Art and Beauty in the Age of Machine Intelligence
Does art made by computers have the same aesthetic value as that of a human artist? Is a robot even capable of portraying the creativity or intelligence needed to create great works of art? At first pass these questions seem nonsensical, and probably unimportant. Yet, the continued evolution and integration of technologies like artificial intelligence and robotics draw our attention to the relationship of art, beauty, and technology.
As humans we have a hard time judging the value of artificial intelligence. It is hard to separate what features of the AI are directly related to its code versus knowledge the system has learned over time. For most of Artificial Intelligence’s existence the Turing Test has been looked to measure a machine’s intelligence. A program passes the Turing Test if it is able to trick a human into thinking it is actually a person. While this feat is impressive, some researchers believe this test is inadequate. A program can pass the Turing Test by mimicking basic language skills, rather than using genuine intelligence to converse meaningfully. Therefore, developers can hack a program to pass the Turing Test rather than build a program that genuinely uses its intelligence to converse successfully with a human.
The Lovelace Test has a more unique criteria for a program to pass as intelligent. The Test’s criteria was based on Ada Lovelace’s idea that a machine should not be considered intelligent in the same way as a human, until a machine can originate an idea it was not designed to do. The most important aspect of this evaluation of machine intelligence is that the program can not be manipulated to trick someone attempting to judge its intelligence. The program has to display genuine independent intelligence. What type of novel output it creates is not important. What is important is that the program’s designers must not be able to explain how the original program came to create the new output. There has been no program to officially pass the Lovelace Test, yet.
Since most machine learning programs at the moment are designed to automate tasks or gain insight from a select group of data, we have fewer programs making a pass at the Lovelace Test. One area of machine intelligence that has the capacity to create new and creative works is computerized art. The arts implicitly seem to fall neatly under the umbrella of human creativity. We value art and its aesthetic due to aspects like the artist’s intentionality and the process that governs the works creation.
A recent study researched participants’ biases and perceived value of machine-made art. Since effort of artist is highly important to value of the art piece’s aesthetic, researchers wanted to see if the physical presence of a robot would impact the value of machine-made art. So the research team set up three conditions, or research groups. One group was able to see the robot create a piece of art, and the work itself. Another group saw the artwork and was told it was created by a machine, but did not see the robot at work. Finally, the last group only saw the artwork, and received no information about the origin or process of the artwork.
The researchers found that computer-made art was not given the same aesthetic value as human art. (Even if the visual image is identical). So those who believed the art was made by a computer saw it as less valuable. Interestingly, the team found that this bias could be moderated by interactions with the agents making the art. The study found that the robot artists were not perceived as human-like, but still seen as intelligent to participants. Which is compelling because this means movements and effects programmed into machines to seem anthropomorphic can actually affect the perceived value of the machine’s output. Showing anthropomorphism, or human-like quality, is linked to perceived aesthetic. Therefore, we can increase societal bond and recognized value of machines’ output by making machines, the content creators, more human-like.
A part of this study important to tease out more is that the anthropomorphic qualities do not have to appear on the surface to have effect. Participants were influenced by movements and mechanics of the robot rather than its direct external appearance. The impression of the machine’s behavior led participants to view the robots as having some intelligence. Further reinforcing the need to empathize with a machine in order to better value its output.
As a society we may just be dipping our toes in the water to our evaluation and appreciation of ideas and art created by machines. Even though our current AI systems are still growing, we can already detect some effect a machine’s humanness has on its work’s value and acceptance. This is truly a revolutionary insight. We are beginning to see how our relationship and perception of machine can adapt over time. The more we are able to value the intelligent processes at work in machines the more the value derived from their work can grow. As these technologies advance, so will our perception of them. Hopefully, we can also continue to advance our own views and expand our outlook on novel work made by machines
