What we see above is a bunch of different average people right?
The catch is none of them is real…
Everyone knows that AI has been implemented in diverse applications, not only autonomic cars, management systems for factories or chatbots we see everyday. Recently it was implemented to a very specific application, namely creating nonexistent people and their faces. It might sound like a useless whim of bored programmers although it actually might be very useful for marketing agencies or graphic designers as they are royalty-free and available for everyone to use.
But how does that work?
Artificial intelligence has made it easier than ever to produce images that look completely real but are totally fake using Generative Adversarial Networks (GAN), a relatively new concept in Machine Learning, introduced for the first time in 2014.
The essential components of every GAN are two neural networks:
-Generator that synthesizes new samples from scratch, a random vector (noise) so the initial output is also noise.
-Discriminator that takes samples from both the training data and the generator’s output and predicts if they are genuine or counterfeit.
Over time generator, as it receives data from the discriminator, it learns how to create more realistic images. Moreover discriminator also learns and improves by comparing synthesized photos with real images. In other words one network generates a fake face, while another decides if it’s realistic enough by comparing it with photos of actual people. If the test isn’t passed, the face generator tries again. To see yourself how well it works go to: https://thispersondoesnotexist.com.
Every time you refresh the page, you get a newly generated face.
And if you get bored with fake faces you can always admire some AI generated cats: https://thiscatdoesnotexist.com, but in my opinion sometimes it gets quite creepy as it’s not as advanced as the system mentioned above…
-Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, Yoshua Bengio; „Generative Adversarial Nets”
-Tero Karras, Samuli Laine, Timo Aila NVIDIA; „A Style-Based Generator Architecture for Generative Adversarial Networks”