Artists working with programs like DALL-E do more than push a button—selecting outputs and engineering prompts are acts of aesthetic expression.
In just a few years, the number of artworks produced by self-described AI artists has dramatically increased. Some of these works have been sold by large auction houses for dizzying prices and have found their way into prestigious curated collections. Initially spearheaded by a few technologically knowledgeable artists who adopted computer programming as part of their creative process, AI art has recently been embraced by the masses, as image generation technology has become both more effective and easier to use without coding skills.
The AI art movement rides on the coattails of technical progress in computer vision, a research area dedicated to designing algorithms that can process meaningful visual information. A subclass of computer vision algorithms, called generative models, occupies center stage in this story. Generative models are artificial neural networks that can be “trained” on large datasets containing millions of images and learn to encode their statistically salient features. After training, they can produce completely new images that are not contained in the original dataset, often guided by text prompts that explicitly describe the desired results. Until recently, images produced through this approach remained somewhat lacking in coherence or detail, although they possessed an undeniable surrealist charm that captured the attention of many serious artists. However, earlier this year the tech company Open AI unveiled a new model— nicknamed DALL·E 2—that can generate remarkably consistent and relevant images from virtually any text prompt. DALL·E 2 can even produce images in specific styles and imitate famous artists rather convincingly, as long as the desired effect is adequately specified in the prompt. A similar tool has been released for free to the public under the name Craiyon (formerly “DALL·E mini”).
The coming-of-age of AI art raises a number of interesting questions, some of which—such as whether AI art is really art, and if so, to what extent it is really made by AI—are not particularly original. These questions echo similar worries once raised by the invention of photography. By merely pressing a button on a camera, someone without painting skills could suddenly capture a realistic depiction of a scene. Today, a person can press a virtual button to run a generative model and produce images of virtually any scene in any style. But cameras and algorithms do not make art. People do. AI art is art, made by human artists who use algorithms as yet another tool in their creative arsenal. While both technologies have lowered the barrier to entry for artistic creation— which calls for celebration rather than concern—one should not underestimate the amount of skill, talent, and intentionality involved in making interesting artworks.
Like any novel tool, generative models introduce significant changes in the process of art-making. In particular, AI art expands the multifaceted notion of curation and continues to blur the line between curation and creation.
There are at least three ways in which making art with AI can involve curatorial acts. The first, and least original, has to do with the curation of outputs. Any generative algorithm can produce an indefinite number of images, but not all of these will typically be conferred artistic status. The process of curating outputs is very familiar to photographers, some of whom routinely capture hundreds or thousands of shots from which a few, if any, might be carefully selected for display. Unlike painters and sculptors, photographers and AI artists have to deal with an abundance of (digital) objects, whose curation is part and parcel of the artistic process. In AI research at large, the act of “cherry-picking” particularly good outputs is seen as bad scientific practice, a way to misleadingly inflate the perceived performance of a model. When it comes to AI art, however, cherry-picking can be the name of the game. The artist’s intentions and artistic sensibility may be expressed in the very act of promoting specific outputs to the status of artworks.
Second, curation may also happen before any images are generated. In fact, while “curation” applied to art generally refers to the process of selecting existing work for display, curation in AI research colloquially refers to the work that goes into crafting a dataset on which to train an artificial neural network. This work is crucial, because if a dataset is poorly designed, the network will often fail to learn how to represent desired features and perform adequately. Furthermore, if a dataset is biased, the network will tend to reproduce, or even amplify, such bias—including, for example, harmful stereotypes. As the saying goes, “garbage in, garbage out.” The adage holds true for AI art, too, except “garbage” takes on an aesthetic (and subjective) dimension.
For his work Memories of Passersby I (2018), German artist Mario Kinglemann, one of the pioneers of AI art, carefully curated a dataset of thousands of portraits from the 17th to 19th centuries. He then used this dataset to train generative algorithms that could produce an infinite stream of novel portraits sharing similar aesthetic characteristics, displayed in real time on two screens (one for female portraits, one for male portraits). This is an example of an AI artwork that does not involve output curation. Still, the meticulous curation of the training data played a fundamental role in its conception. Here, “bias” is a blessing: The dataset was heavily biased according to the artist’s personal aesthetic preferences and taste, and this aesthetic bias is reflected in the final artwork, albeit through the distorting lens of the computer-driven generative process.
