Can Computers Create Art?

Can Computers Create Art?

Aaron Hertzmann
Adobe Research111This essay expresses my own personal opinions and does not reflect those of my employer.
Working draft222 This manuscript is in preparation for submission to the journal Arts, Special Issue on “The Machine as Artist (for the 21st Century)” ( ), by invitation. Please send comments to However, I cannot promise to respond to all feedback.

This paper discusses whether computers, using Artifical Intelligence (AI), could create art. The first part concerns AI-based tools for assisting with art making. The history of technologies that automated aspects of art is covered, including photography and animation. In each case, we see initial fears and denial of the technology, followed by a blossoming of new creative and professional opportunities for artists. The hype and reality of Artificial Intelligence (AI) tools for art making is discussed, together with predictions about how AI tools will be used. The second part speculates about whether it could ever happen that AI systems could conceive of artwork, and be credited with authorship of an artwork.

Recent advances in artificial intelligence for making images have come at a staggering pace lately; along with them, so have come fears about Artificial Intelligence (AI). AI artists are being trained to produce their own art, we are told. In typical news media hype, new neural network algorithms will revolutionize everything and automate everything, thus saving lives and making it easier for people to express themselves, but maybe they will put human workers out of work and, while they’re at it, rob us of our humanity?

Beyond the hype, confusion about how technology influences art pervades serious discussions. Professional artists often are concerned that computers might put them out of their jobs [51], a concern I’ve heard for decades. Computer scientists often view artists in a way that borders on mysticism.

Some researchers present their algorithms as themselves possible artists, e.g., [11, 21]. And, I was recently contacted by a prominent social psychologist who, inspired by recent results with neural networks, wished to conduct experiments to assess whether ordinary people might be willing to buy artwork made by a computer, and, if so, why. It was assumed that computers were already happily making their own artwork.

On the other hand, whenever I have informally asked friends or colleagues the question of whether computers can create art, the answer is usually a decisive “No.” Art requires human intent, inspiration, a desire to express something. Thus, by definition, there is no such thing as art created by a computer… why would anyone worry?

In this essay, I will tackle the question of “Can Computers Create Art?” There are several related versions of this questions worth discussing. Can computers create art now? In the future, will computers replace professional artists or make their technical skills obsolete? Does the current definitions of art allow for a computer artist? Can we imagine a future scenario in which a computer is itself considered the author of an artwork? Will AI change our understanding of what art is? Here I discuss each of these questions.

Given the emphasis on intent as the role of the artist, it is conceptually worthwhile to separate the creation of art into execution versus intent. “Execution” refers to the application of technical skill required to go from an idea to an output, e.g., the ability to paint a picture of a particular subject in a particular style. “Intent” refers to the conception of the idea and the goals of the work, together with the high-level decision-making during an iterative process. The “Intent” is not just the original conception of the work, because most art is iterative: often, an artist will emerge from the process with something very different from their original idea. For example, a painter might begins with a vague idea, rough a sketch, modifies it, paint more, steps back to assess their work so far, change the idea based on how the painting is evolving, and continues to iterate. This separation of decision-making from execution is sometimes artificial, but useful for this essay so that we can discuss their automation separately.

One of my main themes is that technological innovation ultimately benefits art and artists (and thus everyone). Sadly, art and science are often viewed as being separate, or in even opposition or competition [64]. I describe several previous times in history when artists feared new technologies, when, in fact, these new developments ultimately had the effect of dramatically expanding the opportunities for artists. Often, artists are important contributors to these innovations; artists and technologists have much in common, as tinkerers, experimenters, and explorers. Rather than being a foe, technological development stimulates so much of the continued vitality of art. I see every reason to believe that AI technologies will do the same.

Part 1: Execution

I now focus on two related questions: given instruction and guidance, can machines create artwork as well as humans? That is, can they replace the technical skills of paid professionals? And, if so, should human artists — particularly, commercial and professional artists like illustrators and designers — worry about their job security?

For example, consider an art director who must come up with an original magazine cover each month, or original book covers for new novels. Normally, they might start with a rough idea, hire a professional designer or illustrator to explore related ideas and execute on the design brief. In the hypothetical AI-driven world, they would enter their idea into a software package that would execute the idea, and the art director could explore or iterate on different ideas with the software package. The art director performs the high-level decision-making: they decide on the underlying concept, and must have some level of taste and discernment in selecting amongst options. But their need to hire a professional designer is replaced by the computer. There are analogous possibilities in many other work-for-hire scenarios, like the studio musicians performing in a rhythm section in an album recording, or actors performing in a film. There are more far-out scenarios, e.g., where the film script itself is automatically generated [28]. Is this realistic?

I argue that, throughout history, technology has expanded creative and professional opportunities for artists dramatically, by providing newer and more powerful tools for artists. The advent of new technologies often cause fears of displacement among traditional artists. In fact, these new tools ultimately enable new artistic styles and inject vitality into art forms that might otherwise grow stale. These new tools also make art more accessible to a wider sections of society, both as creators and as consumers. These trends are particularly visible in the past two centuries since the Industrial Revolution. Furthermore, I argue that these trends will continue for the foreseeable future with AI technologies.

1.1 How photography became an artform

“From today, painting is dead!” - Paul Delaroche, painter, at a demonstration of the daguerreotype in 1839333The historical information from this section is distilled from two texts: Rosenblum [59] and Scharf [62].

For lessons from the past about AI and art, perhaps no invention is more significant than photography.444This connection has been made previously [1, 32], but I expand upon it much more here. Prior to the invention of photography, realistic images of the world could only be produced by artists. In today’s world, we are so swamped with images that it is hard to imagine just how special and unique it must have felt to see a skillfully-executed realistic painting. The technical skills of painting and other visual arts were inseparable from other creative aspects. Photography automated a task that had previously been solely done by artists, that is, reproducing images of the real world.

In 1839, the first two commercially-practical photographic processes were invented: Louis-Jacques-Mandé Daguerre’s dagguereotype, and William Henry Fox Talbot negative-positive process. They were mainly presented as ways to produce practical records of the world. Of the two, the daguerreotype was more popular for several decades, because Talbot’s process was restricted by patents. Improvements to Talbot’s method eventually made the daguerreotype obsolete and evolved into modern film processes.

Portraiture was a main driver for early adoption. Then, as today, people enjoyed possessing pictures of their friends, loved-ones, and ancestors. Portrait painting was only available to aristocrats and the very wealthy; even today, portrait painting is a symbol of great wealth and status. In the 18th century, a number of inexpensive methods were developed, such as the silhouette, a tiny black representation of an individual’s outline (Figure 1(a)), typically hand-cut by an artisan out of black paper. The daguerreotype offered an economical way to create a realistic portrait (Figure 1(b)). It was very slow and required locking the subject’s head in place with a head brace for several minutes, while the subject gripped their chair tightly, so as not to flutter their fingers. Nonetheless, numerous daguerreotype studios arose and became commonplace as technologies improved; recognizing its potential, many portraitists switched to this new technology. Portrait painters even adopted photography as a helpful crutch for painting. By 1863, a painter-photographer named Henri Le Secq said “One knows that photography has harmed painting considerably, and has killed portraiture especially, once the livelihood of the artist.” Photography largely replaced most older forms of portraiture, such as the silhouette, and no one seems to particularly regret this loss. As much as I appreciate the mystery and beauty at looking at old etchings and portraits, I’d rather use my mobile phone camera for my own pictures than to try to paint everything by hand.

(a) (b)

Figure 1: (a) A traditional silhouette portrait. (b) Daguerreotype portrait of Abraham Lincoln, 1846. Photographic techniques like this completely displaced previous portraiture techniques.
Figure 2: The Unhappy Painter, by Theodor Hosemann, 1843, satirizes the painter, a victim of progress, made obsolete by a daguerreotype.

Another early use for the daguerreotype was to produce souvenirs for tourists: by 1850, daguerreotypes of Roman ruins completely replaced the etchings and lithographs that tourists had previously purchased. As the technology improved, photography became indispensable as a record of engineering projects, disappearing architectural ruins, and documentary purposes, such as Matthew Brady’s photographs of the horrors of the American civil war.

“Is photography art?” This question was debated for many decades, coalescing into three main positions. Many people believed that photography could not be art, because it was made by a mechanical device rather than by human creativity. Many artists were dismissive of photography, and seemed threatened by it. A second view was that photography could be useful to real artists, such as for reference, but should not be considered as equal to drawing and painting. Finally, a third group, relating photography to established forms like etching and lithography, felt that photography could eventually be as significant an art form as painting.

