Try it yourself to find the differences by looking at the 10 images you see in the gallery above. Do you notice anything different than usual? Really hard to do: the protagonists are identical to real cats. Why not, they are not real cats. The images in question, technically, aren’t even real photos: they are produced by artificial intelligence, therefore fake.
It is the latest in a long list of examples of how technology can reproduce a world that apparently, at least through a screen, does not it has nothing different from the real one. To prove it is a new site on which you can stay for hours wondering how it manages to achieve such realistic results, on which it is possible to see a potentially infinite number of fake cats with just one click: is called This Cat Does Not Exist, which means, not surprisingly, “these cats do not exist”.
It was born from an idea of two Australian guys, Stephen Mott e Nathan Glover (the same that is behind Selfie2anime, another fun site that allows you to find out what we would look like if we were a Japanese cartoon), who have decided to show the latest frontier of GAN applications. Gan stands for Generative Adversarial Network, that is, an adversary generative network: it is an automatic learning system capable of reproducing images from scratch by accessing the immense wealth of information relating to their protagonists present on the network. They can be images of any kind, of course human beings too, but Glover and Mott focused on cats. The goal, as they write on the site, was precisely “Try to explain in a practical and fun way to everyone” the possibilities of this system. What better way than the most clicked photos on the net?
As has happened with many deepfakes, there is a way to unmask fake cats: as he pointed out The Verge, generally have a fuzzy fur outline, poorly defined eye coloration, and ears are sometimes disproportionate. What will happen if instead of filling the internet with real photos of cats we are satisfied with fake ones? Browse the gallery