“A robot can’t enjoy art”

Image generated by DALL-E mini from the prompt: “Abandoned city with large figures in distance”

Can an algorithm comprehend art?

Can an algorithm be used as a creative tool?

In this essay, I’d like to argue that these are the same question. But first, a definition of the term is in order.

Defining creativity

In her 1990 book,“The Creative Mind: Myths and Mechanisms”, Research Professor Margaret A. Boden defines 3 types of creativity: “unfamiliar combinations of familiar ideas”, such as creating metaphors; “exploring conceptual space” by exploring what is possible within a structured way of thinking (e.g. poetry, painting, culinary arts, or biology theory); and “transforming the space” by shifting one’s way of thinking entirely. Boden places special weight on “transformative creativity”; if explorative creativity is a street artist honing their style and craft, transformative creativity means creating outside of their style and grasping at something completely new.

However, it may also be worthwhile to consider the meaning of creativity from a lexical perspective. For this, we may turn to analyses of the word “creative” and adjacent words from the Corpus of Contemporary American English:

So what do people think of when they read the words “creative”? First off, we observe from common collocates that the word “creative” unsurprisingly has a positive lean, being associated with words such as “inspire”, “nurture”, “innovative”, “wonderfully”, and “amazingly”. Something or someone that is seen as creative is likely to be seen as doing so to an exceptional degree; furthermore, to describe someone as creative is a mark of praise and a recognition of one’s unique capabilities. Creativity is something that can be “encouraged”, “fostered”, and “inspired”; interestingly, we often use the phrase “human creativity” as though affirming creativity as a unique quality of our species. It’s unsurprising, then, that discussion of whether to recognize AI as creative agents has led to some disagreement.

Is “AI can’t be creative” a fallacy?

from The Conversation: “Is AI-generated art really creative? It depends on the presentation.” (PA Images/Alamy)

There is a wide range of opinions on whether AI can be considered creative. Some critics assert that AI are themselves a product of their programmer and input data and thus lack the free will, agency or initiative to make a truly creative decision, regardless of their capabilities. However, Boden notes that whether or not we choose to recognize AI as being creative in and of its own right, its applications to the three types of creativity still have potential to imitate creativity, so to speak. Furthermore, this question is in line with the traditional pattern of debating “can a robot that can do [insert common human task here] really be considered [insert “uniquely human trait” here]?” whenever we think that technology is becoming advanced enough to replace us, perhaps pointing towards some of our internalized anxieties surrounding what it means to be human. It’s also important to note that we can easily be influenced to feel that an AI has humanlike qualities, as in the case of the Ai-DA robot “artist” (pictured above), so whether or not an AI itself is “creative” becomes quite subjective.

My personal feelings are that it is not difficult to imagine technology progressing to the point that such creative works that are indistinguishable from the works of human artists. When I consider this, I can’t help but hold a lingering suspicion that aggregation and imitation of a body of existing works is a poor stand-in for the thought process of someone who has experienced life as a human; at the same time, I find myself drawing arbitrary lines (“must be a trained artist”; “must live in human society”; “must be an adult”) without any firm conviction, simply to exclude something that is simultaneously super- and sub-human.

What is the future of creators and AI?

Makoto Amano, Hanako Hirata, Ryosuke Nakajima, Yuka Sai
UNLABELED — Camouflage against the machines, 2020

Creative collaboration between humans and AI/ machine learning to produce unique works is an interesting new field. In some cases, such as the development of a cloak that can spoof image recognition models into not recognizing them as a human, the mere existence and growing prevalence of AI in society acts as a catalyst for creation of reactionary works that push back against it. It may also spur introspection, causing us to reconsider our unique status as humans and the meaning of the human condition, while also serving as a tool for creative expression; by integrating AI directly into the work, the line between critique of AI and questioning of humanity is blurred. Lastly, the existence of AI might also be taken as an invitation to the artist to take the stage with them, such as in a live musical performance. All of the above works are exciting in that they represent the final type of creativity mentioned: Transformative Creativity, in which we shift our previous patterns of thinking in order to establish and explore a new style of expression.

As a layperson, I can only offer my opinion, but as we move forward with increasingly capable artificially intelligent technologies, we will be forced to accept a new way of living in which an aspect of humanity — creativity — which was once thought to be a hallmark of advanced sentient life is no longer unique to us. It can’t be seen merely as a tool, but also a growing influence as both inspiration and co-creator of creative works, shining light or holding up a mirror to aspects of humanity that we do not question on a daily basis. In the future, I anticipate that humans will continue to collaborate with AI creatively. However, because AI does not have an intrinsic understanding of humanity, it may be difficult for it to comprehend abstract concepts, understand the context of human culture, or recognize stereotypes and biases, all of which can be considered major aspects of being an artist and a human.

Below is a reflection on my experiences with AI-facilitated creation.

Conclusion: Webcomic made with DALL-E mini

Excerpt. The protagonist seeks to escape a dreamlike, AI-generated virtual world.

Based on observations about the tendencies of machine image generation, I imagined a hypothetical situation, 20 years in the future, where humans live in a machine-generated environment. How would they experience reality?

Although this story may be far-fetched, I was interested in exploring with the metaphor about how AI has already started becoming part of the “landscape” of our society. We interact with these opaque systems everyday, even though we do not fully understand or trust them to be consistent. At the same time, the concept of AI artistic expression is an interesting paradox; art is evaluated by humans who have a sense of the context that AI may not be capable of comprehending. I tried to explore how a piece of art changes depending on the viewer’s interpretation of it through the eyes of the beings inextricably involved in this fictional world.

Some of my inspirations for this were graphic novels and short stories. I was also inspired by the work “Openings !!!” by Szeman Petra, and The Magic Mountain by Takeru Ishiki (both on display at the NTT ICC). All of these harnessed surreal imagery and a feeling of being out of place to immerse the reader. Specifically, the two works at the ICC fused game- and animation-like aesthetics with a personal narrative to elucidate on deeper themes both explicitly and implicitly while playing on the idea of liminal spaces.. The UN-built exhibition in Nihonbashi was also referenced for its detailed, surreal fantasy environments that were based on a real-life location.

Takeru Ishiki, “The Magic Mountain”, 2021–22

Trying to put together a coherent narrative using only generated images was an interesting experience. I did not always get the results that I wanted, so finding a fitting image from DALL-E mini often took some prompt engineering. At the same time, having weird images was both an interesting constraint and a source of inspiration. I’m neither an artist nor writer, but having the setting and mood at least partly dictated through DALL-E mini’s surreal creative choices took away some of the pressure, although I’ll be the first to admit that the combination of dreamcore-esque illustrations and my deliberately but unhelpfully vague written interpretations of character motives made the final product very difficult to read, if unique.

The experience of working with DALL-E mini on a creative project really made me think of AI as a co-creator, and I’m excited for the potential of AI-created designs in our future lives.

--

--