Most of us tend to think that art is a purely human privilege. For millennia, it communicated socially important messages, influenced global moods and perceptions of the world, and was almost the main instrument for broadcasting ideas and experiences. But the question of whether only people can create art is rather rhetorical. This statement can be questioned at least because the first artists were Neanderthals, not Homo sapiens, as scientists recently stated; we also should not forget about the existence of animal-made art, which is often valued no less than man-made art. Furthermore, with the advancement of AI and algorithms artists are beginning to increasingly use these tools in the creative process. Figuring out whether these pieces of code are merely new tools or full-fledged artificial creations is the premise of the article.
How did AI art come to be?
AI art or art generated by artificial intelligence is a relatively new form of artwork created using algorithms and neural networks.
AI art was first heard of in the 1960s, that is, when Harold Cohen developed AARON, a computer program that creates an image based on previously provided visual information. It was during this time that artists began to actively explore ways of creating art that did not require human intervention. Several prominent names are worth mentioning here: Vera Molnar, Georg Nos, Frieder Nake, Manfred Mohr; they were the pioneers who created the first code-based art and blazed the trail in the newly formed world of AI art for generations to come. Later, in the 1980s, the artist William Latham, applying the principles of genetics and evolution, taught simulators to independently modify and develop animated sculptural forms.
And although the practice of using algorithms and computation in art dates back to the middle of the last century, the real breakthrough in the field of AI art has occurred over the past few years. It was caused by the latest innovations in the field of machine learning, in particular the Generative Adversarial Nets (GAN) that can generate and evaluate images. Equally important in the history of AI art was the emergence of the DeepDream program, which uses a convolutional neural network to enhance patterns in images, simultaneously creating a hallucinogenic, a little eerie effect.
An example of an image created in the DeepDream program
How is AI art created?
Having chosen the necessary computer program, the artist writes detailed code in it, taking into account the wishful visual result. Artists do not create algorithms in order to dictate certain rules to the program: by adding an unlimited number of images already created by someone, the author only introduces the program to the aesthetics that they want to see in the final generated image. After analyzing the visual information provided to it, the algorithm produces a number of initial images, leaving its developer the opportunity to weed out unnecessary ones and stop at the one that they consider necessary.
How to create an NFT art on the market where the rules have not yet been set – in our article
While some artists meticulously make catalogs of images to convey the desired aesthetic to the algorithm as accurately as possible, others choose not to do it at all, giving the AI algorithm more freedom. For example, the developers of AICAN (AI Creative Adversarial Network) decided not to introduce the algorithm to specific images, but offered it a few tens of thousands of artworks, representing the Western Canon of the previous five hundred years. This is similar to how art history students are taught.
So, at this stage everything looks simple, but then problems arise. There is a lot of controversy about whose contribution is more important – a person or an AI algorithm. On the one hand, a person independently writes an algorithm, starting from their own creative idea, forms a catalog of images that introduces to this algorithm and controls the process of generating a work at subsequent stages. But on the other hand, the end result strongly depends on the AI algorithm; moreover, it is the algorithm that is responsible for the creative component, generating new images and artistic solutions, when the developer acts as a curator, choosing the necessary information and controlling the process. Now it is difficult to say unequivocally who to a greater extent controls the creation of an AI work – the discussions are still ongoing.
“In my view, AI art is still exclusively human art. It could be perceived as a very advanced tool, allowing us to make inferences from vast amounts of data and generalizing them in a way that exceeds human brain capabilities. Hypothetically, if a huge watercolor painting 100km by 100km wide was painted, using robot hands, programmed by an artist – would it be exclusively human art? In my opinion, yes, as the process itself is defined, curated and finished by human. If an artist is not using his fingers it does not mean he is not creating it. On the other hand, not having full control of the medium and working with unpredictable output due to uncontrollable water flow also does not mean it is human art. I would suggest that these parallels imply the importance of human in the loop in AI art. Without the autonomy of artificial processes it is still early for humans to be excluded,” says AI artist and photographer Ivona Tau.
AI art – an innovation or a return to old methods?
Let’s go back to the Neanderthals. In fact, the process for them to get the skill of creating cave paintings is very similar to the principle by which we taught machines to generate artworks. However, neither one nor the other would have been able to do this without cognitive abilities (memory, pattern recognition, learning ability, thinking), which were passed on to primitive people by their ancestors, and to computer programs by ourselves.
It is worth making another analogy with photography, which has traveled a similar path, at first acting only as a mechanical reproduction tool, then imitating traditional forms of art, such as painting, and finally turning into an independent artistic direction.
