- The New Paper
China’s SenseTime, the world’s most valuable artificial intelligence (AI) start-up, has taken a big step towards fully automating online censorship amid reinforced efforts by the central government to crack down on content not consistent with “core socialist values”.
The Beijing-based company, which is known for providing AI-powered surveillance software for China’s police, on Wednesday unveiled a product it claims can automatically screen out online videos that contain pornographic or violent images, as well as text containing messages deemed sensitive by the authorities.
“With more and more original content being created and published online, it is becoming impossible to use human content reviewers to take down all disturbing videos. That’s why we want to introduce the new product to better protect our next generation,” said SenseTime executive director Dai Juan.
The new product comes amid the Chinese government’s ongoing crackdown on “inappropriate” online media, which has included sexually provocative material uploaded by under aged users, as well as off-colour jokes and content deemed as vulgar. In recent weeks Chinese authorities have ordered online media to remove offending content and step up their self-censorship efforts.
The clean-up in China is running in parallel to a growing debate in the US over the role and responsibility of social networks like Facebook in policing free speech and content, amid wider concerns whether social media is exacerbating divisions in society.
Dai said a human content reviewer can only screen several thousand pictures and videos a day, meaning a platform that receives more than 1 million user generated videos daily would need a team of at least 2,000 people monitoring content, costing around 400,000 yuan (US$63,261) a day.
SenseTime’s AI-based solution, called SenseMedia, only costs 3,000 yuan per day to operate and has an accuracy rate of 99.5 per cent, she said.
Major social media platforms from Facebook in the US to China’s ByteDance have been hiring more human moderators to monitor and judge content after criticism from their respective governments for hosting inappropriate content.
Facebook has increased the number of its content reviewers by 40 per cent to 7,500. ByteDance, the operator of China’s most popular news site Jinri Toutiao, is looking to hire 2,000 content reviewers, with a preference for Communist Party members. The company already has around 4,000 employees dedicated to the task of scrubbing its site of inappropriate material.
AI-based systems with facial recognition software are also being used by Chinese cities and securities agencies to apprehend criminals and cite traffic offenders. Earlier this month a fugitive was arrested in southeast China after facial recognition technology helped identify him in a crowd of about 50,000 attending a pop concert. The country is also exporting such technology, with start-up Yitu Technology selling its body mounted cameras equipped with facial recognition software to the Malaysian police.
As well as SenseMedia, SenseTime announced a solution to prevent drowsy driving by using AI to monitor the driver’s condition. The system sends an audible alert if the driver starts yawning or looking in other directions.
Tang Xiaoou, a professor at the Chinese University of Hong Kong who also serves as SenseTime’s chairman, said the company is already profitable.
Earlier this month, SenseTime received a new round of funding worth US$600 million, led by Chinese e-commerce giant Alibaba Group Holding, which is the parent company of the South China Morning Post. The new funding reportedly gave the company a total valuation of more than US$4.5 billion, making it the world’s most valuable AI start-up, according to analyst firm CB Insights.
For the latest news from the South China Morning Post download our mobile app. Copyright 2018.