Alibaba tests AI tech to identify, remove online pornography
Chinese e-commerce giant Alibaba revealed its voiceprint technology to identify online pornographic content, in a move to crack down on online pornography.
The AI technology can identify multiple languages and Chinese dialects, such as Japanese, Russian and English, or the Hunan and Sichuan dialect, according to a statement Alibaba sent to the Global Times on Monday.
It can identify “the groan sound” as well.
Offline and real-time voices can be converted to words. By matching content on its database, the AI identification program will decide whether it contains any sexually suggestive content, the statement said.
The technology can be used on headlines, comments, product descriptions on social media platforms, news outlets, and e-commerce platforms and live-streams.
Wei Shi, a staff algorithm engineer at Alibaba, told the Global Times the technology can be used to contain risks in terms of pornographic and vulgar content for companies at a low cost.
A trained inspector can examine about 15,000 pictures a day, but an AI program gives people the ability to examine up to 24 million pictures for an annual fee of 60,000 yuan ($8,753), Wei noted.
It’s 99.5 percent accurate in identifying pornographic pictures, but cannot replace people, as machines still find it difficult to comprehend nuance.
The National Office against Pornographic and Illegal Publications ordered websites such as 163.com, baidu.com, bilibili.com, and qingting.fm to remove ASMR-related pornographic content in June, Xinhua News Agency reported.
ASMR – autonomous sensory meridian response – is a tingling sensation that typically begins on the scalp and moves down the body in response to certain stimuli. ASMR content is primarily used for sleeping or relaxing.
Over 80 percent of juvenile criminal cases are linked to online pornography and violence, 2017 China Internet Network Information Center data shows. GLOBAL TIMES
The post Alibaba tests AI tech to identify, remove online pornography appeared first on The Manila Times Online.