Research into artificial intelligence and deep-learning neural networks is now expanding into multimedia manipulation, a field that may have deep implications for news and social media in coming years. AI researchers have been successfully creating 3D face models from still 2D images, generating sound effects from silent video, and even mapping the facial expressions of actors onto other people in videos. So far, these technologies have only been used to make amusing YouTube videos, but in the future they could be used to generate completely believable fake news.
One currently available application of this technology is Smile Vector, a Twitter bot that can make any still photo of a person's face smile. The app uses a neural network powered by deep learning to automatically morph the expressions on a still photo. “I definitely think that this will be a quantum step forward, not only in our ability to manipulate images but really their prevalence in our society,” Tom White, Smile Vector's creator, said. “I don’t think many people outside the machine learning community knew this was even possible.”
“The field is progressing extremely rapidly,” Jeff Clune, an assistant professor of computer science at the University of Wyoming, said. “Jaw-dropping examples arrive in my inbox every month.” Clune and his team have utilized research on the human brain to help teach deep-learning neural networks to recognize objects better. The networks can create new, realistic images of ants, volcanoes, birds, and other things. These images are currently limited to 256 pixel squares, but Clune believes that full photorealistic HD images could be possible within the next few years.
Clune also believes that this new technology will be useful in the creative industries, allowing anyone from furniture designers to video game developers to use it to inspire or even create new ideas. But the technology also lends itself well to being used for hoaxes. The creators of a program called Face2Face have released a demo video showing the facial expressions of an actor being realistically projected onto the faces of Trump, Putin, and Obama. In conjunction with new Adobe VoCo software that allows a user to re-write any recorded speech, anyone could use this technology to have any celebrity or world leader appear to be saying something that they did not in reality.
“I currently exist in a world of reality vertigo,” Clune said. “People send me real images and I start to wonder if they look fake. And when they send me fake images I assume they’re real because the quality is so good. Increasingly, I think, we won’t know the difference between the real and the fake. It’s up to people to try and educate themselves.”