Social Media Companies Feeling The Pressure To Get AI Right

With new laws on hate speech, fake news, and online abuse, social media needs AI


Mark Zuckerberg has been pretty clear about Facebook’s role in the spread of fake news. In a long post in November 2016, he wrote ‘This is an area where I believe we must proceed very carefully though. Identifying the "truth" is complicated. While some hoaxes can be completely debunked, a greater amount of content, including from mainstream sources, often gets the basic idea right but some details wrong or omitted.’ He also added ‘Of all the content on Facebook, more than 99% of what people see is authentic. Only a very small amount is fake news and hoaxes.’

Normally when you talk about a 1% edge case it can largely be ignored or at least put to the back of the queue. However, given that Facebook averages 350 million photo uploads and generates 100 millions hours of videos views every day, that represents 3.5 million fake images and 1 million hours of fake news being watched every 24 hours. These aren’t small edge case numbers, these are big problem numbers.

The spread of fake news, revenge porn, and hate speech are activities that governments are beginning to push back on too, with Germany leading the way. In a statement (translated by Google Translate) the German cabinet said of hate speech ‘The providers of social networks are responsible when their platforms are abused to spread hate criminality and criminal false news. There can be just as little space on social networks for criminal acts as on the street. To do better, we owe the victims of hate criminality. The Internet is characterized by the debate culture and the social climate in our country. Verbal radicalization is often the precursor to physical violence.’

Germany’s government is proposing that if social networks do not remove hate speech, fake news etc, within 1 week of being reported they will face a €50 million fine, with individuals from these companies also personally facing €5 million fines. The move comes after the German government reached an agreement with Facebook, Google, and Twitter in December 2015 that they would remove all hate speech within 24 hours, but have been unhappy with the results.

This represents huge potential damage to these companies. Looking at the numbers above, there could hypothetically be €175 trillion worth of fines every single day for image uploads alone if Mark Zuckerberg’s 99% stat is accurate. It is also not feasible for humans to go through the huge numbers of uploads, which leaves AI and machine learning the only feasible options.

At present we know that Facebook has not perfected this. Firstly there was the debacle in late 2016 when they moved to a pure AI vetting system which allowed several fake news articles to make it onto their trending articles list. They have also recently announced a move to take revenge porn off their site, with Facebook’s Global Head of Safety, Antigone Davis, telling TechCrunch that ‘At this moment, we’re not using AI to go through this particular content…There is significant context that’s required for reviewing non-consensual sharing.’

Therefore this challenge being set forth by the German government is not simply about a social issue but is, in fact, challenging social media companies to speed up and improve their development of AI in this area. It is certainly not an easy task, but something that they will need to work on to perfect. Regardless of whether you believe the fines to be too steep, social media companies have a lot of work to do. 

Looking small

Read next:

Expert Insight: 'An Effective Visualization Results From A Great Deal Of Curiosity And Exploration'