Computational creativity
There’s a long history of artists using principles, or algorithms, or chance to create works, especially in the visual arts. Generative AI, working from a huge set of past images, texts, and/or scenarios, is just the latest.
· automatism in art: Joan Miró & André Masson (painting/drawing from the subconscious) Jackson Pollack (pouring, dripping, and pushing paint very consciously, but with stochastic elements in the conduct of the paint)
· algorithmic art: generated by computers using artist-provided algorithms, with a stochastic element; Vera Molnár, Dóra Maurer, Gizella Rákóczy
· Generative AI using Deep Neural Networks, working from a huge set of past images
· Generative Adversarial Networks, which create content and then assess them for verisimilitude [Gaskin 2018]
Benefits
AI systems allow the artist to experiment widely and repeatedly, by:
· selecting the datasets (training sets) to which the system is exposed – (careful artists use only out-of-copyright works, or even only their own prior works)
· selecting the parameters and degree of creativity or uniqueness requested
· assessing outputs
· revisiting all the above.
I’ll suggest that any artist, in any discipline, follows an analogous sequence – using their knowledge of prior and current works rather than digitized datasets – but deciding on the degree and parameters of uniqueness required, then experimenting, and then assessing outcomes.
“AI is used by artists because of its ability to provide surprising results, unexpected errors and glitches…. the non-deterministic nature of AI leads to errors and accidents that can have a critical role in the creation of an art piece” [Caramiaux & Alaoui 2022: 10]. The artist is “able to exploit the mistakes that AI models are able to produce” [p.11], selecting only those that serve the artist’s goals or vision. The role of the artist or team of artists remains huge.
Creative fields like video production, game creation, moviemaking, or advertising require characters, narrative arcs, set and costume design. Innovative characters, story lines, and visual composition are prized – but so is verisimilitude. Large Language Models and visual generative models (especially Generative Adversarial Networks, which create content and then assess them for verisimilitude) have been put to use in the pre-production of such works, as well as in their actual production. These tools can be used to develop new combinations of the scenarios on which they’ve been trained, or to fathom the structures of those scenarios and create new rules and thus vastly new scenarios [Vijayalakahmi et al. 2026].
Thus, the artistic use of AI tools can be at odds with the goals of many researchers and institutional users. Many artists seek creativity and non-determinate outcomes rather than accuracy, reproducibility, or productivity; many artists have carefully specified the datasets (or training sets) rather than using omnivorous and anonymous datasets.
Implications
Here’s a 1928 quote from Paul Valery, which appears in Walter Benjamin’s 1935 book The Work of Art in the Age of Mechanical Reproduction.
“We must expect great innovations to transform the entire technique of the arts, thereby affecting artistic invention itself and perhaps even bringing about an amazing change in our very notion of art.”
Questions
What are the implications for the future of art and artists?
It’s likely that the expanded universe of artistic creation will render some traditional forms and styles of art less valued, just as photography paved the way for “modern art” and conceptual art, reducing the “wow” value placed on new art that skillfully and painstakingly reproduces what we see in the world [Benjamin 1935; Ajani 2020]. However, “realistic” or mimetic visual art still hangs in galleries and remains popular, despite the success of photographic art.
I’d expect that creative activity that is not reliant on complete novelty – think of much pop music, basic cinematic soundtracks, pulp fiction, and what I might call background art and tourist art – may be generated by nearly autonomous AI. The composers, writers, and visual artists who have made a living from these products may become redundant in financial terms. However, there will always be people who just plain enjoy writing, composing, drawing, and painting: many will retain some, perhaps less remunerative foothold in their sale.
Which artists have the capability to understand the processes, the access to the technology, and the computing power to use them? [Caramiaux & Alaoui 2022].
Does AI art appeal? Does it sell?
Early AI works have been eagerly promoted by galleries and auction houses, and have sold very well [Gaskin 2018]. Some studies have found that viewers’ assessment of a work’s creativity declines when its use of AI is noted. Many art collectors care deeply about the artist’s “story,” motivations, and development over time, and may be put off by work that is largely the product of an automated process. Many art viewers, collectors, and artists appreciate the literally hand-crafted nature of painting and sculpture – yet many sculptures are produced through foundries that produce the work to the specifications of the “artist.” Indeed, “there are already painting machines being developed … that can turn certain digital images into material paintings done in oils or acrylics” [Kalyanaraman 2018].
Finally, there is some evidence that younger viewers are attracted by the possibilities that AI art have opened [de Rooij 2025].
Where does authorship lie?
Using AI to generate art is a collaboration between the “artist,” the algorithms, and the source materials – each of which has many “authors” [Ajani 2020]. But then, painting by hand is a collaboration between the makers of the tools, paints, surfaces, and all the art and writing about art of which the artist is aware. And certainly the production of cinema or music is a many-faceted collaboration of writers/composers, performers, set designers, costumers, directors, sound engineers…
Can the resultant works be copyrighted?
In most countries, copyright law requires a human author. Which of the collaborators would that be? [Gaskin 2018]. In 2023, a US Federal court judge agreed with the US Copyright Office’s refusal to grant copyright to the researcher who developed the AI system that created an original (digital) artwork [Small 2023].
References
Ajani, Gianmaria (2020). Contemporary Artificial Art and the Law: Searching for an Author. Brill Research Perspectives in Humanities and Social Sciences and Brill Research Perspectives in Art and Law. ISBN 978-900444-2689 (e-book).
Benjamin, Walter. (1935). The work of art in the age of mechanical reproduction (epigraph). In Illuminations, edited by Hannah Arendt, translated by Harry Zohn, from the 1935 essay. New York: Schocken Books, 1969.
Caramiaux, Baptiste and Sarah Fdili Alaoui (2022). “Explorers of unknown planets”: practices and politics of artificial intelligence in visual arts. Proceedings ACM Human-Computer Interaction 6: CSCW2, article 477.
Gaskin, Sam (2018). When art is created by artificial intelligence sells, who gets paid? https://scgaskin.wordpress.com/2018/09/19/when -an-ai-artist-makes-the-art-who-gets-paid/
Kalyanaraman, Karthik (2018). AI Art: a new photography moment. Medium. https://medium.com/@info_12534/ai-art-a-new-photography-moment-8d7009bfb696
de Rooij, Alwin (2025). Bias against artificial intelligence in visual art: a meta-analysis. Psychology of Aesthetics, Creativity, and the Arts. https://dx.doi.org/10.1037/aca0000833.
Small, Zachary (2023). As fight over A.I. artwork unfolds, judge rejects copyright claim. The New York Times, 21 August.
Valéry, Paul (1928). La Conquête de l'ubiquité (The Conquest of Ubiquity). Translated by Ralph Manheim, p. 225. Pantheon Books, Bollingen Series, New York, 1964.
Vijayalakahmi, A., M.V., Ulagammai, M., Cai, A., Parhi, M., and Wawage, P. (2026). Large language models for generating creative concepts in visual art pre-production processes. ShodhKosh: J. of Visual and Performing Arts, 7(4s): 161-72.