Artistic Intelligence vs. Generative AI: Old Ideas, New Tensions
Human artistic intelligence (HAI) refers to a “system of capacities for perception, sensing, discernment, insight, activity, choice-making, and divergent synthesis that is developed by, and transcends beyond human intelligence.” Human AI also includes embodied, inherited, and protected wisdoms — knowledge accumulated and transmitted through body, gift, and ritual. When I came up with the brand name “Art & Algorhythms” I was thinking about the tensions between HAI and generative AI (GenAI) that uses sophisticated algorithms to generate novel outputs in response to user prompts. Both types create new content, i.e. text, images, or audio, based on learned patterns, or inherited systems.
It’s been stated that the famous visual artist Picasso gained inspiration for Cubism after viewing traditional abstract sculptures of the Makonde people of north-Tanzania. For hundreds of years, the Makonde people have carved their traditional sculptures of intertwined bodies and faces in varying geometric shapes and from varying angles. Despite this, European artists were widely credited with ‘inventing’ this style/method as a foundation for Cubism and 20th-century abstract art. The copying of the Makonde aesthetic or style (without acknowledgement or permission) is one example of appropriation, which has been a common practice in modern art for well over a century.
The ‘age of mechanical reproduction’ began with the invention of machines such as photostats that were used for making photographic copies of graphic matter or content. Throughout his career, Black American artist Romare Bearden made a practice of copying, redrawing, and reworking images, and he made large-scale photo reproductions (called ‘photostats’, after the machine used to make them) of these collages. Bearden, unlike the Cubists before him, made use of abstraction in a new or different way through collage. This is an example of reappropriation. Both methods were widely accepted in the art world in the 20th century.
Harold Cohen’s AARON is known to be the earliest generative AI program. AARON is the collective name for a series of computer programs written by Cohen that create artistic images autonomously. Cohen understood his work with AARON to be a collaboration, and he devoted his life to exploring the potential of AI to translate an artist’s knowledge and process into code. Like Romare Bearden before him, Cohen used methods that allowed machines such as flatbed plotters to produce physical artwork. Also like Bearden, Cohen colored the images by hand and scaled them up to make larger paintings or murals.
The variety of noises is infinite. If today, when we have perhaps a thousand different machines, we can distinguish a thousand different noises, tomorrow, as new machines multiply, we will be able to distinguish ten, twenty, or thirty thousand different noises, not merely in a simply imitative way, but to combine them according to our imagination. — Luigi Russolo (1913)
In the context of AI, diffusion models learn to generate new data samples (ex. images or text) by progressively adding noise to existing data and then learning to reverse that process, starting with pure noise and progressively removing the noise to reconstruct or remix the original data. The simplification process that we saw with 20th century modern artists like Pablo Picasso has become the vision of Futurists whose works were heavily influenced by, but distinct from, Cubists. Futurist artist and musician Luigi Russolo envisioned the possibility and potential of generative AI over a century before it became a reality. Human artists are still the creators but the potential for collaboration with algorithms (like Harold Cohen) is here.
In this new reality all content on the web is noise (diffusion) and, according to the Futurists, this noise can be used as a ‘deliberate’ artistic element. It could be argued that generative AI is not new, but a continuation of what has come before. Noise is complex, and if engaged with aesthetically, it can produce astonishingly profound experiences and understandings. It can disrupt systems and remix practices according to the designs of the user. The Italian Futurists understood the power of noise just as the Cubists understood the impact of abstraction. Today’s AI artists understand the importance of both processes. If one is tapped into hip hop (culture) there is an additional layer of intentionality in the generative AI art space.
Reappropriation encourages the use of imaginative and innovative methods, which is evident in the creation of artwork from found materials. Below the surface, reappropriation is defined as:
[T]he counterhegemonic practice of repurposing things in ways that revalue, resignify, and relocalize artifacts from mainstream, or dominant, cultures.
Counter-hegemonic AI art refers to the use of AI in artistic creation to challenge existing power structures and dominant ideologies within society and the art world itself. AI-generated art can be used by activists and historically marginalized communities to express their experiences and perspectives, challenging societal prejudices and giving visibility to underrepresented narratives. For example, AI art can help artists from these groups enhance their creativity and reimagine meaning.
Remixing in MidJourney lets users steer how their images turn out. Users can change both their prompt and the parameters they use. Variations allow people to create different versions of the same image by changing little details or the overall composition. I’ve written a lot about this process as well as improvisation in previous entries. Improvisation emphasizes the call-and-response participatory aspect of creative expression and collaboration. Entering a prompt (call) into MidJourney’s “/imagine” field results in a response from the AI generator. With remix mode turned on, users can tweak or adjust their prompts while generating images, as well as make changes after an image has been generated.
In general, today’s generative AI art practice was originally envisioned over a century ago when noise (diffusion) was thought of as an art material, and eventually became a tool for expanded creative possibilities. Also, the use of AI by artists triggers and explores tensions created through the appropriation of data and the reappropriation of styles and content.