Yaren Kılıç / Middle East Technical University
What is meant by making art often varies from person to person. Even the definition of art creates debate among people, and answers vary. This essay will primarily seek an answer to the question of why people make art; then, it will be discussed whether artificial intelligence can make art today. For this, Alan Turing’s Turing Test and John R. Searle’s Chinese Room Argument will be compared, and then it will be concluded that artificial intelligence cannot make art today based on Searle’s argument because artificial intelligence does not do something by understanding, feeling, or thinking, but by applying certain rules. Finally, the counter-argument to this argument will be determined, and the effect of the audience on making sense of the artwork will be discussed. Along with all this, it will be said that the audience is not in a position to make sense of the work of art.
Asking what art is for is like asking what happiness is for. Every person intuitively knows what happiness is and what it is for but has difficulty defining it. The same is valid for art. Most people do not need to ask what art is and what it is for because they assume they know what it is for. According to Aaron Hertzmann (2018), a professor at the University of Toronto and a BA in computer science and art history, when people make art, they express themselves personally. In other words, art is an act of communication that provides self-expression. That is, art is a social act. If art is a social act, it must be made by social agents who can express themselves. Hertzmann thinks that if art is accepted as a social act, every person can make art. This is because humans are social agents. Anything from a child’s drawing that only his family likes to a painting created by the most famous painter can be considered art. Hertzmann then mentions Charles Darwin when discussing what art is and what it is for. According to him, making art is an adaptive product of human biological evolution. It is seen that art has many functions from past to present. For example, when we look at the first ages, art had a function that determined the sexual harmony of people and, as a result, ensured mating. In addition, art has also served to hold communities and societies together. It provided sharing and unity among people. For example, storytelling, dances, and music brought people together. Moreover, storytelling made it easier to tell and convey difficult things. In that case, it can be said that art provides communication and display between people. Besides being a social act of making art, there are also non-social benefits. For some, only doing art for itself can be calming, meditative, enjoyable, and can be seen as a passionate experience.
On the other hand, making art requires a willingness to express one’s feelings, thoughts, or anything, and creativity, growth, and responsiveness. In this case, artificial intelligence must have this skill and exhibit social behaviors in order to make art (Hertzmann 2018, pp. 1-3). It is understood from Hertzmann’s ideas that making art requires certain features. In order for art to exist, it must have mental characteristics such as understanding, feeling, thinking, and empathizing. So, in order for the machine or artificial intelligence to make art, it must have these features. So, does artificial intelligence have these features? If not, can it?
When it comes to artificial intelligence, the arguments of renowned mathematician, computer scientist, and philosopher Alan Turing are very crucial. In order to understand Turing’s idea, computational functionalism is also important because even if he did not become interested in the status of mental states, his arguments could be evaluated in computational functionalism and, therefore, can be stated that according to Turing, mental states are important because it can be deduced that Turing considers mental states according to their function with the Turing machine. According to computational functionalism, mental states, that is, functional states of an organization, are explained by causal relationships that respond to other states. In other words, they produce output in return, according to some calculations. For example, when one is in bodily pain, the response, such as crying, getting angry, etc., is an outcome of stimuli, that is, the pain suffered (Bailey 2014, pp.150-157). It can be said that Turing took thinking in a functionalist way, questioning whether machines can think even if he did not identify himself as a functionalist. He claims that when human behaviors and emotions are considered in this operational way, artificial intelligence can think and act like a human. He presents the thought experiment known as the Turing test to prove this claim. When it is considered human mind, language is the most powerful tool in expressing human feelings and thoughts. Therefore, Turing uses language in this test. According to this test, a computer and a human exist in the same environment, but this computer and human are in places where they cannot see each other. There is also an interrogator besides the computer and the human. By asking questions to the computer and the human, this interrogator will decide which is the real human and which is the computer. The responses of both are projected onto a screen. The computer and the human try to convince the interrogator that they are human with the questions asked. If the interrogator cannot determine who is human and who is a computer after asking questions, the Turing test will be successful because the computer does not reveal itself. According to Turing, the machine can imitate all human behavior, and thus it can be argued that the machine can think, feel and understand. What he means by thinking is that there are inputs and outputs as a result of operations performed with these inputs. In this way, all mental abilities and situations can also be adapted to a machine. The machine receives certain stimuli and generates a response to them, and by doing this many times, it can convince the questioner that it is not a machine. According to him, the important thing is that the machine produces certain outputs suitable for the input, and this means having mental abilities. If the machine contradicts a human as to whether he is a machine or a real human, it will be possible to say that the machine thinks and has cognitive abilities (Turing 1950, pp. 433-442).
