The cellist Jan Vogler believed that art makes us human. But what if the machines also begin to create art? Above, for example, you can see that they already know how to create all sorts of artificial intelligence in the creative field. On the right in the image is a computer with AI, which studies in pictures with graffiti. He runs a plotter that sprays water onto concrete blocks (left). The resulting patterns are a form of computer art.
Is it possible to call this art real? If so, we will have to accept that some part of our humanity – that part, about which Vogler spoke – will be mastered by machines. If not, we can be comforted by the fact that although the machines are engaged in art, there is no subtle art in this art.
When art is created to meet the needs of a third party – in this case, the programmer who teaches the computer – is illustrative or commercial art, and not high creativity. For AI to be able to produce a product of high creativity, it must be its own product: autonomously and independently created by a machine forged by its own aesthetics. Only in this case art will cease to be a passive product of human creativity.
Will there ever be a real artificial artist?
On January 8, 2018, an exhibition dedicated to the aesthetics and art of artificial intelligence was held at the Okinawa Institute of Science and Technology (OIST) in Okinawa, Japan. The focus of the exhibition was the concept of true visual art of artificial intelligence. The only problem for the curators of the exhibition was that there is no art of this category yet.
To get around this inconvenient fact, the exhibition stands were divided into four categories: human art and human aesthetics; human art and machine aesthetics; machine art and human aesthetics; machine art and machine aesthetics. In the first category was a collection of ordinary human art, beginning with the Renaissance. In the second and third categories, a collection of hybrid man-machine art is presented. In the fourth category there was no machine art, because it does not yet reflect the machine aesthetics. In this category, the stands were almost empty.
Each category teaches us its lessons. Art in the first category shows the historical transformation of aesthetics from the perspective of God to man. For the most part, the systemic art of the 20th century in the second category includes minimalism, serial music and visual poetry, characterized by the use of rules or mathematical forms. System art can be considered open in 1889 as the Eiffel Tower. Against the construction of the tower were many well-known artists, including the artist William Adolphe Bouguereau and the novelist Guy de Maupassant, because they saw in it a simple appearance and machine design, as if a hideous denial of a person's aesthetics. The fact that the Eiffel Tower attracts many today is the main lesson of the second category: our aesthetic sense can be changed by mathematics and machines. The third category contains so-called media creativity created by machines and artificial and showing that even in the role of a passive product of human creativity, modern AI is able to create objects of beauty.
In the first three categories of the exhibition describe an incomplete arc. We see the birth of a human author and the birth of an AI writer. But will there ever be a truly artificial artist? Can we expect that aesthetics will one day be born in a completely machine world, without the participation of man? This question was in the center of the exhibition: will AI ever have its own aesthetic autonomy?
Plato argued that "true, good and beautiful" have value in themselves. Beauty has value in itself, and not because it serves some other purpose. We do good for ourselves and so on. In order for the machine to create its own visual art, it must satisfy the dictum of Plato and create without utilitarian purpose. The open question is whether cars will ever be able to do this.
One of the reasons for optimism is that people are not the only creatures that can create without use. For example, a chimpanzee can draw for the sake of pleasure. At the exhibition in Okinawa were shown the drawings of five chimpanzees and bonobos belonging to Kyoto University professor Tetsuro Matsuzawa, and all of them are classified in the fourth category, as if reminding us that in this world, pure creativity without human participation is possible. If animals drew for bananas, they would not be included in this category, because then the art would not be created for the sake of an end in itself.
For AI to catch up with chimpanzees, they need to overcome two steps. First, AI must learn to set their own goals. The goals of modern AI are determined by people who write code evaluating how well or badly the algorithm works. The first object of machine art, which can be introduced into the fourth category, will require its own evaluation functions.
This is not only possible, but has already been achieved. In fact, if you visited the exhibition in Okinawa, you would see it with your own eyes. Kenji Doya, professor of the neural computing unit at OIST and his team, conducted an experiment called "Can robots find their own goals?" They put a collection of robots made from smartphones on the wheels. Robots could roam freely, find their own places for charging and exchange programs, scanning other people's QR codes. Charging was the analogue of a meal, and the exchange of programs is an analogue of reproduction. Robots that did not charge, stopped working, and those that did not exchange programs, did not transfer their "DNA" to the next generation. Over time, the robots began to define their own goals: some stopped charging to pursue other robots, for example. This behavior was not programmed. These results convinced Doya that robots can set their own goals.
When AI starts to create high art, can we understand this?
The second step, which is necessary for AI to create, is to develop secondary goals – which exist only to serve primary goals – in the primary goals themselves. Suppose, for example, that the primary goal of an organism or machine is reproduction. Sex – one of the methods of reproduction, so sex will be a sub-goal. To have sex, you need to attract a partner, and this will be a sub-goal. To attract a partner, you need to be cute, and this will be a sub-podportsellyu and so on. But in people, the sex and beauty of the partner in themselves have become valuable. As sex for sex has become valuable, so art for art's sake. When the AI determines its own goals, and then begins to pursue them at its own discretion, it will be on the path to creating its own high creativity.
How do we know that AI has become a real artist? We can teach AI our own art history to encourage the creation of an output that we will learn and understand. On the other hand, untrained AI will most likely create something completely original or even unrecognizable.
Real creativity for AI will be simultaneously painfully boring and very stimulating, and actually so will progress be seen. Beauty, in the end, can not be quantified, and the search for the definition of aesthetics itself moves art forward. The implementation of AI will bring new dimensions to these issues. And then comes the triumph of materialism, which further destroys the special specificity of the human race and reveals a world in which there is no secret or God in which people are simply machines. If so, we will see a new generation of artists, and with them the new Eiffel Tower beyond our wildest imagination.