Samsung’s deepfake artificial intelligence turns Mona Lisa into a talking, “living” work of art
So-called ‘deepfake’ A.I. turns artworks, photographs into real persons
When more than a year ago ‘deepfake’, algorithm-generated porn movies, which combined actor bodies with faces of famous actors or other celebrities, became big news, the technological process that made this possible was still not fully developed. You still had to use existing moving images in the process, to come up with convincing ‘fakes.’
Now, news coming from Samsung A.I. labs located in Moscow, suggests that this process has grown leaps and bounds since then. According to a paper that institution published, they were able to convincingly turn Da Vinci’s Mona Lisa into a living person that moves, talks, etc. with just the use of a single still picture.
Samsung’s researchers performed the same process on the images of Marilyn Monroe and Albert Einstein. By a process that is called “puppeteering”, they ‘mapped’ the moving facial features of other people onto their images. As the researchers explained the process can be done with just one image but the effect and realism can be even better recreated with the use of more images.
As British Telegraph points out, in this age of fake news the effect that could have on political manipulation is immense, particularly since there was already an instance in which a convincing fake video was created when a speech by former US President Barack Obama was ‘mapped’ onto Russian President Vladimir Putin, making it look like he was actually making that speech.