by Xiaolan Deng
© 2024 Deng, X. All rights reserved
Citation: Deng, X (2024) Original Creativity, Associative Creativity, and AI available at https://www.oxford-aiethics.ox.ac.uk/original-creativity-associative-creativity-and-ai
I believe that the criteria we use to evaluate creativity affects how AI can threaten human creativity. In this essay, I start by introducing two kinds of creativity, which I call original and associative creativity. I will then show that while AI excels at associative creativity, it cannot achieve original creativity, and therefore cannot fully replace human creativity. However, due to our inherent limitations in perceiving original creativity, I argue that AI can still pose a significant threat to human creativity - not through direct replacement, but by reshaping how we produce and perceive creative works.
Before evaluating the role of AI in creativity, I will first clarify what creativity means. It can be classified into two categories: original creativity and associative creativity. The first framework views creativity as the production of something entirely new, a groundbreaking idea or work that transcends conventional thinking and logic. This kind of creativity is often revolutionary, as seen in Albert Einstein’s Annus Mirabilis papers, which redefine fundamental concepts of space, time, mass, and energy in modern science. In contrast, the latter views creativity as the ability to make connections between existing ideas in novel ways. From this perspective, no creation is truly independent of what came before; rather, creativity lies in the novel reconfiguration of familiar elements.
While both forms of creativity often coexist, original creativity tends to dominate human creativity. For example, when a poet composes a poem, the subjective emotions and ideas that inspire the work are entirely original, while the act of writing the poem can be seen as a rearrangement of existing words and symbols. Even so, we generally recognize the poet’s originality as the dominant creative force in the work. Now, imagine a scenario in which a cat randomly walks across a keyboard and accidentally types the same poem. While the arrangement of letters demonstrates associative creativity, we would not attribute original creativity to the cat.
Generative AI excels at associative creativity. By analyzing vast amounts of data, AI can identify patterns and connections that may not be immediately obvious to humans. In fields like music, art, and even writing, AI has demonstrated its capacity to combine familiar elements in unexpected ways. For example, when using DALL·E to generate artwork, one might input a prompt like “a sunset over the sea in the style of Vincent van Gogh.” The AI responds by creating a new image based on combining its understanding of Van Gogh’s style and patterns of light and colour of sunsets. Similarly, AI-driven music tools generate new melodies by blending existing music styles and pieces.
However, AI faces significant limitations when it comes to original creativity. AI is constrained by the datasets it is trained on and the algorithms that govern its operations, making its output inherently derivative from existing data. Even with advancements like reinforcement learning, which aim to push AI beyond its training data, I argue that AI cannot achieve original creativity due to its lack of two necessary conditions for original creativity: uniqueness and intentionality.
First, let's explore uniqueness. Human’s original creativity emerges from the dynamic interactions between the mind and the external world, shaped by emotions, mental states, and personal experiences—what philosophers refer to as "qualia." Qualia are essential to original creativity because they are inherently unique. No two individuals can experience the same qualia in exactly the same way, and even the same person cannot fully replicate their own qualia. For example, I may recall the pleasure I felt when I was admitted to Oxford the first time, but I could never feel that exact emotion again, even if I were to be admitted to Oxford again for a PhD program in the future. In contrast, AI models do not engage with reality in the same way; they rely on statistical analyses of digital data, all of which can be perfectly replicated.
Next, we can consider intentionality. Original creativity also involves making deliberate, mindful choices. When humans are inspired by qualia, they make spontaneous choices to shape it into concrete works. For instance, a poet might be moved by personal feelings of nostalgia and then consciously choose words, metaphors, and rhythms to convey that emotion in a poem. These decisions are spontaneous, guided by the poet’s own emotions. However, we would not view the poem generated by the cat as a work of original creativity because the cat is not consciously selecting words or expressing ideas; it’s simply a matter of chance. Similarly, AI lacks the ability to make intentional decisions. Its outputs are prompted by human trainers and determined by probabilistic models. Hence, unless AI can experience qualia and make intentional decisions, it will remain incapable of original creativity.
If AI can never achieve human-level original creativity, does this mean we can breathe a sigh of relief and confidently conclude that AI poses no threat to our creativity? Unfortunately, I don’t think so. The unreplicable nature of qualia is a double-edged sword: while it prevents AI from attaining original creativity, it also limits our ability to fully recognize and appreciate original creativity in others. When a human creator produces a piece of work—whether it's art, music, or writing—they first develop an abstract, internal concept inspired by qualia (original creativity), which they then transform through symbolic medium into a concrete form (associative creativity). This might involve shapes and colours for visual art, musical notes for compositions, or words and characters for written works. However, when an audience encounters the finished work, they experience the process in reverse. First, they perceive the outward symbols—the visual elements, sounds, or text—and then attempt to decode these into the emotions and ideas the creator intended to express. This translation process is not perfect; it is implicitly affected by the viewer’s own subjectivity. As a result, audiences often miss the creator’s true intent, leading to a partial or even distorted understanding. This subjective gap creates an opening for AI to blur the lines further. As AI’s associative creativity improves, it becomes increasingly difficult for audiences to distinguish between human-original and AI- generated works.
Some may argue that we could simply ask the author to explain their originate source, as a human can articulate their qualia while AI cannot. However, even this is a fragile distinction. Many artists and creators find it difficult, if not impossible, to fully explain the complicated, intuitive, and often subconscious emotions that drive their creative process. This ambiguity further weakens our ability to identify and value original creativity, further blurring the lines between AI-generated and human-created works.
The real threat AI poses to human creativity lies not in its ability to replace human creators, but in how it redefines the creative process. Traditionally, human creativity originates from within—ideas, emotions, and subjective experiences are then shaped into concrete works of art, music, or writing. AI, however, approaches creativity from the opposite direction. It relies entirely on associative creativity to produce outputs without original intention, leaving full interpretation to the audience. If people cannot reliably differentiate between human-made and AI-generated art, the original creativity behind a work may lose relevance. This shift could change how we value creativity—no longer focusing on the originating process, but rather on the final product and how we, as observers, interpret it. In this sense, AI could alter not only how creative works are produced, but also how audiences perceive and value creativity itself.
I understand that it is tempting to use AI for creative work because of its impressive speed and volume of output. AI can provide economic benefits to businesses and industries that prioritize visually impressive results. It can also serve as a powerful tool for human creators by accelerating repetitive process. While AI creativity cannot fully replace human creativity due to its fundamentally different nature, it can still pose threat by redefining the creative process, diminishing the value we place on original creativity that is unique to humans.