As artificial intelligence moderates publication in Facebook
In Facebook is actively fighting the unacceptable content on the pages of their social network. In this struggle, the company's employees are actively help the artificial intelligence. However, not all the objectives of artificial intelligence to cope quite well, it said Facebook CEO Mark Zuckerberg, communicating with analysts during a discussion of quarterly performance of the company.
The involvement of the artificial intelligence in the content moderation on Facebook is growing. However, still can not completely rely on technology. Mark Zuckerberg said that artificial intelligence is not good at all. In particular, there are big problems with the understanding of human speech. In Facebook is very effectively removed the publications that are related to the promotion of terrorism. According to the CEO of social network, it is a merit of AI. In addition, the artificial intelligence is extremely effective at pictures of naked people. But insults and angry Pub - a weak spot technology.
"The easiest way to teach artificial intelligence to find the nipple on the image than to recognize abuse," - says Mark Zuckerberg. With AI 99% of publications that promote terrorism, are removed without the participation of employees, but to achieve efficient retrieval of abuse can take years, says the founder of Facebook.