Artificial intelligence: can it create better art?
With the feverish pace of technological developments, is it finally time for us to start looking at art with a different perspective?
“We are living in a time of unprecedented complexity; our senses are currently whip-driven by a feverish new pace of technological change. The activities that mark us as human, though, don’t begin, exist in, or end by such a calculus.” –Adrienne Rich
Every time humans created technology, it has found its way to art. From 3D painting to conductive paints, artists have found innovative ways to blend technology with art. With developments in artificial intelligence, we are on the edge of a new frontier that can change the way we see and create art. Scientists and several artists have been exploring computer algorithms to create art and increase machine intelligence and creativity. Alongside these developments, critics are also questioning whether machines are our artistic competitors or if they are enhancing human creativity.
Dr Ahmed Elgammal, the founder and director of the Art and Artificial Intelligence Laboratory at Rutgers, believes that human imagination is constrained by our world, but a machine is competent in exploring possibilities. Without imagination, creativity is elusive. With the use of artificial intelligence in art, we can expand human creativity to kick it up a notch.
Scientists, since decades, have been writing algorithms to create art, called ‘algorithmic art’. In this, the algorithm is written while keeping the output in mind, and all the specifications are mentioned in the algorithm. One of the first attempts at this type of art was by a British artist, Harold Cohen. He created the program, AARON, that could generate artistic images. Cohen did not train his program with images, rather he used object and body elements and taught it the relationship between them. Its capacity improved over the years as programmers taught it more difficult techniques. By the 1980s, it could situate objects in 3-D space and it started painting in color from the 1990s.
Later in 2014, Generative Adversarial Networks (GANs) was developed by Ian Goodfellow, which uses two neural networks instead of one. One is the generator, while the other is discriminator. The discriminator is fed with a mixture of real and fake images that helps it in recognizing patterns. Its job is to determine whether the input image is real or fake. The generator’s job is to generate images and trick the discriminator into believing that it’s real. At the same time, it gets feedback on which images are tricking the discriminator and how it should alter its strategy to trick it better. This competition obliges the generator and the discriminator to become better at their jobs.The generator outputs can be surprisingly surreal.
While GANs have shown to create novel art by simulating the given set of inputs, it was, however, limited in its ability to make creative art. Hence, modifications were introduced in the GANs to maximize deviation from established styles. This new network algorithm, Creative Adversarial networks or CANs, works by learning new art styles and becomes more creative by deviating from the known styles. The discriminator still classifies inputs as real or fake, but in addition to that, it now classifies them into the 25 artistic styles too.
This quality of CANs to generate creative art styles successfully enticed several artists into collaborating with AI to produce astonishing results. Recently, Google’s Magenta collaborated with Deeplocal and The Flaming Lips to create Fruit genie, an intelligent musical instrument and melody creator, that combines Magenta’s piano genie with fruits as a physical interface.
Sougwen Chung, an interdisciplinary artist, explores mark-making by both hand and machine to understand the interactions between humans and computers better. For her current project, ‘Drawing operations’, she uses google’s TensorFlow, an OpenSource software library, to train the software to classify her work. The computer then transfers what it has learnt to a robotic arm which draws alongside her.
“My interest in working with robotics came from my practice of drawing. Working with robotics and drawings brings me back to the body — the mark-made-by-hand, and what things like muscle memory and physical instinct can inform about the creative process.” -Sougwen Chung
While some artists could not keep their hands off this technology, the question whether AI created art can be deemed as art still lingers around. CANs might create art that is novel and innovative but it still lacks one important quality, intent.
All great artists had their own unique style and they created art to express. They personalized their art with layers of deep messages hidden in them. The string that connects the artist with his art is what adds value to it.
Developments in technology have always affected how artists produce art, and will always do. When the film camera was invented, it wasn’t accepted as an artistic tool, but when artists got hold of it, they created marvels. The use of artificial intelligence in art can turn out to be something confounding. However, at the same time, the question of whether it can create better art than humans cannot be neglected.