Given Turing’s argument, it seems possible to argue that a machine or artificial intelligence can make art because art requires some skills such as feeling, thinking, and understanding, and the machine can do each of them. It can be taught how to compose a piece of music or draw a picture because what the machine has to do is generate appropriate responses to certain stimuli. In fact, it can be programmed to make mistakes and sometimes act spontaneously when necessary. From this point of view, it does not seem far away that the machine can make art. With the Turing test, just like a human being, a machine can know how to compose a piece of music, draw a picture, make mistakes, act spontaneously, and have a worldview. Moreover, it can feel, understand, and have creativity. However, is Turing’s machine really a machine with cognitive abilities? There has to be a difference between pretending and actually doing something.
At this point, the arguments of the famous philosopher John R. Searle are essential to understanding the difference between imitation and reality. Searle makes a distinction between weak AI and strong AI. Weak Artificial Intelligence is a machine that helps to understand the nature of the mind. For example, if there are some hypotheses about the nature of the mind, they can be tested using computers that process some statistics for these hypotheses. Strong Artificial Intelligence is a machine with the same cognitive abilities as humans. Searle sees no problem in the presence of weak AI, but according to him, there is a problem with strong AI. He denies that machines can think and feel just like humans. To prove this denial, he puts forward a thought experiment called the Chinese room. According to this experiment, an interrogator will test whether a person can speak Chinese. This person does not know how to speak Chinese, but he has a rule book, and this person is locked in a room full of Chinese symbols. Chinese questions are asked to this person from the outside, and this person simply looks at the rule book and puts together the Chinese symbols, and exports the questions in this way. After a while, this person becomes very adept at using the rule book, and his answers to the questions on the outside are correct. In this way, the questioner on the outside is convinced that this person knows Chinese because the person who does not speak Chinese can convince the questioner using the rule book and make them believe that this person knows Chinese. In this case, Searle says, this person does not actually speak Chinese even if they convince the questioner because cognitive abilities like thinking and understanding are not just about following certain rules. In other words, there is a difference between syntax and semantics, and machines only follow the syntax. What computers do is follow formal and syntactic structures, but the mind, besides syntax, also contains semantics. They follow the rules, but they do not know what they mean. Considered from Searl’s point of view, the distinction between imitation and reality is drawn. There is a difference between really thinking, feeling, understanding, and imitating them. The machine may unconsciously follow the rules to convince humans that it is a human, but what humans do is conscious. In this sense, the machine will not come close to being human unless it is conscious (1980, pp. 417-424).
Considering Searle’s argument, artificial intelligence making art will not go beyond imitating human art. In order to make art, the machine will follow certain rules imposed on it and produce a suitable output according to certain stimuli. However, he will not truly have the mental abilities that art requires, such as creativity, desire, feeling, and understanding. The machine can make music, dances, write poetry and perhaps prove itself in many other arts. Nevertheless, while human art is intertwined with emotions, thoughts, meanings, the art of the machine cannot go beyond following a rulebook. Unless the machine’s understandings are like the human’s, what it will do is to imitate art, like all other human actions, rather than art. In this case, machines will not go beyond the tools that people can use while making art.