Photography ultimately had a profound and unexpected effect on painting. Though some painters put more or less emphasis on realism versus expressionism, painters’ mimetic abilities had been improving over the centuries. Many painters of the 19th century, such as the Pre-Raphaelites like John Everett Millais and Neoclassicists like Ingres, painted depictions of the world with dazzling realism, more than had ever been seen before. However, as cameras became cheaper, lighter, and easier to use, they grew widespread among both amateurs and professionals, and realistic imagery became commonplace by the end of the 19th century. If photorealism could be reduced to a mechanical process, what is the real value of art? This question drove painters away from visual realism toward greater abstraction, through movements such as Symbolism and Tonalism, as in the atmospheric scenes of James MacNeill Whistler. Whistler wrote: “The imitator is a poor kind of creature. If the man who paints only the tree, or the flower, or other surface he sees before him were an artist, the king of artists would be the photographer. It is for the artist to do something beyond this.” The Impressionists, who sought to capture the perceptions of scenes, were likely inspired by the “imperfections” of photographs, such as motion blur. Edvard Munch wrote “I have no fear of photography as long as it cannot be used in heaven and in hell. … I am going to paint people who breathe, feel, love, and suffer;” though he did experiment with photography extensively. Vincent Van Gogh, describing his artistic breakthroughs around 1888, wrote to his brother: “You must boldly exaggerate the effects of either harmony or discord which colors produce. It is the same thing in drawing — accurate drawing, accurate color, is perhaps the not the essential thing to aim at, because the reflection of reality in a mirror, if it could be caught, color and all, would not be a picture at all, no more than a photograph.” One can infer a significant influence of Étienne-Jules Marey’s multiple-exposure photography on Futurism and Cubism, e.g., in Duchamp’s Nude Descending A Staircase.

One can argue, in fact, that photography was one of the major catalysts of the Modern Art movement: its influence led to decades of vitality in the world of painting, as artists were both inspired by photographic images and pushed to go beyond realism.

Figure 3: The interplay of painting and early photography. (a) By the 19th century, Western painters had achieved dazzling levels of realism. (b) Early cameras took low-quality (though evocative) pictures. This daguerreotype took over ten minutes to expose. (c) However, camera technology steadily improved, capturing greater and greater realism, in much faster exposures. (d) This challenged painters to create works that were not about hyper-realistic depiction, such as Whistler’s Tonalist Nocturne. (e) The Pictorialist photographers attempted to establish photography as an art form by mimicking the styles and abstraction of painting. Works: (a) Ophelia, John Everett Millais, 1851 (b) Boulevard du Temple, Daguerre, 1838; (c) Portrait of Sarah Bernhardt, Félix Nadar, 1864; (d) Nocturne in Blue and Gold: Old Battersea Bridge, James MacNeill Whistler, c. 1872-1875; (e) Morning, Clarence H. White, 1908

The Pictorialist movement, begun around 1885, was an attempt to firmly establish photography as an art form. Pictorialists introduced much more artist control over the photographs, often manipulating their images in the darkroom. Other times, they used highly-posed subjects, similar to those seen in conventional paintings. Many of their works had a hazy, atmospheric work similar to Tonalism (e.g., Whistler’s paintings), that deliberately softened the realism of high-quality photography. Today, many of these works seem evocative but also very affected in their emulation of the painting styles of the time. They pursued various strategies toward legitimization of their work as art form, such as the organization of photographic societies, periodicals, and juried photography exhibitions [67]. Their works and achievements made it harder and harder to deny the artistic contributions of photography; culminating in the “Buffalo Show,” organized by Alfred Stieglitz at the Albright Gallery in Buffalo, NY, the first photography exhibition at an American art museum, in 1910 [2]. Photography was finally established as an art, and free to move beyond the pretensions of Pictorialism.

This story provides several lessons that are directly relevant for AI as an artistic tool. At first, photography, like AI, was seen by many as non-artistic, because it was a mechanical process. Some saw photography as a threat and argued against its legitimacy. Photography did displace old technologies that had fulfilled non-artistic functions, such as portraiture’s social function. Some artists enthusiastically embraced the new technology, and began to explore its potential. As the technology improved, and became more widespread over nearly a century, artists learned to better control and express themselves with the new technology, until there was no more real controversy over the status of photography. The new technology made image-making much more accessible to non-experts and hobbyists; today, everyone can experiment with photography. Furthermore, the new technology breathed new life into the old art form, provoking it toward greater abstraction. Wherever there is controversy in AI as an artistic tool, I predict the same trajectory. Eventually, new AI tools will be fully recognized as artists’ tools; AI tools may stimulate traditional media as well, e.g., the New Aesthetic [66].

1.2 The technology of live-action cinema

The story of filmmaking and technology has important lessons about how artists and technologists can work together, each pushing the other further. Most of the early photographers were, by necessity, both artists and technologists, experimenting with new techniques driven by their art or to inspire their art. But, in film and animation, the interaction has been much more central to the art form.

The history of film is filled with artist-tinkerers, as well as teams of artists and technologists. Eadweard Muybridge invented the first camera for capturing motion. The Lumière Brothers created the first short film, a simple recording of workers leaving their factory, but also experimented with a wide range of camera technologies, color processing, and artistic ways to use them. The stage magician George Méliès filmed fantastical stories like A Trip to the Moon, employing a wide range of clever in-camera tricks and techniques to create delightfully inventive and beguiling films. Walt Disney employed and pushed new technologies of sound and color recording, and drove other innovations along the way, such as the multiplane camera. Many of Orson Welles’ innovative film techniques were made possible by new camera lenses employed by his cinematographer Gregg Toland. The introduction of the Steadicam in 1975 enabled directors to create a much more first-person sense of movement. George Lucas’ team for Star Wars was an early developer of many new visual effects on a shoestring budget (think of Ben Burtt hitting telephone guy-wires to create the “blaster” sound effect), as well as an early innovation in digital film editing and compositing. Digital and computer graphics technology, have, obviously, revolutionized film storytelling since then, with directors like Michel Gondry and James Cameron pushing the technology further into unforeseen directions. In each case, we see technologies rapidly adopted by directors to create new storytelling techniques and styles, transforming the medium over and over.

(a) (b) (c)

Figure 4: Technological developments in the art of filmmaking: (a) the first captured film, of workers leaving the Lumiere Brothers’ factory (1895), (b) George Méliès’ 1908 A Trip to The Moon filmed like a stage play but with fantastical special effects, (c) Citizen Kane, which used numerous experimental camera and lens effects to tell the story.

1.3 3D computer animation: a tight collaboration

3D computer animation as an artform was pioneered by Pixar Animation Studios, and that success is due to the close collaboration of artists and engineers [56]. It all began with Ed Catmull, an animation enthusiast who received a PhD in computer science in 1974. In his thesis, he invented several core techniques that every major 3D computer graphics system uses today. During his time in graduate school, he quietly set a goal for himself: to make the world’s first computer-animated film [7]. Consequently, he founded the Lucasfilm Computer Division, and hired a team of brilliant engineers to invent computer systems to be used for film-making. However, none of this group could animate, that is, bring a character to life through movement. Hence, Catmull recruited John Lasseter, an animator trained deeply in the Disney tradition. Through tight collaboration between Lasseter and the technical staff, they were able to invent new technologies and discover together how computer animation could start to become its own art form [43]. This group spun out as Pixar, and, over the following years, invented numerous technical innovations aimed at answering the needs set out by Pixar’s artists; in turn, the artists were inspired by these new tools, and pushed them to new extremes, and so on. One of their mantras was “Art challenges technology, technology challenges art” [7].

Pixar, by design, treats artists and engineers both as crucial to the company’s success and minimizes any barriers between the groups. When I worked there during a sabbatical, despite my technical role, I had many energizing conversations with different kinds of artists, attended many lectures on art and storytelling, sketched at an open life drawing session, watched a performance of an employee improv troupe, and participated in many other social and educational events that deliberately mixed people from different parts of the company. This is the culture that, though it still has some flaws to address, achieved so many years of technical and creative innovation, and, ultimately, commercial and artistic success.

Computer animation was another instance where many artists feared new technology. In the early, pre-Pixar days, Catmull’s group made many attempts to interest Disney animators in their work [56]. Alvy Ray Smith later said: “Animators were frightened of the computer. They felt that it was going to take their jobs away. We spent a lot of time telling people, ‘No, it’s just a tool—-it doesn’t do the creativity!’ That misconception was everywhere” [54]. It is a common misconception that computer animation just amounts to the computer solving everything; like a programmer presses a button and the characters just move on their own. In reality, computer animation is extraordinarily labor-intensive, requiring the skills of talented artists (especially animators) for almost every little detail. Character animation is an art form of extreme skill and talent, requiring laborious effort using the same fundamental skills of performance — of bringing a character to life through pure movement — as in conventional animation [43]. Even simulated scene elements, like water and cloth, involve laborious manipulation and adjustment in order to get the “look” just right.

Traditional cel animation jobs did not last at Disney, for various reasons. Disney Feature Animation underwent a renaissance in the early ’90s, starting with The Little Mermaid. Then, following some changes in management, the Disney animation began a slow, sad decline. After releasing duds like Brother Bear and Home on the Range, management shut down all cel animation at Disney, and converted the studios entirely to 3D computer animation. Many conventional animators were retrained in 3D animation, but Disney’s first 3D animation, Chicken Little was still a dud. Following Disney’s acquisition of Pixar several years later, they revived Disney’s beloved cel animation productions. The result, a charming and enjoyable film called The Princess and The Frog, performed so-so at the box office, and, moreover, the animators’ creative energy was focused on the newer 3D art form [7]. Today, traditional cel animation at Disney is dead.555Traditional animation styles are more vibrant in countries like Japan and France that, unlike Americans, do not believe that animation is “just for kids.” Even so, their visual styles of have evolved considerably due to computer technology. Today, computer animation is a thriving industry, and it thrives in many more places than cel animation ever did: at many different film studios, in visual effects for live-action films, in video games, television studios, web startups, independent web animators, and many more. There are now more types of opportunities for animators than ever before. The story here is not the destruction of jobs, but the evolution and growth of an art form through technology.

