DEEPFAKE
What is Deepfake?
Deepfake is a technique that utilizes Artificial Intelligence. It alters the original video of someone and turn them into a different person. By doing so it can appear like someone is saying things or doing things that they actually aren’t, but rather is another person doing it. This is very cool but also can be very dangerous depending on what is used for.
History of Deepfake
Deepfake is named after a Reddit user “deepfakes”. The user posted many pornographic videos of celebrity faces on the bodies of the pornographic actors/actresses. Deepfake really went mainstream in December of 2017 when Samantha Cole published an article in Vice drawing attention to deepfake. Cole had interviewed deepfakes (the Reddit user) and he stated that he used “open-source machine learning tools like TensorFlow, which Google makes freely available to researchers, graduate students, and anyone with an interest in machine learning.”
In January of 2018 a desktop application was launched called FakeApp. It allowed users to create deepfake videos of themselves or celebrities as long as they had a video of the those two people. The app also used Google’s AI framework, TensorFlow. Now there are multiple open source applications for use.
How does Deepfake work?
Deepfake relies on a technique know as GANs, generative adversarial networks. It was invented by Ian Goodfellow, who now works at Apple. There are two separate AIs involved in the GAN algorithm. One AI forms the content, the videos of the people, and the other tries to guess whether the videos are fake or real. Starting off the first AI have no idea how the two people look, but the second AI can distinguish them easily. Then over time the first AI generates content that looks more and more similar until the second AI can’t tell the difference.
Conclusion
Deepfake videos are getting by the day. Many deepfake vidoes are distinguishable, you can tell which ones are fake and which ones are real. But there are some where you can’t telling which one was the original picture with just the naked eye. It is scary to think how deepfake will continue to develop and what level will they reach. Currently the next step is to develop full body deepfakes. Right now full body deepfakes aren’t real enough to fool the naked eye, but one day it will be. Who knows what will happen then?
