How AI holds the key to making the internet safe

Pioneers of new industry sector Viztech are taking the neural network technology behind iPhone X’s facial recognition system to a whole new level to deliver a range of applications including video detection to solve the growing crises in online extremist content and brand safety. David Fulton, CEO of WeSee, explains… 

Could you imagine the Internet without Google? Trying to find what you were looking for would be a lot more painful that it is today – not to mention incredibly time-consuming. Well, that’s the situation we’re facing right now with respect to digital  video content.    

Why should we care? Well, the amount of video content on the web is increasing rapidly. According to recent research by Cisco, by 2020 there will be some 65 trillion images and 6 trillion videos uploaded to the web. This will result in more than 80% of all the traffic on the internet being either image or video-based – and that’s  in less than three years’ time.  

The major engines, such as Google, rely on human-tagged meta-data to carry out their searches. These meta-data tags are written and added when text-based content is created making it visible to search engines and therefore searchable. With more video content currently being uploaded to the internet in 30 days than the major US television networks have created in 30 years, it would take more than 100,000 years to watch all of it. More specifically, it would take over one million years to tag it by hand in the way that we tag meta-data today. This essentially makes this rapidly growing body of image and video content on the internet impossible to detect and categorise using current techniques – and therefore very difficult to find. This could mean that by 2020 some 80% of the internet will be unsearchable using current techniques.    

This inability to find, track and filter video content automatically is causing increasing problems, and this is no more so than in the area of policing dangerous and inappropriate visual material. This includes the videos produced and uploaded by terrorist groups, something which UK Prime Minister Theresa May recently called on Internet giants – including Facebook, YouTube and Microsoft – to come up with a way to detect and remove within two hours or face huge fines. We are seeing this sentiment backed by lawmakers in Germany who have recently approved a bill that potentially gives prison sentences for certain “evidently illegal” content.  

Not being able to spot this kind of content is also hitting business in terms of its potential to damage reputations. Brands are increasingly finding that their advertisements are running alongside inappropriate visual material. This in turn creates guilt by association, and has given rise to the ‘brand safety’ crisis. This is having a negative effect on the big internet players like Facebook and YouTube – along with publishers in general – who are scrambling around desperately trying to find a solution. With advertisers’ trust in their platforms gradually dissipating, a vital revenue stream worth a billion dollars is at risk.    

The answer lies in the new field of Viztech, a fledgling sector that could have ramifications far beyond curbing extremist digital content and brand safety. Viztech pioneers are developing innovative artificial intelligence-based technology that can be trained to spot inappropriate or specific content, such as an ISIS flag or face of a known hate-preacher. This powerful technology can detect and categorise video, as well as still images, quickly and efficiently, processing information just like the human brain, but up to 1,000 times faster… and counting!   

It is driven by deep learning and neural networks, similar to the technology that’s behind the new iPhone X’s facial recognition system but more sophisticated, and like Apple it promises to take advanced AI to the masses. It doesn’t just see visual content, but also understands every multi-layered element within images and videos in the same way we humans do. It allows organisations to automatically ascertain and harness the huge opportunities and value hidden inside all digital images and videos. 

Even at this early stage, technology within Viztech has the power to transform industries. Take broadcasting as an example, where it could enable the development of a proper video search engine and, more importantly, quickly and easily categorise and tag video content on-the-fly. Something that historically took days to be done manually can now be carried out automatically in a matter of seconds. Broadcasters will be able to know the specific video content adjacent to all ads. From an advertising and revenue generation perspective, this is quite simply transformative.    

It goes without saying that it could also power the world’s most advanced adult and violent digital content filter, creating a more child and brand-friendly web. And that’s sure to put a smile on the face of Theresa May and the digital publishers that lawmakers and advertisers are collectively currently putting pressure on to police their content.   

Meanwhile, looking further into our near future, the Viztech sector will soon be unleashing an application of the technology that reveals whether people are lying or being truthful through advanced facial recognition. This is something that could revolutionise the insurance industry (in reference to claims), and also potentially aid law enforcement.   

So, perhaps the future envisaged by the Tom Cruise movie MINORITY REPORT is closer than we all might have imagined. Thanks to the pioneers of Viztech, it is now possible to be predictive rather than reactionary when it comes to monitoring visual digital content – filtering, identifying and categorising video before it has even appeared online. Ultimately, the connection between human intuition and understanding the multi-layers of visual content, together with the processing speed of AI, merely scratches the surface of the revolutionary possibilities within Viztech. 

David Fulton, CEO, WeSee 

Image Credit: Enzozo / Shutterstock