Each of these stories contradicts the popular notion of art and technology operating in conflict, when, in fact, the opposite is usually true. My current employer is Adobe Systems, a company with the primary mission of creating new technology for artists and designers.

1.4 Procedural artwork

In the art world, there is a long tradition of procedural artwork. Jean Arp created artworks governed by laws of chance in the 1910s (or so he claimed), and, beginning in the 1950s, John Cage used random rules to compose music. The term “Generative Art” appears to have originated in the 1960s. Starting in the 1970’s, fine artist Harold Cohen began generating paintings using a program he wrote called AARON [8]. Since the 1980s, many current artists, such as Karl Sims, Scott Snibbe, Golan Levin, Scott Draves, and Jason Salavon, produce abstract artworks by writing computer programs that produce either static images, or create interactive artistic experiences and installation works (Figure 5). In Sims’ and Draves’ work, the artwork “evolves” according to audience input. The popularity of the Processing computer language for artists speaks to the growth of this area.

Figure 5: Procedural artworks in the fine art world of galleries and art museums. Left: Electric Sheep, by Scott Draves, evolves dazzling procedural abstract animations based on thousands of votes. Right: The Top Grossing Film of All Time, 1 x 1 (2000) by Jason Salavon, shows the average color of each frame of the movie Titanic.

In each of these cases, despite the presence of procedural, emergent, and/or crowdsourced elements, the human behind it is credited as the author of the artwork, and it would seem perverse to suggest otherwise. The human has done all of the creative decisionmaking around the visual style, of testing and evaluating alternative algorithms, and so on.

1.5 State of the art: computer science research

Recent developments in computational artistic image synthesis are quite spectacular. But they should not be mistaken for AI artists.

Non-Photorealistic Rendering (NPR) is a subfield of computer graphics research [60] that I have worked in for many years. NPR research develop new algorithms and artistic tools for creating images inspired by the look of a conventional media, such as painting or drawing. Paul Haeberli’s groundbreaking 1990 paper [30] introduced a paint program that began with a user-selected photograph. Whenever the user clicked on the canvas (initially blank), the system placed a brush stroke with color and orientation based on the photograph. In this way, a user could quickly create a simple painting without any particular technical skill (Figure 6(a)). In a follow-up paper, Pete Litwinowcz automated the process entirely, by placing brush strokes on a grid [46]. My own first research paper arose from experimenting with modifications to his algorithm: the method I came up places long curved strokes, beginning with a large strokes that were then refined by small details [31] (Figure 6(b)). The algorithm was inspired by my experience with real painting, and the way artists often start from a rough sketch and then refine it.


Figure 6: Painterly rendering algorithms that process an input photograph, using hand-coded rules and algorithms. (a) Paul Haeberli’s interactive painting system [30]. (b) My automatic painting system [31] processes a photograph without user input aside from selecting some parameter settings.

This type of artistic algorithm design reflects the majority of computer graphics research in this area [33]. The algorithms are automated, but we can explain in complete detail why the algorithm works and the intuitions about artistic process it embodies. This mathematical modeling of artistic representation continues the tradition begun in the Renaissance with Filippo Brunelleschi’s invention of linear perspective, a belief that I have expanded upon more elsewhere [34].

At some point, I found it very difficult to embody richer intuitions about artistic process into source code. Instead, inspired by recent results in computer vision [20], I began to develop a method for working from examples. My collaborators and I published this method in 2001, calling it “Image Analogies” [35]. We presented the work as learning artistic style from example. But the “learning” here was quite shallow. It amounted to rearranging the pixels the of the source artwork in a clever way, but not generalizing to radically new scenes or style (Figure 7). Since then, other researchers have improved the method substantially [22], based on the same principle of rearranging patches of pixels.

Figure 7: Our Image Analogies algorithm [35], which stylizes a photograph in the style of a given artwork; in this case, Van Gogh’s Starry Night over the Rhône.

In 2016, Leon Gatys and his colleagues published a new breakthrough in this space, Neural Style Transfer [24]. Based on recent advances in neural networks, their method transfers certain neural network correlation statistics from a painting to a photograph, thus producing a new painting of the input photograph (Figure 8). The method is still “shallow” in a sense — there is no “understanding” of the photograph or the artwork — but the method was more robust than the classical Image Analogies results in many cases. This paper led to a flurry of excitement and new applications, including the popular Prisma app and Facebook’s Live Video stylization, as well as many new research papers improving upon these ideas. This work is ongoing to this day.

Figure 8: The Neural Style Transfer Algorithm [24], which stylizes a photograph in the style of a given artwork; in this case, Van Gogh’s The Starry Night. This algorithm has led to numerous new apps and research in stylization.

Another development which received considerable attention in 2015 was the invention of DeepDreams by Alexander Mordvintsev [50], who, developing a visualization tool for neural networks, discovered that a simple activation excitation procedure produced striking, hallucinatory imagery of a type we had never seen before. There are many other current projects, particularly those around Generative Adversarial Networks [27] and Project Magenta at Google [48], that also show promise as new artistic tools. For example, Figure 9 shows images that we generated by visualizing trends learned by a neural network from a large collection of artistic images in different styles. Though automatically-generated, these image abstract visual concepts in a way that appears to be novel.

Figure 9: Images that we generated using a neural network trained on different subsets of a database of artistic imagery [74]. Each one is meant to typify a single stylistic category or a medium.

1.6 Algorithms are artists’ tools

In every technology that we currently employ — whether photography, film, or software algorithm — the technologies and algorithms we use are basic tools, just like brushes and paint.666This view reflects, I think, the conventional wisdom in the computer graphics research field, which is my research background. Graphics research has always had close ties with certain artistic communities, especially computer animation and visual effects. Many individuals in the field are experienced in developing tools for artists, or in writing code as artworks. The field is often resistant to attempts to automate creative tasks. In contrast, artificial intelligence researchers use terminology much more aspirationally, historically using words like “intelligence,” “learning,” and “expert systems” in ways that have been different from the human versions of these things. The same is true for the new AI-based algorithms that are appearing. They are not always predictable, and the results are often surprising and delightful — but the same could be said for the way watercolor flows on the page. There is no plausible sense in which current systems reflect “true” artificial intelligence: there is always a human behind the work.

Applying the same standard to the current research in neural networks and neural style transfer, it would seem equally perverse to assign authorship of their outputs to the software itself. The DeepDream software was authored by a human; another human then selected an input image, and experimented with many parameter settings, running the software over and over until obtaining good results. Indeed, in a recent art exhibition meant to promote these methods and their exploration [71], human artists were credited for each of the individual works. The same process of selection of tools and inputs, adjusting settings and even modifying code, and iterating until a desirable output is produced, occurs in all current computer artworks.

Unfortunately, there has been a considerable amount of media hype around AI techniques; in the news media, algorithms are often anthropomorphized, as if they have the same consciousness as humans (e.g., [75, 19]), typically accompanied by an image of The Terminator, or even described as artists [55]. (Often, the hype mainly appears in sensationalist headlines, presumably was chosen by the editors, not the journalists, in order to get more clicks.) In fact, we do not really know what consciousness is (despite many theories), or what it would mean to embody it in an algorithm.

Rodney Brooks has written an insightful essay deconstructing common errors in making predictions about AI [6]. One of his key concepts is the notion of “suitcase words.” When we as researchers speak of “training” an algorithm, or an algorithm that “learns,” it is easy to interpret this as being the same thing as human learning. But these word mean quite different things in the two contexts. At present, nearly all “learning” algorithms are essentially data-fitting procedures. The human specifies a set of data to be fit, and a model with parameters to estimate, and the algorithm fits these parameters (usually, with considerable trial-and-error on the human’s part). But the algorithm has learned no general purpose understanding of the world. It is like a tourist in a foreign country that can repeat and recombine a few phrases remembered from the phrasebook, but lacks true understanding of the language or culture that it is visiting. The way that our current classifiers still fail on cases that violate human common sense is a sign that they are not learning good conceptual representations of the world.

There are, indeed, fascinating parallels between human learning and machine learning, and it does seem likely that humans are, in some way, machines for optimization of evolutionary principles [3, 42, 70]. But going from these high-level analogies to actual machine intelligence is a problem for which the solution is not even on the horizon.

In short, in our present understanding, all art algorithms, including methods based on machine learning, are tools for artists; they are not themselves artists. AI artists are currently a science fiction that I’ll discuss more in Part 2.

1.7 Technology helps art stay vital

Rather than being afraid of the new technologies, we should be enthusiastic about the new artworks that it will enable artists to produce. When we think of art of having external influences, we normally think of social or political influences, but ignore the effect of new tools. In contrast, I argue that, in the 19th and 20th Centuries, technological developments have played a pivotal role in advancing art, in keeping it vital and injecting fresh ideas. The stories I gave of photography and cinema include many of examples of this. However, the effect is far more widespread.

One of the most important breakthroughs in the history of Western art was the invention of oil paint by Flemish painters such as Jan van Eyck in the 15th century [13]. Previously, painting had been done primarily with tempera, which lack subtle coloration, and fresco, which was very cumbersome to paint. Oil paint had existed in some form for centuries before, but van Eyck and others found new techniques and compounds that gave them a very practical new medium that was fast drying and allowed them to create richer colors, sharp edges, hard surfaces, and a much wider gamut of colors. The rich light and color that we associate with the Northern Renaissance and the Italian Renaissance are due to this technology (Figure 10).

