Google computer scientist is working on an app capable of detecting fake footage of celebrities and politicians, generated by the advanced software
Google engineer Supasorn Suwajanakorn demonstrated his software he had trained on 14 hours of footage of former US President Barack Obama during a TED conference in Vancouver. The system uses a neural learning network to mimic functions of human brain, studies the subject’s teeth, mouth movement, and facial expressions and learns to mimic them uncannily. Suwajanakorn is working with the AI on an app called Reality Defender, which could run automatically in web browsers to detect fake videos or doctored pictures. “Video manipulation will be used in malicious ways unless counter-measures are in place,” he told AFP. “We have to make it very risky and cost-ineffective.”
University of Washington is concerned regarding how easily such technology could be manipulated to create videos which have real appearance. “These results seemed intriguing, but at the same time troubling; it concerns me, the potential for misuse,” he said. “We don’t want it to be in the wrong hands, so we have to be very careful about it.” One example he used was real-seeming footage of a world leader announcing nuclear war circulating on social media and across the internet.
Similar technology is being used to edit the face of female celebrities onto the bodies of pornographic actors, creaking videos known as ‘deepfakes’. Twitter and Reddit announced crackdowns on the videos in February following a spike in uploading. While people have been doctoring pornographic pictures with female celebrity faces for decades, advances in machine learning, a branch of AI that enables computers to observe and learn new behaviors they have not been programmed to complete, have facilitated the creation of convincing porn footage. However, Suwajanakorn acknowledged the technology could also be harnessed to resurrect deceased actors in films or family members, for more positive ends.