Deep Fakes – Possibilities and Dangers of Realistic AI Faces
- 0
- 0
- 0
- 0
- 0
- 0
The term artificial intelligence (AI) is often used for simple machine learning or mimicked intelligence. But there are also really amazing uses for so-called deep fakes.
Also read: Deepfake Technology
Deep fakes with quality control
Then the game starts: One neural network calculates completely new faces from all the possible human features that it has memorized while learning, from face shapes to skin characteristics and eye colors to hairstyles. The other neural network then evaluates the result using its own empirical values. If the fake face passes the “reality check”, it is output as finished. Otherwise the first network will have to be reworked again.
More than 90 percent of the results are so realistic that someone who doesn’t know how the portrait was created would have no doubts about its authenticity. But here and there flawed faces come out again, which is somehow reassuring. Often there is something wrong with the facial symmetry, the peak of a baseball cap grows out of someone’s forehead or there are texture errors on the skin, at transitions or in the background.
Artificial art and realistic ancestors
Other projects bring generated faces to life, for example on the AI portal Altered Qualia, which presents artificial art. There the portraits begin to smile before they are overlaid by the next face (morphs). In addition, the eyes follow and the heads move with the mouse pointer. Even more exciting examples can be found on the main page of the portal.
And on Twitter, users are currently sharing countless AI animations that come from a genealogy portal. The company has recently started offering its customers the opportunity to bring their own uploaded portraits to life, with head and eye movements and here and there with a smile – and it calls it Deep Nostalgia. This works quite well and realistically even with faces from paintings or sculptures.
The dangers of misusing deep fakes
Deep fakes may be a technical achievement. However, like most others, this can also be misused. The images and videos that are most frequently clicked on are often fake or taken out of context. While one is at least a little bit more sensitive to photo montages on the internet, deep fakes are a rather new phenomenon in the video sector. Many do not expect such manipulation to be possible at all. Video deep fakes made headlines, for example, when fraudulent groups faked the faces of Hollywood actresses in pornographic films. The good news: At the moment, video fakes are not that mature and are often easy to spot. But that will change in the future. So stay alert.
GO TO COMMENTS (0)