Another novelty spurred by the recent progress of generative algorithms is the ability to produce images by describing the desired result in natural language. This has come to be known as “prompting,” or guiding the algorithm with text prompts as opposed to sampling random outputs. Consider the illustration accompanying this article: The collage features several images generated by prompting DALL·E 2 with the phrases “an AI image generation algorithm, conceptual art,” “collage with images made by a generative AI model, illustration from Wired magazine,” and “an artist curating artworks produced with an AI algorithm, conceptual art.”
In some ways, being able to prompt a generative algorithm with words makes the creative process both easier and more focused. It may reduce the need for the curation of outputs, as one can directly describe one’s vision. However, prompting is not a silver bullet that trivializes artistic creation. It is more akin to a new kind of creative skill. AI researchers even talk about “prompt engineering” to describe the process of crafting good prompts to obtain desired results.
Prompt engineering is more of an art than a science, especially when it comes to creative uses of AI. It has even been compared to alchemy, or incantation. In addition to having a unique vision for the final products, one must get a feel for the right combination of magic words that will unlock specific styles or subjects with any given algorithm. Therein lies the third and perhaps most novel form of curation introduced by AI art: carefully designing and collecting personal prompts, or prompt fragments, that elicit desired results from an algorithm.
As the use of pre-trained algorithms like DALL·E 2 starts to obviate the need for dataset curation, prompt curation offers an alternative way of developing a personal artistic style. Interestingly, it also places images in dialog with text, as traditional museum curation does, although in a less academic and often more poetic format. Like art commentary, prompts can be very literal (“A man standing in a corn field, low angle, 35-mm portrait photography”) or very abstract (“The unbearable lightness of being”). Either way, prompts impose a novel layer of interpretation on artworks. Some artists like to share their prompts and may even use them as titles for their works; others prefer to keep them to themselves and leave the resulting images open to interpretation.
The curation of prompts and the curation of outputs often become entwined in a creative feedback loop. One might try out a given prompt, get a sense of the images it can produce, then use that new knowledge to iteratively refine the prompt, picking out interesting outputs in the process. This cycle can be repeated over and over, ad infinitum. It is reminiscent of traditional artists exploring variations on a common theme, such as Picasso’s lithograph series The Bull (1945), in which he depicted a bull at various stages of abstraction. One noteworthy difference is that the quasi-alchemic prompting procedure always involves an element of surprise guaranteed by the stochastic nature of generation: No prompt will produce the exact same result twice, and slight variations in the prompt may have an unexpectedly large impact on the outputs.
The blurring of boundaries between artists and curators is not new. While curation was initially seen as a merely custodial endeavor, tasked with preserving and displaying a catalog of artworks in a museum, since the 1960s it has come to be recognized as a creative gesture in itself. Curating an exhibition often involves deliberately adopting a particular concept or perspective to shine a new light on a set of artworks. Star curators such as Carolyn Christov-Bakargiev and Hans Ulrich Obrist approach their work like artists and have had an influential role in shaping contemporary discourse about art and curation. Conversely, artists such as Marcel Duchamp curated iconic events themselves and played a pivotal role in modernizing the exhibition medium. As a creative process in its own right, curation can become a deeply personal expression of artistic taste. The progress of generative algorithms creates additional opportunities for cross-pollination between art and curation by introducing new curatorial gestures that channel the artist’s aesthetic sensibilities at several stages of the creative process.
These curatorial aspects of AI art may eventually percolate through curatorial practices in museums or digital exhibitions. For example, institutions exhibiting AI art will need to decide how much information to provide about the datasets on which algorithms used to produce specific artworks were trained. Sotheby’s catalog note for Memories of Passerby I mentions that the training dataset contained 17th- to 19th-century portraits, which provides relevant context to understand the artwork and its art historical lineage. If a prompt was used to produce a piece and was communicated by the artist, curators may decide to include and reflect on it in their presentation. In line with the idea of the curator as (AI) artist, one could also conceive of an exhibition in which traditional artworks are selected on the basis of the similarity of the captions an algorithm assigns to them (see Google Arts & Culture for similar experiments in digital curation). One thing is certain: Technological innovations from AI research will continue influencing artistic creation and curation in exciting and unpredictable ways that provide fertile ground for novel forms of creativity.