In September, Facebook, in partnership with Microsoft and a list of academic institutions, announced a $10 million research effort to help build a better Deepfakes detector. To that end, companies like Google, Facebook, and Twitter have a critical role to play. In this arms race, it is clear that the entity that has the most data will win - the more data that you have, the more accurate and sophisticated a model you can build. “As of now, we lack automated ways to detect Deepfakes in a reliable and scalable fashion”, Dawn Song, a Computer Science Professor at the University of California Berkeley said, “It will be an arms race between those that create Deepfakes and those seek to detect them.” In fact, the technology used to detect Deepfaks can be used to generate them.
Modified or photoshopped images are certainly nothing new, but the difference between Deepfakes and these prior image manipulation techniques is that there is no reliable way to detect Deepfakes today. With technologies like Deepfake, revenge porn or cyber smear campaigns can take on a whole new dimension and potentially impact a wider population of victims. As we saw in the case with Representative Katie Hill, revenge porn can end someone’s political career. In addition to violating the victim’s rights to her own images and privacy, a major concern is Deepfakes’ potential impact on revenge porn. Not surprisingly, the main victims of the fake videos are women, whose images and likenesses are used without their consent, and often without their knowledge. According to the Deeptrace report, 96% of the Deepfake videos on the Internet are pornographic videos. The painting was created by a group of French students, using a dataset of approximately 15,000 portraits from the 14th to the 20th century.Įven though AI techniques like Deepfake have many applications, including photo editing, image repair, and 3D transformation, unfortunately, their main application today is generating sexually explicit videos for cyber exploitation. In October 2018, Christie, the auction house, sold an oil painting generated by GAN for $435,000, nearly 45 times its highest estimate.
0 kommentar(er)
