The Deepfake videos such as Obama’s public service announcement video, Nancy Pelosi slowed down video, Donald Trump lectures Belgium video, Yang Mi travels in time video, Salvador Dalí comes back to life video should not be so worrying if taken as amusement, as much if taken as threat.
Deepfake technology is the combination of deep learning, which is a subset of machine learning and Artificial intelligence.
Artificial intelligence captures how a source face looks like at different angles in order to transpose the face onto a target, usually politicians and celebrities; the reason is simple, these politicians and celebrities have heavy social media presence and their public profiles provides ample source material for AI to learn from, but with the amount of selfies the average person takes in a lifetime and technological advances, perhaps soon anyone could be used as a source, as if it were a mask.
Deep learning is a branch of machine learning that applies neural net simulation to massive data sets, to create a fake.
Machine learning is a subset of AI.
To further simplify these terms, let’s take the example of different human actions and how these technologies can be mapped to these.
Understand Artificial Intelligence as a technology, that will simulate or copy how you look like. Understand Machine Learning as a technology that will use statistical technique on your personal data to come up with your fake behavior like , how you speak, how is your body movement, how you smile.
Understand Deep Learning as technology that will use the neural networks to behave completely like you, or say, will be 100% replica of yours, in terms of actions that you perform at a particular situation, steps you take after an incident etc. In Deep Learning the machine itself trains to perform task, like speech and image recognition.
These are the examples of why the potential misuse of Deepfake technology has been so feared in politics and celebrates alike. Sometimes, small proctored video like the original video can be slowed down and the pitch can be altered to make it seem like the original video is a fake video.
FakeApp and DeepFaceLab alike software tools offers interesting possibilities that range from dubbing, improving and repairing video to solving the uncanny valley effect in video games, avoiding actors having to repeat fluffed line, the creation of apps that allow us to try on clothes or hairstyles, and even to train doctors, but many fear it being used for nefarious ends. These are the fakes that have scared (and amused) people the most so far.
Watch below related video on Deepfake technology.