[ad_1]
Deepfake technology can reach a point where a single profile picture on Facebook is material enough for someone to literally put words in your mouth.
Researchers at the Samsung Artificial Intelligence Laboratory in Russia published an article this week describing a technique that would make a portrait appear as if its subject were speaking in a (semi-) realistic video sequence. The possibility of making such clips from photographs is not new, but the methods used for this purpose were previously based on training devices with large sets of image data of the subject in question.
While this technology has many harmless uses in areas such as design and content creation, experts are also concerned that its potential spread of misinformation may herald a future in which the separation of truth from online fiction becomes impossible. Until now, however, the need for many images of the subject in question had limited targets for celebrities, politicians and other personalities photographed intensively in the media.
The creators of Deepfake generally use what is called a generative adversary network, a configuration in which two neural networks are formed on the same set of data images. One then tries to create one's own image that looks like what it has seen, until the other can no longer differentiate between that creation and the actual images.
Instead of training this system on multiple images of the same person, Samsung researchers provided him with a host of head photographs from which he was able to map some of his facial features and facial expressions. This preloading process has fine-tuned the software so that it is agile enough to move closer to anyone's facial movements, from the Mona Lisa to Salvador Dali.
The results are not perfect; A fake clip of Marylin Monroe does not capture his distinctive mole, for example. And there are periodic flashes of distortion in facial movements that can also betray the artificiality of the footage. Yet, it is sometimes realistic enough to leave an occasional spectator who does not pay much attention to the situation.
Brands and agencies have already begun to explore less sinister uses of deepfake technology. Earlier this month, the Dalí Museum in Florida partnered with Goodby, Silverstein & Partners to create a faithful version of the artist to make selfies with visitors. R / GA and the start of the video Synthesia also created a video last month in which football superstar David Beckham seemed to speak nine different languages for a campaign against malaria.
"It's one of those things where technology itself is neither good nor bad; that's what you do with it, "Nick Coronges, global technology manager at R / GA, told Adweek.
[ad_2]
Source link