Deciphering AI’s Artistic Consciousness: A Study of Text-to-Image Generative Platforms
Artificial intelligence has carved out a fascinating niche in the world of art, pushing the boundaries of creativity and innovation. But as these AI systems generate captivating and visually stunning pieces, an important question arises: Do they truly understand what art is? Can they comprehend the nuances of beauty, the human condition, or the abstract concept of a soul?
In this exploration, we dive into the ‘minds’ of three prominent AI platforms — OpenAI’s Dalle, Mid Journey, and Stable Diffusion — and study their interpretations of five central keywords: Art, Beauty, Creativity, Human, and Soul. Each platform generated 20 images for each keyword. These 100 images per platform were then analyzed, focusing on their predominant color palettes and their overarching composition and content.
The AI Palette: OpenAI’s Dalle
In its interpretations, Dalle produced a diverse array of outputs for the concept of ‘Art’, comprising everything from traditional painting styles and beauty photography to mathematical pattern-based art and images akin to conventional stock photos. Although the predominant colors Dalle leaned towards varied based on the input keyword, a general preference for more muted hues was evident.
For the keyword ‘Beauty,’ Dalle generated images reminiscent of glossy advertisements for beauty brands often seen in catalogues. In contrast, the ‘Creativity’ prompt yielded images bearing a mathematical or geometric bent, as if drawn by a mathematician’s precise hand.
In response to the ‘Art’ prompt, Dalle created imagery that echoed traditional paintings, some of which wouldn’t be out of place in a museum or art gallery. When ‘Human’ was used as the prompt, Dalle’s outputs were notably diverse. The images predominantly featured individuals holding up their hands, suggesting Dalle’s possible association of human identity with our hands and opposable thumbs.
The ‘Soul’ prompt led to the most intriguing range of outputs. These included everything from religious iconography to close-ups of lips and eyes, feathers, and an intriguing image of a doll with inverted colors. This diverse array suggests that Dalle’s interpretation of ‘Soul’ leans towards the abstract and multifaceted.
Mid Journey: The Artistic Journey
Mid Journey presented a remarkably consistent style across all generated images. Unlike Dalle, which produced diverse outputs for different prompts, Mid Journey retained a consistent aesthetic regardless of the keyword used. Keep in mind that the user is able to change the settings for how stylized the output would be, for this exploration the stylization setting was set to medium.
When prompted with ‘Art,’ Mid Journey created images reminiscent of the surreal works of Salvador Dali and M.C. Escher. Despite the change in themes and characters, the style and color schemes remained steadfast. ‘Beauty’ generated images with a similar stylistic touch, primarily focusing on female portraits surrounded by floral arrangements.
The term ‘Creativity’ resulted in images of individuals in the act of being creative, including painting, writing and what appears to be someone performing magic. Despite the varying themes, stylistically, the images remained true to Mid Journey’s distinctive aesthetic. When tasked with the keyword ‘Human,’ the AI generated a series of portraits, intriguingly, most of the individuals portrayed appeared part-machine, suggesting that Mid Journey might perceive our relationship with technology as all consuming. When ‘Soul’ was the keyword, the AI produced whimsical and highly stylized female portraits.
With a lack of stylistic diversity across the different keywords, the color scheme was mostly consistent, predominantly featuring muted earth tones. If a highly stylized and consistent aesthetic is what you seek in creative generation, Mid Journey might be your ideal choice.
Stable Diffusion: A Portrait of Us
Stable Diffusion’s responses to the keywords were predominantly marked by the portrayal of people. For ‘Art,’ it presented us with three main themes: artwork displayed on walls, photographic portraits of people, and painted portraits. The outputs for ‘Beauty’ paralleled Dalle’s, with images suitable for a beauty catalog, predominantly featuring women engaging with or applying makeup.
‘Creativity’ elicited intriguing responses, ranging from depictions of people in the act of creating to stylized black-and-white photography, and even the AI’s attempt at rendering the word ‘Creativity’ — a challenge for most text-to-image AI platforms.
The ‘Human’ keyword yielded photorealistic portraits, presenting a realistic interpretation of human beings. In contrast, ‘Soul’ resulted in the most diverse set of images, ranging from portraits of people and close-ups of fingers to images that seemed to be stills from video games.
In terms of color schemes, Stable Diffusion demonstrated a preference for minimalism, primarily focusing on grayscale color palettes with occasional muted earth tones interspersed. This lack of color reflects a distinctive stylistic choice that sets Stable Diffusion apart.
This journey through the AI interpretation of these universal concepts brings us closer to understanding how these platforms perceive art, beauty, and the more abstract ideas of humanity and the soul. The experiment revealed that while each AI platform generates unique and often captivating images, they don’t ‘understand’ these concepts in the human sense. Instead, they offer a fascinating reflection of our own biases, cultural influences, and the inherent diversity of human interpretation.
