July 31, 2019

How AI Can Detect Fake News

With the increasing popularity of social media across the world, fake news has been spreading at an alarming rate, especially over the past couple of years. Although a quick search online can help with fact-checking information, new technology has been able to produce fake videos from image portraits. 

 

From a single headshot, the latest tech innovations can create a moving image that depicts semi-natural and real behaviour. This makes differentiating fake from real even more cumbersome and complex.

 

This is known as “Deepfake technology” and although it is not new, recent research has allowed it to become easier to produce. While creating videos out of still portraits is groundbreaking for various types of content and design, some experts worry that it could be used by the wrong hands to spread fake news. 

 

How fake news spreads

A recent report by MIT has identified that fake news and genuine news online spread differently, with false news spreading faster than legitimate news stories. 

 

The report found that false news was ‘newer’ and more novel than genuine news, and often “inspired fear, disgust, and surprise in replies” while genuine stories “inspired anticipation, sadness, joy, and trust.” [1]

 

The study also unfolded that humans are more likely to spread false news quicker than real news.

 

AI saves the day

Labs and startups are working towards utilizing AI to detect fake news by using machine learning algorithms – specifically ‘geometric deep learning’, an emerging field that is more complex than traditional machine learning techniques. 

 

Fake news has characteristics that distinguish it from real news, and this is where AI systems can tap in to identify fake news online. Geometric deep learning works alongside network-structured data, thereby identifying and categorizing data according to its features. 

 

Image source: TechCrunch

 

The above image identifies how fake news and real news spreads; online users who often share fake news are in red while users who share real news are in blue. This showcases how there is a recognizable pattern that differentiates each group. [3]

 

Partnering with third-party fact-checking NGOs or websites can help AI startups create faster systems for identifying fake news. 

 

AI detects fake news across borders

Social media networks have employed thousands of reviewers to remove inappropriate content on its platforms; however, its staff primarily speak English – the world’s global language, and there may be a shortage of reviewers in other languages. 

 

Due to this, content reviewers are unable to identify fake news in other languages. With emerging AI tech that identifies fake news due to patterns and data – language barriers are eliminated. Hence, fake news can be spotted online globally as it is not language-dependent.

Essentially, AI can utilize geometric deep learning to spot patterns in how content online – predominantly fake news – spreads.

 

Want to integrate AI into your business? Get in touch http://saal.ai/contact/

 

Sources:

[1] Massachusetts Institute of Technology (MIT), The spread of true and false news online, Science, Vol. 359, Issue 6380, pp. 1146-1151, March 2018, DOI: 10.1126/science.aap9559 

[2] The Telegraph, How artificial intelligence has the ability to detect fake news, March 2019

[3] TechCrunch, Fabula AI is using social spread to spot ‘fake news’, February 2019