Mr Ng has a flair for highly quotable soundbites. He recently tweeted: “Pretty much anything that a normal person can do in less than a second, we can now automate with AI.” I am not inclined to argue.

他后來去了谷歌(Google)，啟動了谷歌大腦項目，然后又成為百度的首席科學(xué)家。去年，他離開百度去追求其他事業(yè)。吳恩達聰明、富有魅力，而且最重要的是精力特別充沛。吳恩達有著能夠說出被人們廣泛引用的金句的天賦。他最近在Twitter上表示：“一個普通人可以在不到一秒的時間內(nèi)做的幾乎所有事情，現(xiàn)在我們差不多都可以使用人工智能自動化。”我不想爭論。

In 2017, he declared that AI is the “new electricity”, and that “just as electricity transformed many industries roughly 100 years ago, AI will also now change nearly every major industry”. If that is the case, it is very likely that China will be the generator that powers AI in the decades ahead.

2017年，他宣稱人工智能是“新電力”，而且“就像電力在大約100年前改變了許多行業(yè)一樣，人工智能現(xiàn)在也將會改變幾乎所有主要行業(yè)”。如果真是這樣的話，中國很有可能會成為未來幾十年人工智能的發(fā)電機。

The writer is professor of computer science at the University of Oxford and author of ‘Artificial Intelligence: A Ladybird Expert Book’

本文作者是牛津大學(xué)(University of Oxford)計算機科學(xué)教授，并著有《人工智能：瓢蟲專家手冊》(Artificial Intelligence: A Ladybird Expert Book)一書