AI and Art II: The CreatorShare
A look at AI and art within the context of theories that focus on the artist.
Many theories of art include the creator of the work as a critical part of sorting art from non-art. An excellent example of this is Leo Tolstoy’s theory of art. As he sees it, the creation of art requires two steps. First, the creator must evoke in their self a feeling they have once experienced. Second, by various external means (movement, colors, sounds, words, etc.) the creator must transmit that feeling to others so that can be infected by them. While there is more to the theory, such as ruling out directly causing feelings (like punching someone in anger that makes them angry in turn), this is the key part for deciding whether AI created works can be art. Given Tolstoy’s theory, if an AI cannot feel an emotion, then it cannot create art. This is because it cannot evoke a feeling it has experienced nor can it infect others with that feeling. For it does not feel. However, if an AI could feel emotion, then it could create art—under Tolstoy’s definition. The known AI systems of today clearly do not feel, hence the works they create would not be art. But what about the future?
While the focus of research is on artificial intelligence, there is also interest in artificial emotions—or at least the appearance of emotions. As such, there could be an AI system that produces music, poems or paintings and seems to express emotion. In the context of Tolstoy’s theory, the key question would be whether it feels emotion or merely appears to feel. Interestingly, the same question also arises for human artists for this is a variant of the problem of other minds. This is the philosophical problem of determining whether other beings think. In this variant, the problem of other hearts, the challenge is determining whether other beings feel.
Tests have already been created for discerning intelligence such as Descartes’ language test and the famous Turing Test. While it might be objected that a being could pass such tests by faking intelligence, the obvious reply is that faking intelligence so well would require intelligence. To use an obvious analogy, if I could “fake” successfully performing complex repairs on various vehicles over and over it would be odd to say that I was faking—in what way would my fakery differ from having skill? The same would apply to intelligence. As such, theories of art that are based on intelligence rather than emotion would allow for a means to test whether an AI could produce art.
Testing for emotions is rather more challenging since a being could adequately fake emotions by knowing how to respond without feeling them. There are, after all, humans who do this. Some are actors, some are sociopaths. Presumably, some are both. As such, testing for emotion (as opposed to testing for responses) is rather problematic. Because of this, if Tolstoy’s theory or other emotional based theory is used to define art, then it seems impossible to know whether a work created by an AI would be art. In fact, it is worse than that.
Since the problem of other hearts and minds applies to humans, any theory of art that requires knowing what the artist felt (or thought) leaves us forever guessing—it seems impossible to know if the artist was feeling a specific feeling or feeling at all. It is possible to take a more practical approach and make guesses about what an artist might have been feeling and whether this is what the work is conveying, and this weak standard would certainly make it easier to regard AI created works as art.
Critics of Tolstoy have made the obvious criticism that artists can create works that seem to be art without meeting his requirements. That is, an artist might have felt a different emotion from what the work seems to convey. For example, a depressed and suicidal musician might write a happy and upbeat song affirming the joy of life. Or the artist might have created the work without any particular emotion in their heart. Because of these and many other reasons, Tolstoy’s theory does not seem to offer the true and ultimate account of art. That said, he does provide an excellent starting point for a general theory of AI and art in the context of defining art in terms of the artist. While the devil lies in the details, any artist focused theory of art can be addressed in the following manner.
If an AI can have the qualities an artist must have to create art, then an AI could create art. The challenge is sorting out what these qualities must be and determining if an AI has or even can have them. If an AI cannot have the qualities an artist must have to create art, then it cannot be an artist and cannot create art. As such, there is a straightforward template for applying artist focused theories of art to AI works. But, as noted in the previous essay, this just allows one to know what the theory says about the work—the question always remains as to whether the theory is the one true view. In the next essay I will look at work focused approaches to art.
Read also AI and Art I: The Setup