Thread by @WIRED

Facebook has a lot of problems. One of the most serious and ongoing is that the people in charge of moderating its platform for hate speech, violence, porn and more are suffering severe trauma. (via @verge) 1/
wired.trib.al/o4wFXG4

The @Verge’s new piece goes into graphic detail for US based Facebook moderators. But social networks have outsourced this work for years, raising questions about the psychological toll of repeated exposure to violent and disturbing images. 2/ wired.trib.al/HcsuzV9

Social networks rely on a hidden workforce of contractors who make much less than full-time employees, and who often sign confidentiality agreements that stay in effect even after their employment ends. 3/ wired.trib.al/yAjkZMg

And of course it’s not just Facebook. Users upload 600,000 hours of new video to YouTube daily. The site still relies on human moderators to review content—and to train its AI systems to do the same work. 4/ wired.trib.al/HGDcI0P

When Mark Zuckerberg testified before Congress in April 2018, he said having AI identify hate speech was “one of the hardest” problems. He was optimistic that in 5-10 years, "we will have AI tools that can get into some of the linguistic nuances.” 5/ wired.trib.al/QjikBpZ

For now, though, machines are still pretty easy to fool. In one example, called the “love” attack, researchers were able to trick AI into thinking a hateful sentence isn’t just that. 6/ wired.trib.al/gzduTRw

Facebook has faced a dizzying series of accusations and scandals over the past year. And many of Facebook’s apologies exhibit a common theme: Artificial intelligence will help solve the problems incubating on the company's platform. 7/ wired.trib.al/wUcE0jp

Counting on AI is a gamble. Algorithms have proved capable of helping to police Facebook, but they are far from a cure-all. And until then, the people who manually moderate FB will be subject to the horrible things that the rest of us put on there. 8/ wired.trib.al/9RMFZW9