中国人脸识别背后的“数据标记工厂”

文章导读

本文选自《经济学人》1月4日刊文章。中国是人脸识别的国度，在人脸识别领域，中国已远远走在世界前列。这并不是因为中国有技术优势，而是因为中国有庞大的廉价劳动力和数据源。这些人脸识别是基于人工智能系统对庞大的人脸照片的自我学习，但这些输入系统的照片必须先进行标记，系统才能知道人脸的五官特征，并最终进行自我学习。这就催生了“数据标记工厂”，即人工对图片特征进行标记，然后投喂人工智能系统，系统基于这些特征标记进行自我学习，并实现智能识别。现在通过数据标记，人工智能不仅可以识别人脸，还可以自动识别色情影像、医学疾病检测图片等，还正在为自动驾驶汽车上路铺路。人脸识别的背后，是这些“数据标记工厂”员工的大量标记工作。

由于篇幅较长，本文节选全文重点段落进行精讲。

选文精讲

Data 数据

China’s success at AI has relied on good data

中国在人工智能领域的成功依赖于好的数据

But cheap labour has also played an important part

但廉价劳动力也发挥了重要作用

CHINA IS THE land of face recognition. Cameras able to extract face prints from passers-by are common in the streets of large cities like Guangzhou and Shenzhen. Boxyvending machinesat airports offer to let you pay for a cup of orange juice, robot-squeezed for perfect freshness, by scanning your face. From December 1st all people applying for an account with one of China’s telecoms companies such as China Mobile must have their face scanned. Previous regulations required proof of identity, but the possession of users’ face prints will let firms verify identities inreal-timevia smartphone cameras.

vending machine: 自动售货机

real-time: 实时、同时

中国是人脸识别的国度。在广州和深圳等大城市的街道上，摄像头从行人身上提取面部图像很常见。机场四四方方的自动贩卖机可以通过扫描你的脸，来支付一杯由机器榨出的新鲜果汁。从12月1日起，所有向中国移动等电信公司申请开户的人都必须进行面部扫描。以前的管理要求提供身份证明，但现在拥有用户的面部图像的公司，可以通过智能手机摄像头实时验证身份。

Considering the oppressive purposes to which this technology is being put, it would not be appropriate to call China’s rapid adoption of it anything more than a technical success. The underappreciated fact that companiesleaping aheadin the field are more reliant on cleverly deployed cheap labour for their progress than on any technological edge, suggests another reason for caution before declaring a Chinese victory in the tech wars. But understanding how China has got face recognition to flourish is nonetheless instructive. Two of the world’s most valuable startups, Megvii and SenseTime, worth $4bn and $7.5bn respectively, are Chinese AI companies specialising in the field. Their application of it alone would make it one of the most widely deployed forms of artificial intelligence in the world.

anything more than: 不仅仅是

leap ahead: 引领、飞越向前

因为这一技术还会用于一些压制性的目的，中国迅速采用这一技术不仅仅是技术上的成功。在这一领域取得领先地位的企业，其进步更多地依赖于廉价劳动力，而非任何技术优势，这意味着，中国在科技战争中取得胜利之前，我们有理由保持谨慎。了解中国是如何让人脸识别蓬勃发展起来的，仍然具有启发性。全球最有价值的两家初创公司——分别价值40亿美元和75亿美元的Megvii和SenseTime——都是专注于该领域的中国人工智能公司。他们的应用将使其成为世界上应用最广泛的人工智能形式之一。