SenseTime says its AI can censor politically sensitive messages
China’s SenseTime is best known for providing the country’s police with AI-powered surveillance software. Now it says it has software that can help China’s strict online censors filter out controversial chatter on social media.
The world’s most valuable AI startup is releasing SenseMedia, a “smart content reviewing solution,” according to the South China Morning Post. The company boasts that it can block pornographic and violent content with a "99.5% accuracy" rate -- and also filter politically sensitive discussion.
It comes as time when services around the world grapple with the challenge of filtering disturbing content. The sheer volume of things uploaded to social media every day would seem too much for humans to wade through -- but designing software to solve the problem has proven difficult. Facebook now has 7,500 content reviewers working in 40 languages, while YouTube plans to hire over 10,000 people to watch over the platform.
But it’s even more of a problem in China. What you can and can’t say online is tightly controlled by the government -- and the penalty for sites that break the rules is severe.
SenseTime says its software will filter discussion on “politically sensitive figures.” Social media in China is often censored, with various terms blocked as users use seemingly unrelated words to discuss controversial topics.
News aggregator Toutiao was shut down entirely for 24 hours after the government accused it of spreading “pornographic and and vulgar content,” leading the company to hire 2,000 content reviewers in January (in addition to its existing team of 4,000).
SenseTime also told us that the product is partnering with China’s TCL, the world’s third largest TV manufacturer, to review content on the TV’s video platforms. But it refused to reveal what that meant -- or explain how it might work.
(One of SenseTime's backers is Alibaba. Abacus is a unit of the South China Morning Post, which is owned by Alibaba.)