As an argument against this claim, it can be argued that even if the art of the machine is not art for the machine, it can be an art for the person perceiving what the machine does. Even if the artist machine does not have some human abilities in real terms, people who perceive the work they produce can see this product as a work of art and enjoy it. In this case, it does not matter if the machine has a real thought about the product it produces, whether it made this product by using its understanding or feeling. Instead, a product can be considered a work of art if it evokes certain emotions, meanings, and thoughts in the viewer. At this point, it is necessary to reconsider the question of what art is for. At first, art was defined as self-expression, an act of communication. Herein, art existed to describe, transform, or display emotions, thoughts, feelings, or anything around society or people themselves. However, is it the artist or the art audience that gives meaning to art?
Let us call the personal expression, feeling, meaning, pleasure, and thinking that is used in the creation of art the necessities of art. So, for whom are these requirements essential, for the audience or for the artist? If the artist obtains these experiences through the experience of making art and they do not worry about the audience, this experience will be an inner experience. If the artist is just concerned that the audience understands and enjoys these requirements, this experience will be functional. It is understandable that the artist presents their work to the audience and expects appreciation. However, at this point, it would not be correct to claim that art is for the audience. For example, the famous writer Franz Kafka did not want his works published, and his works were published after his death thanks to a friend. However, the fact that Kafka does not want his works published does not mean that the art he produces is worthless or meaningless. The artist himself experiences art and creates meaning. Moreover, statues made in ancient times were only buried with the dead and not exhibited. This shows that a work of art can exist without being presented to the audience’s discretion. Rather than a factor that determines the meaning of art, the audience is a second phenomenon that necessarily arises with the existence of art. An example of this is kinship relations. An uncle does not have to do anything to be an uncle. When a brother necessarily has a child, that person becomes an uncle. This person does not need to be an uncle for the meaning of his existence but has to become an uncle as a result of existence. Therefore, what gives meaning to art is not the audience but the art producer themselves (Zangwill 1999, pp. 316-320).
When we turn back to artificial intelligence from here, it is seen that the art of artificial intelligence is far from being an art. Even if the audience sees the product of artificial intelligence as art, this does not mean that it is art. As a result of the product presented, the taste of the audience is not enough to determine the artistic value of this product. What gives meaning to the work of art is the artist himself. In the case of artificial intelligence, there is no artistic meaning or value since there is no mention of a work of art and an artist.
As a result, in this essay, first of all, the question of what art is for was examined. Then, it was concluded that art is a form of expression as an act of communication and that art has both functional and personal effects in the lives of societies and individuals from past to present. Later, it was claimed that art is the whole of people’s feelings, displays, thoughts, and search for meaning. Along with this claim, it was focused on whether a machine or artificial intelligence could be an artist. In doing so, the arguments of Alan Turing and John R. Searle were compared, and it was argued that Turing’s argument was insufficient to claim that the machine could make art because the machine’s cognitive abilities could not go beyond imitation. Searle’s Chinese room argument was presented as the basis for this claim. This argument showed the difference between understanding something and following something by rule. However, the importance of the ideas of the art audience as an argument against the inability of the machine or artificial intelligence to make art was discussed. At this point, it has been said that it is not the art viewer who gives meaning to the artwork. The reason for this was determined as the existence of the art audience as a result of the work of art. Therefore, it will not be possible for artificial intelligence or a machine to make art unless it has real consciousness.
Bailey, A. (Ed.). (2014). Philosophy of mind: The key thinkers. A&C Black.
Hertzmann, A. (2018, June). Can computers create art?. In Arts (Vol. 7, No. 2, p. 18). Multidisciplinary Digital Publishing Institute.
Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–424. https://doi.org/10.1017/S0140525X00005756
Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59, 433–460. https://doi.org/10.1093/mind/LIX.236.433
Zangwill, N. (1999). Art and Audience. The Journal of Aesthetics and Art Criticism, 57(3), 315-332. doi:10.2307/432197