In China, human vs. machine face-offs are just as much about entertainment as they are about technology. On Thursday, Ant Financial, the financial affiliate of Alibaba Group, held a facial recognition contest between T.V celebrity Wang Yuheng and “Mark” (蚂可), the facial recognition AI for Alipay, Ant Financial’s mobile payment system.
Over the course of three half hour rounds, Mark and Mr. Wang identified internet celebrities at the event from hundreds of photographs. Both human and AI guessed correctly for the first two rounds, which involved 150 and 300 photographs, respectively. However, in the third round, when Mark and Mr. Wang were asked to identify the childhood photographs of two livestreaming hosts, and Mark lost.
“We wanted to see how the recognition abilities of super humans like [Wang Yuheng] compared to those of a machine,” Dr. Chen Jidong, the Senior Data Expert at Ant Financial, told TechNode.
“We want to absorb and incorporate their special recognition abilities into our algorithm so that our AI can more safely and conveniently service users,” he says.
Mr. Wang is known for his appearance on the Chinese talent show, “The Brain” (最强大脑), which pits geniuses from different countries against each other. To win, participants complete various challenges, such as identifying which glass a judge drank out of given 520 glasses of water – the challenge that earned Mr. Wang the nickname “Water Brother” (水哥, our translation).
“I believe that Mark has learned a lot in interacting with Water Brother, but it has a long way to go when compared to humans,” said Dr. Chen after Mark lost in the final round. “This event isn’t really a [battle]. We’re just hoping that more people will understand this technology.”
The contest is largely a publicity stunt, similar to the one by Alibaba in April. Using information like social media content and song popularity, the Alibaba’s AI accurately predicted the finalists and winner of Chinese reality singing show I’m A Singer. Mark’s algorithm was developed by Face++, one of Ant Financial’s partners. The Beijing-based startup specializes in face recognition technology and also powers Alipay’s “smile to pay” service.
Alipay’s facial recognition feature has been around for about a year, but is still being perfected, says Dr. Chen. As a finance-related application, Ant Financial has “very strict” requirements when it comes to its false acceptance rate (FAR), or identifying the wrong person. The technology also has to mesh well with the Alipay app and deliver a smooth user experience for all kinds of users. The point is to someday make passwords obsolete, says Dr. Chen.
“We hope that users will try [our facial verification feature] under different light settings, different angles…even while wearing makeup to see if it will pass,” says Dr. Chen. Improving Mark’s performance under challenging environments is one of Ant Financial’s ongoing initiatives. In the case of the childhood photographs, many were shot in dim lighting, contributing to Mark’s errors.
In addition to photo quality, data security is one of the key challenges for financial applications of biometric technology. Simply using one biometric method – a face scan – to verify someone’s identity is not ideal, says Dr. Chen. In the future, Ant Financial plans to incorporate other biometric methods as well, such as behavioral patterns and the user’s social network.
In addition to Ant Financial, other Chinese companies, such as Ping An, are also looking at using biometric identification. In April, Ping An launched its face recognition loan technology, which cuts down the loan application process to six minutes and can differentiate between twins, according to the company.
Image credit: Ant Financial
from TechNode http://ift.tt/29iyKKV