A new algorithm capable of creating generate fake talking-head videos with just a few images of a person’s face was just released by Samsung’s AI center in Moscow.
A.I. is getting better and better at producing fake videos, for everything from amusingly adding Nicholas Cage into movies to maliciously spreading fake news.[…]
Hollywood has been creating fake videos using special effects for years. The argument for this particular AI is that it is merely democratizing special effects. However, watch out for increasing use in news stories. As this technology gets better and better, it inevitably leaves you wondering who and what you can trust.
The fake of Nancy Pelosi on Facebook required a huge amount of data to look real, using Convolutional Neural Networks (CNN’s). However, Egor Zakharov and his team simplified this system by creating a system that can replicate that using just a few images. Their system is trained on a big dataset of videos in order to learn facial profiles and create an adversarial training problem. From just a few images, the system is then able to create a generative adversarial network (GAN) in a person-specific way.