Recently, China has made international headlines for utilizing a number of deep learning solutions to crack down on crime. However, a recent incident suggests its highly touted facial recognition technology has a few bugs in it. Earlier this month, a well-regarded businesswoman was incorrectly identified as a jaywalker.
When Deep Learning Makes a Mistake
Like many of China’s urban hubs, the city of Ningbo, Zhejiang has been outfitted with surveillance equipment, including facial recognition systems, designed to identify criminals. One camera captured Dong Mingzhu, CEO of Gree Electric Appliances, as she seemingly jaywalked through an intersection. Subsequently, her photo, name, and part of her government ID number were displayed on a digital billboard intended to shame local lawbreakers.
But as it turned out, the system had made a false positive. The facial recognition system actually detected Dong’s face on the poster of a local bus. The executive was not actually there. Moreover, the malfunctioning system incorrectly listed her surname as “Ju.” As the billboard made the system’s error obvious, users of the popular Chinese social network Sina Weibo began mocking the high-profile mistake.
Once the gaffe was discovered, the Ningbo traffic police acknowledged the error and explained that the record of Dong’s violation had been deleted. The police also noted that the facial recognition system would be upgraded to prevent future false positives. In response to the Kafkaesque mistake, Gree issued a statement thanking the Ningbo police for their hard work.
China is All In on Artificial Intelligence
Despite the unsettling implications of Dong’s misidentification, China seems dedicated on utilizing deep learning solutions to further its long-term goals. In the summer of 2017, China’s State Council announced plans to make itself the worldwide leader in artificial intelligence innovation.
As part of the initiative, China has already committed $2.1 billion to build a state-of-the-art 135-acre AI technology park in Beijing.
Although Chinese officials have named smartphones, smart houses, and autonomous vehicles as focus areas for their deep learning research, surveillance and military applications are also on the agenda. However, despite all the money and resources China is dedicating to AI, the results have been mixed.
The Future Isn’t Ready Yet
Earlier this month, it was announced that cutting edge “gait recognition” technology has been deployed for use in Beijing and Shanghai. The AI-powered surveillance tool is designed to identify individuals by their unique body shape and movements.
Though gait recognition tech has been found to accurately recognize criminals at distances up to 165-feet away, it doesn’t work as quickly as other biometric systems. The software utilized by the Chinese government can’t perform analysis in real time; it makes identifications in recorded videos at a rate of 10 minutes per 1 hour of footage.
Similarly, it was recently found that one of China’s AI partners may not be on the level. Remark Holdings, a Las Vegas-based company that also owns swimwear retailer Bikini.com, has established itself as an AI player in China. Remark has worked with the Chinese police to catch motorcycle-riding lawbreakers and has developed AI-powered security solutions for casinos and smart checkout systems in supermarkets.
But in spite of these partnerships, Remark has underperformed financially. In 2017, the company reported revenues of $70 million, but their earnings dropped to $4.6 million in the first six months of 2018. Additionally, corporate analysis firm J Capital Research reported that Remark has exaggerated its connections to major Chinese companies and that the firm’s declared backend resources do not match its spending.
With malfunctioning facial recognition systems, less than optimal gait recognition, and partnerships with reportedly questionable companies, China’s AI revolution might define the future, but it’s struggling in the present.