Facebook Is Using AI To Detect “Hateful Memes”

In its latest maneuver against hate speech and misinformation, Facebook announced this week that it has created a database of 10,000 “hateful memes” that will be made available to researchers. Facebook is also launching a $100,000 prize for researchers to develop further AI models using the data.

Facebook relies heavily on artificial intelligence to remove offensive content, but this kind of technology can have a difficult time parsing images and video. AI first needs to be shown many examples of the offending content before it can take action. And even still, slight changes to an image can help it avoid being found.

“People sharing hate speech often try to elude detection by modifying their content,” Facebook said in a statement. “As we improve our systems to address these challenges, it’s crucial to get it right. Mistakenly classifying content as hate speech can mean preventing people from expressing themselves and engaging with others.”

Thankfully, Facebook says that its newly-trained AI model can recognize copies of hateful memes and even near-duplicates. Considering that the company says a “substantial portion” of the hate speech on its platform lives in images and video, this could potentially be a big breakthrough for cleaning up the site.




Recommended Resources

bitdefender The Choice of Tech Experts Worldwide. Try 90 days free of Bitdefender 2020 and experience the highest level of digital safety.

PIAPrivate Internet Access is an award-winning, cost-effective VPN solution. The use of an anonymous and trusted VPN is essential to your online privacy, security and identity protection.

System Mechanic 14 – Make your computer run like new. Winner of 200+ Editor’s Choice awards!

Facebook Cracks Down On Conspiracy Video — But Not Before It Racked Up 1.8 Million Views Previous post Facebook Cracks Down On Conspiracy Video — But Not Before It Racked Up 1.8 Million Views Facebook Warns That COVID-19 Has Hampered Its Content Moderation Efforts Next post Facebook Warns That COVID-19 Has Hampered Its Content Moderation Efforts