AI can be called a simulator, which we taught to recognize our own visual images. AI algorithms don’t need an artistic background – they just get to know a certain aesthetic and then transfer it to other images by combining styles.
So, AI art is not such a new phenomenon as it might seem at first glance, therefore it is difficult to call it revolutionary from a historical and artistic point of view. Perhaps we will better understand the nature and specifics of AI art if we turn to early examples of computer art, with which it is directly related, because computer artists also worked with a large number of images and understood the creation of algorithmic art to create generative aesthetics.
Prejudice to AI art
Can AI art be considered real art? Does the truth of an artwork determine a person’s involvement in its creation? What is the face of the artist’s intervention in the generation of AI works? Such questions can often be heard from many skeptics who give AI art the status of “non-art” even when the works generated by the algorithm make a positive first impression on them. According to surveys, people who have previously rated a particular art artifact highly tend to change their minds when they learn that it was generated by the AI algorithm. This happens for several reasons.
We tend to sacralize the process of creating an artistic artifact, because it is difficult for us to explain how and why creative ideas are born. But AI art is not associated with such ephemeral things as muse, inspiration or enlightenment, so it destroys the magic that envelops the creativity and personality of the artist, which we obviously do not like. But the fact that we do not fully understand how creativity emerges does not mean that there is no explanation. Everything happens for a reason: both behind the work of traditional art and behind the AI work, there is a certain historical and cultural scheme; in fact, it is the result of our visual experience and the ability to interpret what we have once seen.
Our reluctance to put AI art alongside other, more traditional areas is also due to the fact that algorithms do not create an artistic artifact, guided by a desire to to express feelings. They create artworks mechanically; algorithms generate aesthetically appealing images, but cannot endow them with semantics. And if in almost every work of traditional art there is a certain idea, then behind the work of AI, even the most ingenious, there is often a simple coincidence.
Algorithms can answer the question “how to do it?”, but they cannot say why they do it. Unlike humans, algorithms interact with a limited number of visual images; they cannot reflect on current topics and relate their works to broader contexts. Although, on the other hand, is not this the main task of the developer of the algorithm?
But it is not necessary to strongly distinguish between traditional and AI art – the reasons for which a person indignantly calls an artwork “non-art” are often the same for both spheres. For example, in traditional painting, as well as in AI art, there are low-quality works and authors who do not have artistic abilities. However, such cases more often make people question the artistic value of only a particular artifact, and not the whole sphere, because it still has a large number of artists who create skillful works and place them in the right contexts.
In general, many insist that the value of AI art lies not so much in the final image, but in the very existence of such a phenomenon and form of interaction between man and artificial intelligence in the field of art. The process of creating an algorithm itself becomes a kind of art form.
Prospects of AI art
Many people are already humanizing artificial intelligence, so it is not surprising that most predictions suggest the transformation of AI into an autonomous artist. This is likely to be a matter of the next few years.
The most futuristic scenario – with the advent of technological singularity, artificial intelligence will begin to create emotionally, it will create when it wants. Machines will stop imitating human art and create their own artistic language, which we may not understand. But if we understand, it will radically change our perception of art.
A more plausible scenario which is already becoming a reality – AI will become an independent artist. For example, there are more and more programs that significantly limit human intervention in the creation of AI images, bringing the algorithm to the fore. Instead of using AI solely as a tool, artists begin to treat it as a co-author, creating new ways of synergizing the human and the computer.
It became clear that AI is not omnipotent either – it is in an isolated creative space and only people are able to integrate AI art into the modern art space. The fact that artificial intelligence can create artworks independently does not mean that it will replace artists – it will only create a new form of collaboration, which is likely to bear much interesting fruit.
Here is what the new media artist, curator, and assistant Arts Professor in the Department of Photography and Imaging at New York University Tisch School of the Arts Snow Yunxue Fu says: “I think technologies like AI, VR, and other imaging tools open a lot of new possibilities for artmaking and they are becoming more and more accessible for people to use for artmaking. There could be a complicated relationship between AI art and the human factor. If the human is the one to create the code for AI for the purpose of making artwork, the human could be considered the original artist for any art the AI technology would make. However, if the AI then created their own codes and artworks outside of the original design, then I think we go into a bit of a muddled area. There perhaps also need to be a more in-depth definition of the concept of “AI” and even “Art” to better answer the question. Throughout the technological change in history, many general concepts get expanded while still having a deep relationship to the more primal approaches”.
For a better immersion in the topic, we offer a resource AIArtists.org, which covers various aspects of AI art, as well as the portal Ai gAng and a special issue of AI magazine “Computational Creativity: Coming of Age” (2009).