Related Articles
Samsung's research centre at Moscow has successfully developed a breakthrough in what we can call deep fake technology. The AI created by the South Korean technology giant can create talking avatars from single pictures and paintings without using any 3D modelling or a bunch of pictures. Samsung dubbed this development as "realistic neural talking heads".
Deep fake is a technology by which a talking head model can be created of anyone by using a large number of their photos and videos, but Samsung has achieved the feat with even one still image. So if available to the general public someone can create a deep fake video of you by simply having access to your Facebook profile photo.
Software for creating deep fakes or fabricated clips that make people appear to do or say things traditionally requires a large data including sets of images from different angles in order to create a realistic deep fake. With Samsung's new artificial intelligence system, a fake clip can be generated by just one photo.
Samsung published the paper about this new technology developed in the company's Moscow AI lab, the paper contains several proof of work that were developed by just a single photo, including the classic portrait of Mona Lisa which exists as a single image. It is animated in three different clips to demonstrate the new technology.
The technology can be used for media, fun or educational purposes, but the downside of it can also not be ignored. This kind of technique increases the risk of misinformation, fraud, election tampering, revenge and bringing bad name to someone. Experts say that these are photorealistic digital puppets and there must be a way to detect a deep fake.