(a) (b)

Figure 10: The development of the oil painting technology changed painting as an art form. (a) Fresco painting by Michaelangelo on the Sistine Chapel (ca. 1508). The fresco process was difficult and achieved limited tonal range. Today, fresco is defunct as a medium. (b) Oil painting by Jan van Eyck for the Ghent Alterpiece (ca. 1430). Much richer colors and lighting are possible with oil paint.

In each decade since the 1950s, many of our culturally-important works used technology that had only been invented within the previous ten years. For example, most technology used in today’s feature films did not exist ten years ago (e.g., widespread use of HD digital cameras; facial performance capture); the same goes for artworks using smartphones and crowdsourcing; artworks involving white LEDs and Arduino controllers; DJs performing on stage behind their laptops; and so on. Even the most vérité-seeming romantic comedies frequently involve recent digital video editing and digital backdrops.

Conversely, artistic styles that fail to change often become stale and lose their cultural relevance; the adoption and exploration of new technology is one of the main ways that art stays vibrant. For example, the introduction of synthesizer music into 1980s pop music created a new sound that was exciting and modern. The sound diversified as the tools improved, until grunge became popular and made the 80s synthpop sound seem superficial and old-fashioned. Nowadays, a recent revival of 70s and 80s instruments by bands like Daft Punk and LCD Soundsystem seems most exciting at times when they are creating new types of music using old instruments. In contrast, the swing music revival of the 1990s never went anywhere (from bands like Big Bad Voodoo Daddy and Squirrel Nut Zippers) in my opinion, because the bands aped classic styles with classic instruments, without inventing anything particular original themselves.

In each era, radical technological innovations are met by artists with both enthusiasm and rejection. For example, when the Moog synthesizer became popular, it was adopted by big-name bands like Emerson, Lake, and Palmer. Other bands felt that twisting knobs to make music was “cheating:” Queen’s album covers proudly state that the band did not use synthesizers. Robert Moog described one New York musician who said of the instrument “This is the end of the world” [65]. It now seems silly to imagine that people might have ever categorically objected to synthesized music, or to the scratching and sampling of hip-hop DJs, just as it now seems silly that people once rejected waltzing, the Impressionists, and the Rite of Spring as invalid or immoral.

In addition to stimulating professional artists, new tools make art more accessible to larger portions of society. Photography was accessible only to the most determined early adopters, but has continually become easier, faster, and more compact, to the point where nearly everyone carries a mobile phone camera in their pocket or purse. The same goes for the tools of cinematography (from hand-cranked to cameras to handycams to iPhones), and so on. Modern computers give nearly everyone access to digital equivalents of darkrooms, mixing studios, painting studios, and so on; these were formerly highly-specialized technologies requiring laborious effort.

Figure 11: Les Paul, inventor the solid-body electric guitar in 1943; Robert Moog, pioneer of the electronic synthesizer, respectively. Their technologies transformed popular music in ways they could not have foreseen.

1.8 The general story of jobs and technology

Concerns of how technology displaces jobs has been around since at least the 19th century, when Luddite protesters destroyed mechanical weaving machines, and, in folk songs, John Henry competed against a steam drilling machine. Yet, despite many technological disruptions, overall unemployment has not continually risen over the past 200 years. This is because, as old roles are erased, many more arise in their stead. At any given time, while it is easy to imagine losing specific jobs, it takes superhuman imagination to forecast what new opportunities will be created by transformative new technologies. Nowadays, most of us do jobs that would be hard to even explain in detail to a 19th-century worker.

The real concerns should not be about technology, but whether the economic system shares the benefits of new productivity fairly across society versus concentrating wealth only among the very richest [68]. When displacements due to new technology occur, their effects can be eased by social safety nets and better educational foundations (for employment flexibility and retraining). Conversely, a society which fails to distribute wealth and economic gains fairly has much bigger problems than just the impact of AI.

1.9 What does this tell us about the future?

Some general trends around the evolution of technology and art seem to be quite robust. As discussed above, current AI algorithms are not autonomous creators, and will not be in the foreseeable future. They are still just tools, ready for artists to explore and exploit. New developments in these tools will be enthusiastically adopted by some artists, leading to exciting new forms and styles that we cannot currently foresee. It is possible that some tasks performed by human artists will gradually fade out, but these will generally be mechanical tasks that do not require much creativity because they fill societal functions other than artistic expression. Some traditional arts may fade simply due to seeming old-fashioned; this is the nature of art: nothing is fresh forever, which is not to be blamed on technology. But, conversely, new technologies will enable new styles, aesthetics, and job descriptions. Novices will have access to new simplified tools for expression. Artistic technology is a “imagination amplifier” [9] and better technology will artists to see even further than before.

Aside from general trends, it is hard to make specific predictions about the art of the future. Les Paul, who invented the solid-body guitar in the 1940s, himself primarily performed light pop, country, and showtunes, and could hardly have predicted how the electric guitar would be used by, say, Led Zeppelin, just as it’s hard to imagine Daguerre predicting Instagram. More generally, making predictions about how AI technologies might transform society is very hard because we have so little understanding of what these technologies might actually be [6]. Even the science fiction writers of the 1950s and 1960s completely failed to imagine the transformative power of the Internet and mobile computing [40]; for them, the computers of the future would still be room-scale monstrosities that one had to sit in front of to operate. But they did predict moon colonies and replicants by 2018.

In short, we cannot predict what new inventions and ideas artists will come up with in the future, but we can predict that they will be amazing, and they will be amazing because they make use of technology in new, unpredictable ways.

Part 2: INTENT

As discussed, authorship of all current algorithmic art is assigned to the human author behind the algorithms or the users of the tools. I now turn to more philosophical and hypothetical discussion: could we ever say that an AI itself created art? Will we ever recognize a piece of software as the author of a work of art?

Hypothetically, if we ever develop AI with human-level artificial intelligence and consciousness, it ought to be able to create art. But, as discussed in the previous section, this scenario is science fiction and we have no idea if this is possible or how it would be achieved. Making meaningful predictions about this world is impossible [6], because we have so little idea of how specifically this AI would actually operate. Moreover, this AI would transform society so much as to make it unrecognizable to us. We may as well speculate about what kind of artwork will be made by outer-space aliens when they come to visit.

Hence, the interesting question is whether there could be computer-authored artwork without human-level artificial intelligence or consciousness.

2.1 What is art?

The answer to whether computers can create art depends on what we mean by “art;” different definitions could lead to different answers. Definitions of art could provide helpful clues to the question posed in this section.

For art world outsiders (i.e., most people), today’s definition of art seems to involve several overlapping but competing ideas: the image of the artist as lone genius struck by a mystical inspiration (i.e., “it’s art because I made it”); conceptual art, an inscrutable and bizarre enterprise (i.e., “it’s art because it’s weird”); art as technical mastery of a skill like realistic painting (i.e., “it’s art because it’s difficult”); art as financial instrument and status symbol (i.e., “it’s art because art collectors will pay for it”). There is some truth to each view, but none of them is complete.

The precise contours of what activities we define as art are culturally determined: the definition of art has changed over centuries and differs between societies. We think of classical visual art as highly-cultivated expressions of technical skill and cultural statements (e.g., moral lessons or statements of wealth and status), such as Raphael’s paintings in the Papal Apartments of the Vatican Palace. Modern art allows conceptual forms unconnected to technical skill; it seems unlikely that a Renaissance noble would recognize Piero Manzoni’s 1961 famous canning of his own feces as artwork. Yet, some of the cans have been shown in major institutions and sold for tens of thousands of dollars each (two markers of status as art). Likewise, many practical crafts have passed from roles as everyday activity to the status of fine art, e.g., with the exhibition of famous fashion designers in fine art museums.

(a) (b)

Figure 12: The ever-broadening Western conception of art. (a) In Renaissance art, such as Rafael’s Vatican Palace paintings, beauty, technical skill, and, often, moral/religious instruction are all embodied in an artifact. (b) Modern works defy some or most of these requirements, like Marina Abramović’s performance art The Artist Is Present (2010), which is not a physical artifact.

Modern definitions of art.

In the modern era, there have been several attempts to form concise definitions of “art” that capture everything that we describe as art, including both representational painting and conceptual art. The prototypical conceptual work is, arguably, Marcel Duchamp’s 1917 “Fountain,” an ordinary urinal that he inverted, signed with a fake name, and submitted to an annual exhibition. The Fountain made a clear statement that the artist’s concept and intent are all that are needed to create an artwork; no technical skill or significant effort are required. In the rest of the century, artists explored this idea and pushed the boundaries of precedent even further. Marina Abramović’s performance artworks produce no physical artifacts, but are still considered artworks that are “exhibited” in fine art museums; conversely, Robert Smithson’s earthworks exist only in specific locations outside of gallery spaces. Outsider art and folk art are specifically the work of untrained amateurs, such as Grandma Moses and Henry Darger. A current definition of art must include all of these examples… and also, dance, music, literature, and so on.

Modern definitions of art are all meant to cover the same classes of works, and so they may all be valid; indeed, it seems unlikely that there is one single, compact definition [52]. However, some definitions may be more “predictive,” and thus give us more guidance about what might be valid art in the future.

Institutional definitions and an inclusive definition.

