Among Silicon Valley tech giants, Google maintains leadership in new fields such as artificial intelligence and machine learning. The search giant has deployed AI techniques in key areas like object recognition and detecting photo-shopped images. Now, the company wants to expand the technology to recognizing fake videos as well.
A Google engineer named Supasorn Suwajanakorn has developed a new tool which can create a realistic fake video when fed with the right input. The tool has the capability to mimic the way a person talks by closely observing existing footage of their mouth and teeth and then recreate a perfect lip-sync. The technology is seen capable of spotting fake videos using artificial intelligence.
Since the technology has the potential to recreate perfect lip-sync, it has the great potential to create harm as well. In order to ensure, the technology does not fall in the hands of bad actors, Suwajanakorn is working with the AI Foundation on an app called ‘Reality Defender’ that AFP reports would run automatically in web browsers to spot and flag fake pictures or videos.
“I let a computer watch 14 hours of pure Obama video, and synthesized him talking,” Suwajanakorn said while describing his work at the TED Conference in Vancouver on Wednesday.
The technology could be used to recreate virtual avatar of those who have passed and there have been instances where AI researchers have created their deceased parents or friends using machine learning and old archive. Suwajanakorn says the technology can be used to recreate grandparents, who could be asked for advice; actors, who can return to the screen; great teachers can once again give lessons while authors read their works aloud. He also notes a ‘New Dimensions in Testimonty’ project that allows people have conversations with holograms of Holocaust survivors.
Considering the potential of technology created by him, Suwajanakorn is also developing a counter technology that will detect fake images and video. He says ‘Reality Defender’ will automatically scan for manipulated images or video and allow users to report apparent fake in order to create a defense mechanism. “Video manipulation will be used in malicious ways unless counter-measures are in place,” he told AFP.
Watch: All you need to know about Google Go
The details of Suwajanakorn’s work comes as the world continues to battle spread of fake news and the misinformation created by them. He says while it is easy to write fake news, it will be expensive to create a manipulated video.