Researchers from Samsung have developed a new AI, that can create fake videos from a single source image. The videos shared by Samsung Researchers are made from the same AI and are very much identical to the picture. Although they are fake videos but looks like they are indeed original videos of the person in the image, this is something unique from Samsung, but the technology could create chaos amongst the people as there would be nothing to make people believe if it is real or not.
The researchers at Samsung AI Center in Moscow developed a new technique, which copies the facial data from Source image and implements it on the Target image. Whatever the target image does, the data from Source Image would be imposed on it, making a fake video as we’ve seen already. The animated image is then adjusted using the AI to look like it’s real, and with this AI, the researchers have already animated some older adults like Mona Lisa and others. Samsung researchers have already created the AI, and this version is the improvement to the same. The New AI uses the “single-shot learning,” which is the technology that doesn’t require much data, except the source image.
Unlike older AI, this image won’t need the whole bunch of facial data, and can automatically detect required everything from the same source image. Currently, it only uses the facial data to create head level videos and would take some more time to create AI that is capable of animating the whole body. To make videos more reliable and realistic, it uses “Generative Adversarial Network,” which competes for the two images with each other to eliminate the differences. As of now, Samsung doesn’t know where to use this AI with practical use. If people got access to this AI if it gets public, then there would be absolute chaos with fake videos of people.
Phone: +1 512-845-8160