Of the modern definitions, a good baseline definition of art is the Institutional Theory of art: art is anything that is culturally accepted as art by society’s cultural institutions [12, 15]. A variant on this theory is Levinson’s Historial Definition of art: “a work of art is a thing intended for regard-as-a-work-of-art: regard in any of the ways works of art existing to it have correctly been regarded.” [44] This definition allows for art styles to evolve, but not the mechanisms of appreciation, e.g., a pleasurable drug is not art because it never has been art. Modern theories have the role of the human artist — and their intent — as implicit to varying degrees, but they never consider the possibility of “art” not created by humans. The question posed in this essay then comes down to the question of whether computer-authored works would ever be regarded as art.

My own version of this definition of art as: the only prerequisite for being an artwork in the fine art world is to be some specific expression, artifact, or activity designated by a human as art; that human is the artist. This definition differs in that it makes the role of the human explicit, and it does not require cultural institutions or broad acceptance, merely the statement by that person that their activity is art. This definition includes everything that we currently consider as art. It may be over-broad, but not in ways that matter, particularly for the discussion here. And, by this definition, computers cannot create art, because it assumes a human artist.

This definition of art is extremely inclusive. A five-year old scribbling in crayon is making art. A teenager moodily shooting BBs into a lake could be described as performing art. An angry driver swerving in traffic dramatically could describe their act as art. Of course, just because we could call an activity art does not mean that we want to — my point is that there are few meaningful boundaries for what could be called art. For the teenager shooting BBs into a lake, most people wouldn’t bother to call it art, and there’d be no reason to do so. It would not satisfy the institutional definition of art that I gave above. However, if that teenager happened to have taken a lot of art classes and wanted to sell tickets to their situationist performance about environmental degradation, I would not waste my time denying their work the status of “art.” Instead, if pressed, I would question whether the work was worthwhile on any axis: is it interesting? Is it thought-provoking? Is it impressive? Does it say something interesting about current topics?

The purpose of my definition is avoid wasting time on debates about what human-made thing is or isn’t art. As a rule-of-thumb, if you someone presents a thing as their artwork, then it almost certainly is. The important question is, almost always, not “is it art?” but “Is it good?” How successful is it on different axes, e.g., is it technically impressive? Is it interesting or provocative? Is it beautiful? Is it immoral? How much money is it worth? etc.

These definitions have the advantage of being broadly true, but have the disadvantage of being so vague as to be almost tautological. Quite understandably, they generally assume that art is the product of a human artist without discussing alternatives. They provide little guidance to the question at hand. The Institutional definition tells us that anything that people accept as art is art, but doesn’t tell us when people will accept something new as art.

Cluster Concept.

In search of a more useful definition of art, more recent philosophers have proposed that art can be defined as a “cluster concept.” [25, 26, 17, 18]. A cluster concept is something defined by a list of properties that a work may have, such as being a source of pleasure, requiring techincal skill, expressing of emotion, being bracketed off from ordinary life, having novelty and creativity, creating “imaginative experience” in the audience, and so on. In this view, no single property is a requirement for something to be called art, but it must have most of the properties. In addition, these authors also state some additional required criteria for artwork, including being a “product of an action” [26] and intended for an audience [18].

Overall, the Cluster Concept gives only limited guidance for our problem. Some of these properties could be viewed as prerequisites for computer art, e.g., the ability to explore deep themes in new and unexpected ways. But the Cluster Concept definition is really meant to list attributes that distinguish, say, spectator sports from theatre. It does not tell us when we have a worthy automatic playwright.

The problem with definitions of art.

These definitions are attempts to concisely describe those things that we currently call art. They all assume a human artist; the philosophers behind these discussions do not much discuss this requirement, since there are no other cases to consider. But art is a cultural practice, and definitions of art follow the practice, not the other way around. (In machine learning terminology, these definitions cannot generalize to non-human art, because they were trained only on examples of human activity as art, i.e., they are overfit.)

Hence, the real question is not “Does AI-generated art fit the definition of art?” but “Will the definition of art ever be expanded to recognize AIs as artists?”

2.2 Other useful cases

There are a few specific cases of art/non-art that are relevant to these questions.

Not art.

It might seem that, in the art world, anything goes, and anything can be called art. But there are some instructive counterexamples.

Natural processes, including landscapes like the Grand Canyon or the HuangShan Mountains, are not art, even though they may be extraordinarily beautiful and change one’s perspective immensely. Beautiful structures and mating rituals made instinctively by animals, such as honeycombs and birds’ songs, are not considered art.

In some cases, higher mammals have been trained to paint, including chimpanzees, elephants, and dolphins [14]. Often, the animal’s owner or handler steers the process, e.g., stopping the painting when they believe it is done and selecting which works to show. It often seems that the animal is doodling on the canvas while the human is making the editorial decisions that identify lucky doodles to be called artworks. Animal artwork has not had any significant cultural impact or popularity; it seems to have been largely sensational stunts. (People for the Ethical Treatment of Animals have recently tried to claim copyright in favor of a monkey, but failed [49], as US copyright law only allows humans to claim copyright.) Based on this, together with evidence that human guidance is usually involved, I am skeptical as to whether we should consider any of this to really be art. More likely, it should be considered the art of the zookeeper and not the animal.

Despite attempts to promote animal artwork, human intent seems to be crucial for the time being. However, the discussions around animal artwork are empirical rather than definitional: is the thing that the monkey is doing art? Does the monkey conceive of the artwork as a thing in itself? Does it care about aesthetics, beauty, or visual expression? These are empirical questions. It’s not that the human element is really fundamental, just that we haven’t yet found any creatures (or robots) that seem to be making something that we consider to be real artwork.

Figure 13: Not art. (a) Beautiful landscapes are not considered art. (b) Pierre Brassau, the monkey painter. He was part of a 1963 hoax that satirized modern art.

Conceptual art.

For modern and conceptual art, how that work speaks to current cultural debates is crucial to the artwork. Marcel Duchamp’s Fountain would not have meant anything to a Renaissance noble; if presented as an artwork today, it would seem like a trite stunt. But, in 1917, it had enormous impact on discussions around art and we understand art. Conceptual pieces are significant for their place and time, and would not have been considered serious art at some other time. Many of these artworks attract attention and high prices simply because they provoke discussion and controversy, and pose fascinating riddles about the nature of art. As a recent example, Damian Hirst’s 2017 exhibition “The Wreck of the Unbelievable” left my head spinning for days with different interpretations and assessments, and wondering whether one can simultaneously be an artistic genius and a craven charlatan.

In other words, while anything, in form, could be art, effective conceptual art is highly tied to the specific content in which it is presented, and, often, the artists’ statements around it.

In a way, conceptual art is the distillation of years of tradition, reducing art to “pure intent;” skill, technique, and aesthetic pleasure may be irrelevant to it.

Collaborations, contributions, crowdworking.

In some cases, multiple people are involved in the creation of an artwork, which is instructive for how we might credit humans working with AI or crowdworkers. First, artworks can obviously be collaborations between multiple people. Equal collaborators usually get equal credit (e.g., Leonard Bernstein and Stephen Sondheim); in a more hierarchical relationship, e.g., a movie director and their actors and crew, the director gets authorship, though the contributions of actors and crew get recognized as well. Sometimes artwork involves collaboration with technicians that are not really considered artists themselves, such as camera designers and manufacturers (and sometimes the members of an artists’ atelier simply aren’t credited).

Crowdworking involves automating requests for many small tasks from online human workers, and has been described as a technique to solve AI-hard problems with human workers, essentially, turning people into algorithm subroutines [45]. Scott Draves’ Electric Sheep artwork runs a genetic algorithm that optimizes procedural patterns, with a fitness signal provided by hundreds of thousands of users voting on which patterns are most appealing [16]. Aaron Koblin’s Sheep Market [38] involved many crowdworkers paid two cents for each sheep drawing. In each case, credit for the artwork goes to the coordinator/creator, not the crowdworkers. (As livestock-themed Internet-scale works go, Ian Bogost’s satirical Cow Clicker video game [69] is also singular.)

An extreme and controversial example is Richard Prince [47], a conceptual artist who directly copies others’ work and regularly sells it for millions; the other photographers that he directly copies do not receive any of the credit or profit (except, potentially, through lawsuits).

These examples all show that, in our current definitions of art, no matter what contributions of the work came from other people or machines, the artist who conceives and directs the work gets ultimate authorship credit. Hence, it is not enough for an AI to create most of the work to be an author of it. Conversely, an AI that directs the work, but relies on crowdworkers to execute the individual tasks, could be considered an artist.

“Art” as colloquial judgment.

The term “art” is often used in another way, i.e., to separate good from bad or to elevate the ordinary to the exalted. I recall one of my college art professors, in several instances enthusiastically declaring one amateur or another as “a real artist!” For example, he gave this praise to Scott Weaver, a hobbyist who spent many decades building an awe-inspiring model of the San Francisco Bay Area out of toothpicks and Elmer’s glue. Likewise, one might praise their friend’s drawings by saying “you are a real artist!” In these cases, “art” and “artist” are judgements of technical skill. Paradoxically, this usage is both inclusive — it allows untrained hobbyists to be elevated to the status of artist — but also elitist because it excludes people based on skill. When spoken by, say, an artist or professor, it furthermore seems meant to elevate the speaker, to show that they are authorized to decide what is or isn’t art.

