According to Stephen Wilson, Sean Cubitt, and Roger F. Malina, there are three ways artists employ scientific methods to create work, although there is no clear cut boundary between them. The first is to continue a modernist practice of art with contemporary adjustments. The second is to develop a unique postmodernist art built around deconstruction at its core. The third is to develop a practice focused on elaborating the possibilities of new technology. Under this framework, we can identity artist Natalie Jeremijenko’s practice as a modernist approach to poetically manifest the beauty of nature through the use of technologies, which remain invisible in most of her works. Scientific methods, such as using light to attract moth, are means to protect the moths from the industrial urban environment and to create cinematic effect. Heather Dewey-Hagborg’s Stranger Visions both critiques the invasion of technology into privacy and spookily visualize the vulnerability of individual in face of technological hegemony, straddling a postmodernist deconstruction of technology and a creative imagination of technology. The use of AI could span across all categories, although most practices cluster around the first modernist practice, be it composing like classical music masters or painting like Van Gogh.
Can AI replace artists? It is way too early to answer the question yet. One the hand, we still rely on the writing system today to maintain the basic operation of the society, albeit countless prophecies declaring its death whenever a new medium is invented. This historical inertia can not be easily eliminated. Meanwhile, most AI art, created based on GAN or CAN is not creating new artistic ideas but an averaged product of the input dataset based on visual assimilation. Works generated by CAN might dazzle viewers with its visual novelty, but it doesn’t necessarily mean originality or progression. The product is just one of billions of unrealized possible images given birth by computational power. Instead of advancing people’s understanding of art, it merely provides another version of reality, a branch of parallel universe. One the other hand, we should not underestimate AI’s potential in assimilating human intelligence, emotion or even aesthetic, given its rapid development in recent year. A better question to ask, is how can AI assist artists in producing works?
One dimension to think about is control and randomness. There was debate about artist’s hand since conceptual art and generative art, and AI certainly intensified it. As an extension of human intelligence, AI can both control or randomize the resulting artwork, depending on the intention of artists. While the final product is greatly determined by the input database used to train the AI, artists can also select the infinite results generated by AI. In this way, whether the artwork turns out to be great largely depends on how artists exert artistic intention on the final product, and how artists allow AI to be open-ended and explorative enough to generate surprising outcomes. While they may sound paradoxical at first, the two criterions interplay with each other, existing simultaneously when artwork is produced.
Using AI to produce artworks revolutionizes the process of making art. The transformation from laborious handcraft work to intellectual judgement signifies a shifting focus from the process to the end result. Artist Manfred Mohr once said "it is not the necessarily the system or logic of my work I want to present, but the visual invention which results from it. My artistic goal is reached when a finished work can dissociate itself from its logical content and stand convincingly as an independent abstract entity.” It seems like Mohr uses technology as a tool to efficiently generate the image he desires. However, the increased efficiency isn’t necessarily the triumph of art. When process becomes something to be skipped during the creation of art, does the nature of art change? Well, certainly for some kinds of art. Imagine creating a Jackson Pollock with just a few clicks. The evidence of his personal trace is completely erased. Also think of photography generated by AI, as much as they look like real photos, their documentary value is none. By substituting social value with surface aesthetic value, AI could ruin many artwork whose core is a representation of personal emotion/consciousness, experimental actions or specific locality.
How does this change inform curatorial practice? When exhibiting AI generated works, institutions should be aware of its procedural literacy, which is a system of rules that generate the end result. Through admitting the vulnerability of machine vision, curators will demystify the technology and engage audiences from a critical start point. That is not to say AI art is doomed to produce process-irrelevant works. Conversely, artists could also explore the system that forms the basis of technology and be adventurous about the final result, instead of having a rigid expectation. Using AI can be start point, not the destination.