News by VICE

Mark Zuckerberg says it could be a decade before Facebook fixes its hate speech problem

AI can't save us from this one.

by Christianna Silva
Apr 10 2018, 10:02pm

Facebook CEO Mark Zuckerberg said Tuesday that the company is struggling to figure out the best way to deal with hate speech — and the answer might be a decade away.

The social media giant is increasingly relying on artificial intelligence to identify and remove content posted to the site that violates its policies. But AI isn’t able to handle hate speech yet, Zuckerberg said during a joint hearing of the Senate Judiciary and Commerce committee on Tuesday.

Zuckerberg told senators that AI works great for some things, like removing terrorism posts, which he says are largely caught by their AI system before any human views the post. He claimed that more than 90 percent of pro-ISIS content is automatically flagged by Facebook’s AI, and that the companies’ tools can even detect when users may be at risk of self-harm and can help intervene. But hate speech is different, Zuckerberg says, and the technology just hasn’t caught up with the language.

“I am optimistic that over a five-to-10-year period, we will have AI tools that can get into some of the linguistic nuances of different types of content to be more accurate in flagging content for our systems, but today we’re not just there on that,” Zuckerberg said. “Until we get it more automated, there’s a higher error rate than I am happy with.”

Ideally, AI would catch hate speech before users do, like it does with terrorism posts. For now though, Zuckerberg said most hate speech is caught first by users, who flag it as offensive content. When Republican Sen. John Thune of South Dakota asked Zuckerberg what steps Facebook was taking to improve its ability to define hate speech, the CEO said it’s all a work in progress.

“Hate speech is one of the hardest,” Zuckerberg said. “Determining if something is hate speech is very linguistically nuanced. You need to understand what is a slur and whether something is hateful, and not just in English…”

He added that one of the biggest problems with identifying hate speech on Facebook is that the network operates all over the world, and they just don’t have an AI that can understand nuances in multiple languages. Zuckerberg said they’d be hiring content reviewers who are knowledgeable in multiple languages.

During his testimony, Zuckerberg stressed his dedication to solving the issue, particularly in the aftermath of hate speech that polluted Facebook in Myanmar and facilitated ethnic violence against Rohingya Muslims.

Mark Zuckerberg
hate speech
facebook hate speech
mark zuckerberg testimony
facebook ai