I think this usage is risky as a definition of art. Here is an analogy. Suppose that it were common, when describing a movie that one does not like, to say that it “wasn’t a movie.” For example, when coming out of, say, “Star Wars I: The Phantom Menace”, many people would say that it wasn’t a movie. A few fans might say that it was a movie, and then they would get into arguments about whether it was or wasn’t a movie. Each person might have their own slightly different idea of what makes something a movie or not. This seems like a confusing and ridiculous argument to have: we should be arguing over whether it was a good movie: whether it was fun, meaningful, exciting, or provocative. When arguing over whether or not something is an enjoyable movie or not, the subjectivity inherent in the question is obvious, but not when arguing whether or not it is a movie at all. Arguing over whether something is or isn’t art seems equally ridiculous.

Using an inclusive definition of art greatly improves communication. For example, getting into debates about whether something is or isn’t art is usually a quagmire. Saying that, say, a work by Thomas Kinkade or Damien Hirst “isn’t art” is phrased as an objective judgement. But it is really a subjective choice, e.g., it asserts that Kinkade or Hirst’s work is unoriginal, superficial, or pretentious. The objective-seeming statement hides the subjective nature of the judgement, and obscures the real reasoning behind the judgement777An additional disadvantage of art as judgment is that it discourages amateurs from pursuing art as a hobby, without risking the possibility that their amateur watercolor or crocheting isn’t art. Everyone can create art, even if only for their own satisfaction and enjoyment, or for sharing with friends and family; it need not be museum-quality to be a worthwhile activity..

2.3 What would it take to recognize a software artifact as an artist?

The fact that there is debate over whether animals’ painting are art, and whether AI could create art, indicates that we are open to the possibility of expanding the definition — or that there is a more fundamental definition of art that we have not yet articulated. Here I’ll explore a few possible criteria that might describe our “real” criteria for artwork — or what the “real” definition might be.

Social agents.

Why do we make art? Dutton [18] argues persuasively that creating art is a product and outgrowth of our biological evolution. Creating art serves several functions, including communication and sharing information, displays of skill (e.g., for mating), and other social functions. Since creating art is a fundamentally social act of expression and communication,888It is true that one often creates art for one’s own amusement and experience. However, one may also sing or talk to oneself while alone, but talking and singing are still fundamentally social activities. it makes sense that an art creator can only be a human — or something else that we have a social relationship with.

So, perhaps, the intent required for artwork is the intent of a social agent to express to, share with, and/or create responses in another social agent. Since we often have social relationships with animals, animals could create art if they were communicating through that art, or creating beauty for others to appreciate. For AIs, we would need to perceive them as social agents.

The extreme version of this statement is: we want to believe that the AI creating the art is an independent being, with its own consciousness, emotions, and things to say.

Shallow AI agents.

People are sometimes “fooled” by shallow AI. The classic case is Eliza, a simple text-based “psychiatrist” program developed in 1964, based on simple pattern-matching and repetition of what the user types [73]. It was meant as a demonstration of the superficiality of the AIs of the time, but, unexpectedly, many people attributed human-like emotions to the machine. Since then, many “chatbots” have fooled humans in online settings. But, once the veil is lifted, it is clear that these chatbots don’t exhibit real intelligence.

A related effect is that people behave toward their computers as if they were social agents in certain ways [57], even when they don’t believe that they are intelligent. For example, dialogue systems like Siri and Alexa all use female voices by default, based on many findings that people (both male and female) respond better to female voices [29, 57].

Perhaps, for the “average” user, the system doesn’t need to be truly intelligent, it just has to be perceived as a social agent, like a Siri or Alexa that you can ask to make you an artwork via natural language [76]. At the moment, however, people (aside from headline writers) don’t credit Siri or Alexa with intelligence.

Quality, Creativity, Novelty.

It seems that we would make several more requirements of our artificial artist, that it should produce work that is high-quality, creative, and novel.

In classical and popular art forms, quality requires aesthetic pleasure of some form. Conceptual art must speak to current concerns of the art world or the larger society. (Often, the most successful works achieve both aesthetic and conceptual goals.)

Either way, effective art works must be novel and original in some way, not just a copy or interpolation of existing works. Another way to express the goal of novelty is that the software system has somehow “surpassed” the system’s author’s teachings.

For example, Colton [10] proposes that we judge the creativity of a system, in part, by whether the system’s output surprises the system’s author. I believe that this is too weak a criterion, since many systems may be surprising to their own author, but later explained by someone else. This is a common theme in scientific discovery. For example, analysis and understanding of why deep networks work so well is an ongoing research activity [4].

As a more fundamental example, chaotic systems contain enormous, unpredictable complexity. Consider Mandelbrot set images: the basic algorithm that produces them can be described in a single sentence999Color the image location according to the number of steps the iteration requires to reach , starting from , where is a large constant. yet produce dazzling animations of infinite complexity.101010For example: The Mandelbrot set is very surprising and produces beautiful, unprecedented images, but no one would call its iteration equation an artist.

Here is another way to phrase the standard. At present, we credit the author of a piece of automatic software with the output of that software. This usually acknowledges the skill and effort required to engineer and iterate with software so that it produces good outputs. Giving credit to the software would be depriving the author of their deserved credit. When do we believe that, by crediting the system, we are not depriving the author of credit they deserve? That is, when are they more like a parent than like a director?

This level of independence implies novel creations: not just new instances of a given style, but new styles, forms and continuing creativity. We want the sense that the artist is progressing, and not just doing the same thing over and over again (Mandelbrot images get boring after awhile), and, ideally, responding to the culture and their environment as well.

On the surface, these criteria seem to contradict the highly inclusive definition of art. We no longer put a quality requirement on human art, so why put it on AI art? I think it’s because we need to believe that the AI is intelligent. In this case, having AI create good, original artwork would be the proof, like an artistic Turing Test.


There are a few ways in which the definition of art could change to include AI as artists.

Perhaps a curator at a well-known museum would download or otherwise acquire various artifacts from software “artists,” and list the software systems listed as the authors. There would be controversy, and discussion in newspapers and journals. Perhaps other curators and galleries would follow suit. Perhaps people would find enough value in these computer-generated artworks, while also being convinced that no human could be rightly given credit for their works. This sort of process has happened for things like abstract expressionism, and not for chimpanzee art. Could it happen for computer art?

Perhaps a software company will market image-generating software as “AI artists.” Maybe, with enough market penetration, people will accept these software tools as artists. One can imagine many ways in which average users might use automatically-generated artwork, e.g., generating clip art, personalized imagery (like Bitmoji), and so on. Even though these systems would not really be social, independent agents, and their creativity could be very limited, this sort of software could be marketed as “art.” With enough marketing, and popular success, the definition of “artist” could thus be changed.

There is a danger in with shallow AIs being called artists: it risks misleading people as to the level of intelligence exhibited by the system.

2.4 Can we do it with current software paradigms?

At present, we have several different paradigms for authoring software systems, including AIs that generate images, music, and other artifacts.

The classic paradigm is procedural art, in which a programmer authors algorithms that produce images. This is the standard technique of “classic AI” techniques, such as “knowledge-based systems.” Examples include many of the original procedural and generative artworks, as well as the painterly rendering algorithms in Section 1.5. Optimization is a way to specify constraints on a system rather than specifying the synthesis procedure directly (the classic AI technique of constraint-satisfaction is a special case of this). An example of this is the solvers used in modern design software for adaptive interface layout, e.g., Cassowary; optimization has also often been used in non-photorealistic rendering [33]. In the past decades or so, machine learning has come to define AI, including example-based image-generation methods like Image Analogies and Neural Style Transfer (Section 1.5). Crowdsourcing (Section 2.2) often plays a role in modern systems, either as a subroutine of an algorithm [45], or as a way to produce training data.

Nowadays, most real systems will often be hybrids of these different software paradigms. For example, we might engineer an architecture that places brush strokes, but uses a learned model to evaluate the system, or we might use a neural network to directly shade pixels, but define a procedure for the color selection [23]. Occasionally, systems authored this way are meant to be “creative” or to intended someday to be seen as artists [21, 11].

Can we create an “artist” using our current programming methodologies?

Whenever we design software with our current systems, we understand and architect them to achieve specific tasks. In Section 2.6, I’ll sketch out what an artist-AI might look like in this system. Despite the use of machine learning, every stage of the process is carefully designed and experimented upon. There is no sense in which the system is independent. These software systems are not true social agents, or independent beings with consciousness or intelligence.

One big missing element in current art experiments is the notion of novelty: systems that can go beyond a fixed space of concepts or styles, and innovate in a meaningful way. Most machine learning is essentially sophisticated data interpolation, and extrapolation is much harder than interpolation. A machine that drew input from current trends, and could effectively “trendspot” to identify good trends might be the closest we could get.

The Creative Adversarial Network [21] produces very intriguing and “artistic” images. This is very nice work, but it is presented as a form of artificial creativity, and I disagree: this is just optimizing a human-designed energy function to remix trained models, using the image biases of convolutional neural networks [72]. If anything, “creativity” is another “suitcase word”, and the system is “creative” in a narrow, technical sense, not in a human sense.

Hence, by the standards of social AI, or of creativity and novelty, I do not believe that it is possible to create true AI artists with our current paradigms. It may be possible to create “shallow AI” systems that are marketed as artists.

2.5 Future paradigms

Several future goals drive AI research. We do not know if they are achievable, or what they would precisely look like. But we can hypothesize whether, if achieved, they would let us create “artists.”

Emergent intelligence.

Emergence is a property in which simple rules lead to complex aggregate behaviors. For example, complex swarming behaviors for bird animations can be defined from very simple rules for each bird without requiring global coordination [58]. Human intelligence must be, in a way, an emergent property: we can understand the workings on individual axons and dendrites and chemical signalling, but studying these elements on their own does not tell us how intelligence works. It is possible that “intelligence” could emerge from a system designed using the current paradigms.

