and other books including Einstein, Picasso: Space, Time, and the Beauty That Causes Havoc.Miller applied the “Page 99 Test” to his new book, The Artist in the Machine: The World of AI-Powered Creativity, and reported the following:
From page 99:Visit Arthur I. Miller's website.
“Translating one image into another … is like translating between languages, like between English and French. They are two different representations of the same world,” says Phillip Isola.My personal litmus test for reading a book is to be gripped by the first sentence in the first substantive section. If I had to make this decision on the basis of page 99 of The Artist in the Machine I would be tempted to put it back on the shelf. While what is on this page is highly interesting – almost magical – the reader will have to have gone through preceding chapters to know, for example, what a GAN (Generative Adversarial Network) and style transfer are. Opening up to page 99 could give the prospective reader the impression that this is a book meant for a computer scientist. Not so. I wrote The Artist in the Machine for the educated layperson interested in learning about the up-to-date ways in which AIs (artificial intelligences) can amazingly create art, literature and music of a sort that we cannot imagine. My book focuses on concepts and uses no mathematics. So, the page 99 test does not work for my book.
Isola and his coworkers invented a variation on GANs that he calls conditional generative adversarial networks (CGANs). They are conditional because instead of starting the generator network (G) from noise, from nothing, they condition it by using an actual image. Rather than feeding the discriminator network (D) on huge caches of images, they use pairs of images, such as a black-and-white image of a scene and the same scene in color. Then, they input a new black-and-white scene into the generator network. Initially D rejects the new scene, so G colorizes it. In other words, the output is conditioned by the input, which is what GANs are all about. As a result, Pix2Pix, as Isola calls his system, requires a much smaller set of training data than other supervised learning algorithms.
Thus Isola discovered how to translate an image of one sort into another sort: Pix2Pix, pixels to pixels. As he puts it, all those “little problems in computer vision were just mapping of pixels to pixels.” While style transfer transfers the style of one image onto another, creating an image “in the style of” a painting by Picasso, for example, Pix2Pix goes further. Like Leon Gatys, who invented style transfer, Isola is interested in perception, how we see.
Those aficionados of AI who open to page 99, will be struck by how an artificial network machine, trained on pairs of images, one in black and white and the other in colour, can colorize an image. The process is called Pix2Pix. Artists use it to create highly interesting works. Examples are in The Artist in the Machine. Pix2Pix also shows how a machine can take on its own creativity to produce an original work of art.
For some decades I have written about creativity in humans and touched on creativity in machines. In this book I have addressed both with focus on machines. I discuss how my own theory of creativity can apply to machines which, in the future, will create art, literature and music with emotions and consciousness. These works will interest us and their brethren too and perhaps will be even better than we can produce. In this way humans and machines can bootstrap each other’s creativity.
--Marshal Zeringue
Posts
