In just one Over the years, the number of artworks produced by self-described AI artists has increased dramatically. Some of these works have been sold by major auction houses at staggering prices and have found their way into prestigious curated collections. Originally led by a few tech-savvy artists who adopted computer programming as part of their creative process, AI art has recently been embraced by the masses as image-generating technology has become even more effective and easier to to be used without coding skills.
The art movement of artificial intelligence travels over layers 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, takes 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 included in the original data set, often guided by texts that explicitly describe the desired results. Until recently, the images produced through this approach remained somewhat lacking in coherence or detail, although they possessed an undeniable surrealist charm that attracted the attention of many serious artists. However, earlier this year the Open AI technology company revealed a new model – dubbed DALL·E 2 – that can generate extremely stable and relevant images from almost any text. DALL·E 2 can even produce images in specific styles and imitate famous artists quite convincingly, as long as the desired effect is adequately specified in the request. 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 truly made by AI—are not particularly original. These questions echo similar concerns once raised by the invention of photography. By simply pressing a button on a camera, someone with no 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 almost any scene in any style. But cameras and algorithms don’t make art. People do. AI art is art, created by human artists who use algorithms as another tool in their creative arsenal. While both technologies have lowered the barrier to entry for artistic creation—which is something to celebrate, not worry—the amount of skill, talent, and intent involved in creating interesting artwork should not be underestimated.
Like any new tool, generative models bring significant changes to the art-making process. 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 AI art-making can involve curatorial acts. The first, and less original, has to do with results curation. Any generating algorithm can produce an indefinite number of images, but not all of these will usually achieve artistic status. The process of curating results is very familiar to photographers, some of whom routinely capture hundreds or thousands of photos from which few, if any, can be carefully selected for display. Unlike painters and sculptors, photographers and AI artists must deal with a plethora of (digital) objects, the curation of which is an integral part of the artistic process. In AI research in general, the act of “cherry-picking” particularly good results is seen as bad scientific practice, a way of misleadingly inflating a model’s perceived performance. When it comes to AI art, however, cherry picking may be the name of the game. The artist’s artistic intentions and sensitivity can be expressed in the very act of promoting specific results to the status of artistic works.
Second, curation can also occur before any images are created. 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 designing a dataset on which to train a network artificial nerve. This work is essential because if a dataset is poorly designed, the network will often not learn how to represent the desired features and perform adequately. Furthermore, if a data set is biased, the network will tend to reproduce, or even reinforce, such bias—including, for example, harmful stereotypes. As the saying goes, “garbage in, garbage out.” The adage is true of AI art as well, except that “garbage” takes on an aesthetic (and subjective) dimension.