I am skeptical of this. Conversational agents like Siri and Alexa give us a sense of the extent of emergent intelligence with current technologies: they have super-human ability to draw on vast stores of human knowledge to answer questions. While their implementations are not public, there is no one software engineer at these companies who could explain everything that the system does in full detail, but, overall, the overall system architectures must surely follow familiar patterns, and the systems are clearly constrained to avoid creativity or originality. (They do include some whimsical answers to certain questions that are presumably written by humans.)

Artificial general intelligence.

There are many ways to define intelligence, to state the goals of AI research [61].

One definition is that a system should be able to perform general tasks without requiring retraining. Such a system would have learned some representations of underlying entities and relationships in the world, and would be able to make general inferences and predictions about these entities. (The wide range of simple adversarial attacks, as well as frequent failures on cases that should be obvious, show that current neural networks are far from achieving this.) Such a system could learn from the worlds’ artworks to reason about styles, contexts, and artists’ intents. It could generate high-quality new images according to instructions, e.g., generate interesting images in a new style that is similar to existing styles but not too similar. Such a system could be an amazing tool for exploring and creating art. It would provide amazing capabilities to novices. But I do not believe that it would be an artist — unless marketing and PR convinced people otherwise.

Human-Level or Conscious AI.

An artificial intelligence that had human-like consciousness and independence would, by definition, be able to create art. However, this is utter science fiction, and nothing like this is currently on our horizon.

2.6 Building an Intent Machine

In order to make speculations concrete, let us now conduct a thought experiment of building a computer artist. This artist must be able to generate an “intent.” I suspect that actually defining what we mean by intent, in a comprehensive and complete way, is very hard. For some artworks, the intent might be very hard to articulate, or a matter of speculation. However, no artist needs to be able to have every possible intent, so we just need to create a sufficient rich set of “intents.” I will focus on representational painting, in order to narrow the scope of the discussion. (A few previous systems have tried to create intent, e.g., [41]).

Intent representation.

I first need to define, in some way, a set of possible intents, and how I’d represent intent in software. Let’s say that the system’s intent can be one of the four possible goals: to depict a specific scene and the feeling that it conveys; to express the artist’s emotional state; to make a statement through the scene (e.g., glorifying nature, portraying the city as exciting or as disgusting); to make a conceptual statement (to glamorize or protest an event, or a concept, or a person). The intent could be represented by a simple data structure (e.g., ), an English-language sentence (“I wish to express melancholy over human mortality”), or a learned feature encoding (e.g., non-linear embedding) that captures hard-to-articulate concepts inferred by latent variable modeling and weakly-supervised learning. Recent methods that convert from images to English-language descriptions [36] or vice versa [77, 76] could be inspirational here.


As for execution on the intent, there are many options. We could train a neural network model that takes a given “intent” and a photograph, and creates an appropriate painting for simple concepts; or, if this model isn’t good enough, the system could hire a professional through an online crowdworker platform in order to execute on the intent. The input image itself could be automatically selected from online image collections, instead of being provided to the system. The system might also include a classifier or ranker to identify which artworks it produces are the best. Although the synthesis model might not be very good at first, as we have seen, authorship of artwork is defined by the intent and editorial choices, not by execution.

Intent generation.

The simplest intent generator would be hand-authored. I could write some lines of code that, given an image, randomly selects which of the four types of intents to apply, and then randomly generates appropriate qualities to go with it (e.g., which emotional state to depict). I might make the sampler conditional on the image content and photographic style.


The next step would likely to be train better models of intent. Perhaps I would scrape artist statements from the web, or from analysis of Wikipedia pages, or, more likely, pay crowdworkers to annotate real-world artworks with their intents according to a rubric that I’d specify. I might train a model that infers latent parameters of intent and style that can’t be manually annotated. Even better would be to first generate an overall meta-intent and style, representing this artist’s particular inclinations, and a natural-language “Artist’s Statement” expressing this concept, and then a collection of individual artworks that elaborate on that theme. The system would also need to automatically select starting images from online social images, or, better yet, generate them with a generative model (e.g., a Generative Adversarial Network [27] or Variational Autoencoder [37]). The system could also be refined by using crowd workers to evaluate, label, and score its outputs, in order to improve each step of the process. Many other continued refinements would be possible.

I could then submit the generated artworks to a gallery or art exhibit, describing the process by which they were created, and claim that the software is itself the artist.

I don’t believe that this resulting software artifact would be recognized as an artist. It is not a social agent, it is not creating intents that are culturally relevant, though it could give the appearance superficially. It is too easy to understand the system architecture as designed for a certain model of art.


One of my main goals in this essay has been to highlight the degree to which technology contributes to art, rather than being antagonistic.

Fears of new technology seem to be human nature. I suspect many people view the “normal” state of things as being how they were when they came of age, and they view any significant change as scary. Yet, nearly all of our familiar modern technologies were viewed as threatening by some previous generation.

The fear of human-created life has been with us for a long time. Notably, 18th-century scientists discovered electricity. As they searched to understand it, they discovered the life-like effect of galvanism, that the muscles of dead frogs could be stimulated by electrical currents. Had the secret of life been discovered? This inspired Mary Shelley’s novel Frankenstein; or The Modern Prometheus in 1813, in which a student uses modern science to create new life [63]. Today, the story is vivid and evocative, but, intellectually, we recognize it as preposterous. The fear of AI is essentially the same irrational fear; SkyNet is Frankenstein’s monster, but with neural networks as the Promethean spark instead of galvanism.111111In fact, Frankenstein is presented as a cautionary tale about the quest for knowledge in general. Victor Frankenstein tells his story as a warning when he learns that Captain Walton is himself driven by an obsessive quest for knowledge that is entirely unrelated to Frankenstein’s. At present, the autonomous AIs of “The Terminator” are only slightly more plausible than their ability to travel backwards in time.

There are real dangers to current AI technologies, particularly, misuse of big data and machine learning that inadvertently magnifies existing inequality and unfairness in our social and governmental structures [53, 5]. These are problems not with technology per se, but with how it is used by humans and organizations.

I do not believe that any software system in our current understanding could be called an “artist.” But this could someday change. I mean this as a warning against misleading oneself and others about the nature of art. However, the ambitious reader could take this as a challenge: I have laid out some of the serious objections that you must overcome if you wish to create a software “artist.” I don’t think it can be done anytime soon, but I also know that proving critics wrong is one of the ways that art and science advance.

We are lucky to be alive at a time when artists can explore ever-more-powerful tools. Every time I see an artist create something wonderful with new technology, I get a little thrill: it feels like a new art form evolving. Danny Rosin’s Wooden Mirror, Jason Salavon’s The Top Grossing Film of All Time, 1 x 1, Bob Sabiston’s Snack and Drink, Michel Gondry’s Like A Rolling Stone, Kutiman’s ThruYOU, Amon Tobin’s Permutation, Ian Bogost’s Cow Clicker, Christian Marclay’s video installations, Íñigo Quilez’s procedural renderings, and Wesley Allsbrook’s and Goro Fujita’s Virtual Reality paintings are a few examples of artworks that have affected me this way over the years. Today, through GitHub and Twitter, there is an extremely fast interplay between machine learning researchers and artists; it seems like, every day, we see new tinkerers and artists Tweeting their latest creative experiments with RNNs and GANs (e.g., @JanelleCShane, @helena, @christophrhesse, @Salavon).

Art maintains its vitality through continual innovation, and technology is one of the main engines of that innovation. Occasionally, the avant garde has tremendous cultural impact: electronic music and sampling was once the domain of experimental electronic and musique concrète pioneers, like Wendy Carlos and Delia Darbyshire. Likewise, at one time, computer-animated films could only be seen at obscure short-film festivals. Today, we are seeing many intriguing and beguiling experiments with AI techniques, and, as artists’ tools, they will surely transform the way we think about art in thrilling and unpredictable ways.




