An Opinion Piece on AI Art and Culture: Is AI-generated art less valuable than human art?
Generated through AI text-to-image program Neural Love
Recently, AI art has been the subject of colossal attention; the world was taken by storm as more intricate images produced by AI continue to provoke divisive public opinions. This resulted in the resurfacing of a question that has already been brewing for a while: is AI-generated art less valuable than human art?
There is no doubt that AI art can possess aesthetic qualities – that beauty can be generated through the harmony of shapes and colours, all blended together to magically elicit a feeling of pleasure for the observer. Hume (1757) considered the cause of this relationship between the observer and the observed to be the normative sentiments shared among the living, the artistic ideals of beauty shared by the consensus of men and distinguished by connoisseurs and critics. This means that regardless of whether it is due to essential features of our being or cultural disposition, we find ourselves partial towards certain sentiments associated with observations that claim beauty.
Understood that way, beauty can connote anything from an artwork to something of a natural emergence such as a landscape. Costelloe (2007) considered these sentiments as “ideal evaluators” or general rules, which, when assorted in harmony, construe an aesthetic experience. These sentiments or evaluators can range from simple concepts such as symmetry and colour co-ordination to more complex ones such as cultural allusions and fractal-patterning. The beautiful was considered as the harmony structured according to the rules of sentiment.
That being said, AI art can produce artworks that are considered beautiful. But, to answer the question of value, one ought to consider what AI art does not have rather than what it does have. In order to properly go about this inquiry, it is imperative that we first acknowledge the all-too-human dimension of culture. Culture is, after all, the front where the human expression and the human condition, within a social framework, come to meet.
At first glance, we can assume that AI is impartial and does not relate to a certain culture, but rather is subject to the accumulation of data associated with its parameters. However, upon further reflection, we find that herein lies the problem: the accumulated data itself is subject to biases from the dominant modes of production and knowledge, which, in turn, realizes the generation of seemingly partial or biased products.
Generated through AI text-to-image program Neural Love
AI art-producing tools, such as Midjourney or DALL-E, utilize existing human inputs, alter them according to the prompts (which are also subject to human concepts), and synthesize the appropriate inputs to produce a unique image. Clearly, this means that the images produced by AI can reflect the data’s existing biases. However, one can still argue for the impartiality referred to earlier, if one considers AI’s lack of agency. In other words, we can consider AI impartial if we judge the system itself (its agency, or in this case the lack thereof) rather than the tools and data it employs. At the same time, it is worth noting that the collected human inputs form the datasets, and come from all over the internet, possibly encompassing a wide range of cultures.
Biased or not, what role does culture play in the generated AI art and, by extension, its value? In his social critique, Pierre Bourdieu (1977 & 1979) outlined the importance and degree of influence our social structures and situated cultures have on our aesthetic taste and artistic expression. He also noted how our social surroundings ingrain within us certain habits and dispositions, of which aesthetic taste is of no immunity. He referred to these dispositions as the Habitus. Bourdieu’s theory of the Habitus offers us some interesting insights when it comes to the question of AI art’s value.
Following his work, we find that on one end, the generated images are not an expression linked to a “habitus” or social structural influence and are therefore devoid of the expression of the human experience. This is due to the AI’s utilization of diverse data that are inclusive of more than one singular cultural/social reference point.
So, AI art might not involve habitus which, at least for humans, determines aesthetic taste or value. Interestingly, by virtue of the fact that it’s not informed by cultural habituations, one can argue that the produced art becomes immune to cultural capital. Given the supposed cultural impartiality, mass distribution, accessibility, and appeal of AI art, on an open platform on the internet, any claim to exclusivity goes out the window. This means that with AI art’s lack of a singular cultural reference as well as its wide access, social classes and structures won’t be able to capitalize on it, whether in monetary form or social hierarchical correspondence that one is used to with more traditional forms of art.
Now in terms of value, this means that AI art inherits social justice value yet loses cultural value (in the creative sense). This, in turn, means that the value gained serves an ethical standard, and the value lost is that of cultural expression.
Here we find a divide in the form of the question: In what manner of “value” do we speak?
Let us consider value in the lines between aesthetics and culture. Integral to the discussion would be Kieran & Lopes’s (2004) formulated aesthetic epistemology of a twofold modality: Knowing Through Art and Knowing About Art. They argued that when a subject interacts with a work of art, these are the two types of knowledge infused with the aesthetic experience at play.
The first is knowing through art, that is the awareness of the aesthetic experience in itself, it is gained in virtue of the interaction with work of art. Fundamentally, this kind of knowledge is rooted in the subjective experience associated with the work of art.
The second is knowing about art, that is the cognitive knowledge of a work of art, its cultural and historical contexts, its intended aspirations, as well as its designated style and method. This renders this mode of knowledge as dependent on the successful communication of all fronts with which the work of art is associated.
In most cases, knowledge through art is based on the formal aesthetic properties of the object such as colours, composition, and shapes. Here is where the most basic of aesthetic experiences reside. For example, a simple landscape can instigate the subjective perception of beauty – and AI art does not miss the mark. It is difficult to deny the formal beauty generated by AI, since it is mere replication of the existing formal aesthetic rules and properties.
The crux of the problem, however, resides in the considerations of the nonformal aesthetic properties. This is because we do not know much about AI art, save for that it is generated by AI from variations of existing human made inputs. That is not to say that there dwells mysteries in the art generated by AI but rather, to say that there aren’t any nonformal aesthetic properties beyond that. There are no historical, cultural, socio-economic influences or personal reflections to reflect on for a better experience of the art, nor can we decipher the intentions or aspirations of the artist because there are none. Although the inputs are subject to nonformal references, the AI, being considered as the artist, knows not the human condition, or any condition, with which expression can find a manifestation that elicits nonformal aesthetic properties. The generation of art itself expresses nothing.
Walton (1970) takes it a step further with his analytic postulations for artworks. In his paper, he establishes an analytic outline for what can constitute a work of art where he formulates categories and their logical constituents. With this argument, he argued that the correct categorization of a work is essential for its proper aesthetic perception as a piece of art, and that within this boundary of categorization, nonformal aesthetic properties are indispensable. A person familiar with Cubism may have a better aesthetic appreciation for a Cubist work of art in virtue of his knowledge of nonformal properties pertaining to the artwork in question.
Perhaps, however, things are not as black and white in terms of value as they seem. If we take an honest look at what Walton was saying, we can uncover a hidden nuance within his main argument: AI art does not exactly fit the categorization of normal art. Walton asserts that if a new form of expressive presentation reveals itself (at the time it was collages in contrast to paintings) as a variation from the standard category, and such a form becomes common, it forges a new distinct category. This is where AI art seems to be. A whole new category of art with its own standards.
So perhaps the question framed to compare the values of AI art and human made art misses the same mark as the question framed to compare the values of painting and collages.
Hume, David. Of the standard of taste. 1711-1776.
Costelloe, Timothy M. Aesthetics and morals in the philosophy of David Hume. Vol. 6. Routledge, 2007: 2-21
Bourdieu, Pierre. Outline of a Theory of Practice. Cambridge University Press, 1977
Bourdieu, Pierre. La distinction: Critique sociale du jugement (Le Sens commun) (French Edition) (0 ed.). Éditions de Minuit, 1979
Kieran, Matthew, and Dominic McIver Lopes, eds. Knowing art: essays in aesthetics and epistemology. Vol. 107. Springer Science & Business Media, 2007
Walton, Kendall L. "Categories of art." The philosophical review 79, no. 3 (1970): 334-367