  • [1] Agüera y Arcas, B. Art in the Age of Machine Intelligence. Arts 6, 4 (2017).
  • [2] Albright-Knox Art Gallery. International exhibition of pictorial photography. Retrieved January 10, 2018.
  • [3] Alexander, R. M. Optima For Animals. Princeton University Press, 1996.
  • [4] Bau, D., Zhou, B., Khosla, A., Oliva, A., and Torralba, A. Network Dissection: Quantifying Interpretability of Deep Visual Representations. In Proc. Computer Vision and Pattern Recognition (2017).
  • [5] Bolukbasi, T., Chang, K.-W., Zou, J. Y., Saligrama, V., and Kalai, A. T. Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings. In Proc. Neural Information Processing Systems (2016).
  • [6] Brooks, R. The Seven Deadly Sins of AI Predictions. MIT Technology Review (2017).
  • [7] Catmull, E., and Wallace, A. Creativity, Inc: Overcoming the Unseen Forces that Stand in the Way of True Inspiration. Random House, 2014.
  • [8] Cohen, H. The further exploits of AARON, painter. Stanford Humanities Review 4, 2 (1995).
  • [9] Cohen, M. F. Imagination amplification. IEEE Comp. Graph. & App. (2000).
  • [10] Colton, S. Creativity Versus the Perception of Creativity in Computational Systems. In Proc. AAAI spring symposium: creative intelligent systems (2008).
  • [11] Colton, S. The Painting Fool: Stories from Building an Automated Painter. In Proc. Computers and Creativity (2012).
  • [12] Danto, A. The Artworld. The Journal of Philosophy 61, 19 (1964).
  • [13] de la Croix, H., Tansey, R. G., and Kirkpatrick, D. Gardner’s Art Through The Ages, 9 ed. Harcourt Brace, 1991.
  • [14] Desmond, J. Zoos make money selling paintings made by animals. Are they art? The Washington Post (2016).
  • [15] Dickie, G. Defining Art. American Philosophical Quarterly 6, 3 (1969).
  • [16] Draves, S. The Electric Sheep Screen-Saver: A Case Study in Aesthetic Evolution. In Proc. EvoWorkshops (2005).
  • [17] Dutton, D. But They Don’t Have Our Concept of Art. In Theories of Art Today, N. Carroll, Ed. University of Wisconsin Press, 2000.
  • [18] Dutton, D. The Art Instinct: Beauty, Pleasure, and Human Evolution. Bloomsbury Press, 2009.
  • [19] Dvorsky, G. This Artificially Intelligent Robot Composes and Performs Its Own Music. Gizmodo (2017).
  • [20] Efros, A. A., and Leung, T. Texture synthesis by non-parametric sampling. In Proc. International Conference on Computer Vision (1999).
  • [21] Elgammal, A., Liu, B., Elhoseiny, M., and Mazzone, M. CAN: Creative Adversarial Networks Generating “Art” by Learning About Styles and Deviating from Style Norms. In Proc. Int. Conf. on Computational Creativity (2017).
  • [22] Fišer, J., Jamriška, O., Lukáč, M., Shechtman, E., Asente, P., Lu, J., and Sýkora, D. StyLit: Illumination-guided Example-based Stylization of 3D Renderings. ACM Trans. on Graphics (SIGGRAPH) 35, 4 (July 2016), 92:1–92:11.
  • [23] Gatys, L., Ecker, A. S., Bethge, M., Hertzmann, A., and Shechtman, E. Controlling Perceptual Factors in Neural Style Transfer. In Proc. Computer Vision and Pattern Recognition (2017).
  • [24] Gatys, L. A., Ecker, A. S., and Bethge, M. Image Style Transfer Using Convolutional Neural Networks. In Proc. Computer Vision and Pattern Recognition (2016).
  • [25] Gaut, B. Art as a Cluster Concept. In Theories of Art Today, N. Carroll, Ed. University of Wisconsin Press, 2000.
  • [26] Gaut, B. The Cluster Account of Art Defended. The British Journal of Aesthetics (2005).
  • [27] Goodfellow, I. J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y. Generative Adversarial Nets. In Proc. Neural Information Processing Systems (2014).
  • [28] Goodwin, R. Adventures in Narrated Reality, Part II. Medium (2016).
  • [29] Griggs, B. Why computer voices are mostly female. CNN (2011).
  • [30] Haeberli, P. Paint By Numbers: Abstract Image Representations. In Proc. SIGGRAPH (1990).
  • [31] Hertzmann, A. Painterly Rendering with Curved Brush Strokes of Multiple Sizes. In Proc. SIGGRAPH (1998).
  • [32] Hertzmann, A. Algorithms for Rendering in Artistic Styles. PhD thesis, New York University, 2001.
  • [33] Hertzmann, A. A Survey of Stroke-Based Rendering. IEEE Computer Graphics & Applications 23, 4 (2003).
  • [34] Hertzmann, A. Non-Photorealistic Rendering and the Science of Art. In Proc. NPAR (2010).
  • [35] Hertzmann, A., Jacobs, C. E., Oliver, N., Curless, B., and Salesin, D. H. Image Analogies. In Proc. SIGGRAPH (2001).
  • [36] Karpathy, A., and Fei-Fei, L. Deep Visual-Semantic Alignments for Generating Image Descriptions. In Proc. Computer Vision and Pattern Recognition (2017).
  • [37] Kingma, D. P., and Welling, M. Auto-encoding variational Bayes. In Proc. Int. Conf. on Learning Representations (2014).
  • [38] Koblin, A. M. The Sheep Market. In Proc. Creativity and Cognition (2009), pp. 451–452.
  • [39] Kosinski, D., Ed. The Artist and The Camera. Dallas Museum of Art, 1999.
  • [40] Krauss, L. Why Science-Fiction Writers Couldn’t Imagine the Internet. Slate (2017).
  • [41] Krzeczkowska, A., El-Hage, J., Colton, S., and Clark, S. Automated Collage Generation - With Intent. In Proc. Int. Conf. on Computational Creativity (2010).
  • [42] Körding, K. P., and Wolpert, D. M. Bayesian decision theory in sensorimotor control. Trends in Cognitive Sciences 10, 7 (2006), 319 – 326. Special issue: Probabilistic models of cognition.
  • [43] Lasseter, J. Principles of traditional animation applied to 3d computer animation. In Proc. SIGGRAPH (1987).
  • [44] Levinson, J. Defining art historically. The British Journal of Aesthetics 19, 3 (1979).
  • [45] Little, G., Chilton, L. B., Goldman, M., and Miller, R. C. TurKit: Human Computation Algorithms on Mechanical Turk. In Proc. User Interface Software and Technology (2010).
  • [46] Litwinowicz, P. Processing Images and Video for an Impressionist Effect. In Proc. SIGGRAPH (1997).
  • [47] McDermon, D. You Can Now Buy a $90,000 Richard Prince Instagram Work or a $90 ‘Copy’. The New York Times (2015).
  • [48] Metz, C. How A.I. Is Creating Building Blocks to Reshape Music and Art. The New York Times (2017).
  • [49] Millward, D. Monkey selfie case: British photographer settles with animal charity over royalties dispute. The Telegraph (2017).
  • [50] Mordvintsev, A., Olah, C., and Tyka, M. Inceptionism: Going Deeper into Neural Networks. Google Research Blog (2015).
  • [51] Nicholas, G. These stunning A.I. Tools Are About to Change the Art World. Slate (2017).
  • [52] Novitz, D. Art by another name. The British Journal of Aesthetics 38, 1 (1998).
  • [53] O’Neil, C. Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. Crown, 2016.
  • [54] Paik, K. To Infinity and Beyond! The story of Pixar Animation Studios. Chronicle Books, 2007.
  • [55] Perez, S. Microsoft’s new drawing bot is an AI artist. TechCruch (2018).
  • [56] Price, D. A. The Pixar Touch. Vintage, 2008.
  • [57] Reeves, B., and Nass, C. The Media Equation: How People Treat Computers, Television, and New Media Like Real People and Places. CSLI, 1996.
  • [58] Reynolds, C. W. Flocks, herds and schools: A distributed behavioral model. In Proc. SIGGRAPH (1987).
  • [59] Rosenblum, N. A World History of Photography. Abbeville Press, 1984.
  • [60] Rosin, P., and Collomosse, J. Image and Video-Based Artistic Stylisation. Springer, 2013.
  • [61] Russell, S. J., and Norvig, P. Artificial Intelligence: A Modern Approach, 3 ed. Pearson, 2010.
  • [62] Scharf, A. Art and Photography. Penguin, 1968.
  • [63] Shelley, M. W. Frankenstein; or, the Modern Prometheus. Lackington, Hughes, Harding, Mavor & Jones, 1818.
  • [64] Snow, C. P. The Two Cultures. London: Cambridge University Press, 1959.
  • [65] Snowden, D. Robert Moog: ’i wouldn’t call this music’ – a classic interview to mark a Google doodle. The Guardian (2012).
  • [66] Sterling, B. An Essay on the New Aesthetic. WIRED (2012).
  • [67] Sternberger, P. S. Between Amateur & Aesthete: The Legimitization of Photography as Art in America, 1880-1900. University of New Mexico Press, 2001.
  • [68] Stiglitz, J. E. The Price of Inequality. Norton, 2012.
  • [69] Tanz, J. The Curse of Cow Clicker: How a Cheeky Satire Became a Videogame Hit. WIRED (2011).
  • [70] Tenenbaum, J. B., Kemp, C., Griffiths, T. L., and Goodman, N. D. How to Grow a Mind: Statistics, Structure, and Abstraction. Science (2011).
  • [71] Tyka, M. Exploring the Intersection of Art and Machine Intelligence. Google Research Blog (2016).
  • [72] Ulyanov, D., Vedaldi, A., and Lempitsky, V. Deep Image Prior, 2017. arXiv:1711.10925.
  • [73] Weizenbaum, J. ELIZA—–a computer program for the study of natural language communication between man and machine. Comm. of the ACM (1966).
  • [74] Wilber, M. J., Fang, C., Jin, H., Hertzmann, A., Collomosse, J., and Belongie, S. BAM! The Behance Artistic Media Dataset for Recognition Beyond Photography. In Proc. International Conference on Computer Vision (2017).
  • [75] Wilson, M. AI Is Inventing Languages Humans Can’t Understand. Should We Stop It? Fast Company (2017).
  • [76] Xu, T., Zhang, P., Huang, Q., Zhang, H., Gan, Z., Huang, X., and He, X. AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks. arXiv:1711.10485.
  • [77] Zhang, H., Xu, T., Li, H., Zhang, S., Wang, X., Huang, X., and Metaxas, D. StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks. In Proc. International Conference on Computer Vision (2017).
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description