Will A Computer Ever Be An Artist?

Can Computers Create Art? Part 3

As I described in the previous part of this series, computer AI algorithms do not create art; they are tools for artists. There is no existing algorithm or AI software that can credibly be called “an artist,” and there is no current technology to make this possible.

Could this ever change? Perhaps future AI technology will become good enough to be called an “artist.” The definition of art is fluid and changes over time and across cultures; perhaps we will become more accepting of software artists. In this essay, adapted from a longer academic essay, I discuss this question.

What is the question?

The question I address here is: “Could a piece of computer software ever be widely credited as the author of an artwork? What would this require?” This is a question of the psychology and philosophy of art.

One might think that we only need to look up existing definitions of art to find the answer. However, these definitions are not helpful. Art is a complex phenomenon that varies across cultures and changes over time, and encompasses a broad variety of forms. Contemporary philosophers, instead of defining art prescriptively, have instead attempted to reverse-engineer definitions of art from existing practice, and these definitions are not particularly useful for judging whether machines can be artists.

Instead, the definition of art changes over time as a result of a cultural consensus. Over the next decade, our society might decide to define, say, fishing to be an art and sculpture to not be an art. But this seems very unlikely. I believe we have a sort of “gut” definition of what can or can’t be art, based on both our culture and our evolution. My goal here is to predict how people might agree to change the definition of an artist, rather than simply to prescribe what I think it ought to be.

A Social Theory of Art

I will now describe my own theory of what it takes to be an artist. In the following sections, I’ll discuss several other common alternative theories that keep coming up.

It is helpful to consider why we create art. Darwin first suggested that art-making is, in some way, a product of our biological evolution. This argument was fleshed out very persuasively by Denis Dutton, in his book The Art Instinct. While I cannot do justice to his argument here, the main conclusion is that creating art served several functions for our Pleistocene ancestors that persist today, including:

  1. Art-making provide for sharing and gifts
  2. Art-making serves as a fitness signal for mating and sexual selection.
  3. Art can also be used as a display of wealth, status, and tribal affiliation.
  4. Storytelling, music, and dance strengthen social bonds within a group. Storytelling additionally plays a very important role of communicating information that would otherwise be hard to share.

I observe the following: each of these roles of art are social activities. Hence, I argue art is an interaction between social agents, and only social agents can create art. A “social agent” is anything that has a status akin to personhood; someone worthy of empathy and ethical consideration, and someone that we communicate and share with. Many of our behaviors are social, such as gifts, conversation, and social relationships like friendship, competition, and romance. In contrast, while we can get emotionally attached to our computers and other possessions, we feel no real empathy for their needs, and no ethical duty toward them. Objects participate only in shallow versions of these relationships; I might “love” my bike or talk to my phone, but I don’t feel empathy towards their feelings.

This means that our current computers cannot create art until they have personhood, just as people do not give gifts to their coffee makers or marry their cars. Conversely, if there were ever such a thing as human-level AI, with the same thoughts, feelings, and moral status as a human, then it would be able to create art. But “human-level AI” is pure science fiction right now, and we are nowhere near achieving it.

(Art-making does also have non-social benefits to the artist, such as practicing skills or providing a meditative activity. But these benefits are secondary to their social benefits: they are not the reasons that evolution has produced art as a human activity. Similarly, one may also talk or sing to oneself while alone, but talking and singing are still fundamentally social activities; binging on potato chips is not good for one’s health, even though it is explained by our Pleistocene origins.)

Other Possible Criteria

When asking the question whether computers can create art, there are a few common arguments I hear for why computers cannot create art: art must have an intent behind it; it must convey an emotion. Conversely, some authors argue that, once their software satisfies a few basic criteria, such as autonomously creating great work or being “creative,” then they’ll have succeeded at making a computer that is the author of an artwork.

However, each of these proposed theories, in my opinion, do not hold up under scrutiny.

Produces great work

It is tempting to judge whether a computer is an artist based solely on the work it produces. If an algorithm outputs a continual stream of diverse, stimulating, and beautiful artworks — without requiring a human to select the great results from a sea of duds — maybe this algorithm is an artist?

Skill is clearly not the real requirement for someone or something to be an artist. Any human can make art, including unskilled amateurs and children. There is lots of artwork that is unoriginal and derivative (e.g., “trashy” movies or novels; kitschy paintings and photographs), but still art. Conversely, computers can already be programmed to create infinite sequences of dazzling realistic or abstract imagery, exhibiting technical proficiency way beyond typical human ability. Nature creates many stunningly beautiful and perspective-altering phenomena, from natural landscapes to orchids to spiderwebs, but we do not consider them art.

Intent and Expression

Some people say that only humans can create art because art requires intent, or, it must express something, such as an emotion. However, it would be easy to build artificial systems that do this.

For example, in many artworks, the intent can be summarized by a short sentence: “depict a specific beautiful landscape,” or “convey an experience of living through the horrors of war.” It would be straightforward to build systems that generate “intents” like these, and then fleshes out the representation, or hires crowdworkers. (As we have seen, an artist is not required to execute on the intent to be the author of the artwork.) The system could work from recent news articles or Tweets to keep itself fresh and current. Feedback from online users could be used to update its style over time. It is easy to imagine building more and more sophisticated improvements to this idea.

This artificial system supplies intent and selects emotions to convey. But I think most people would agree that this system is not “an artist;” it is still a human-engineered system to create art, and the authorship really belongs to the author of the system.

Kiros’ “Neural Storyteller” experiment is trained from romance novels to generate stories from images. A network could easily be trained to generate artwork intents from random sampling. (This result from samim)

Growth and Surprise

Toward the end of his life, Harold Cohen described the evolution of his thinking about the software that he built. His software was continually producing high-quality, unique art, some of which has been acquired by major collections, and he, at first, suggested that this made it an artist. But, as it continued to produce new work over many years, he observed how the style was completely determined by its programming and never changed unless he changed it. He began to feel that a software artist would have to grow over time, and have its own worldview over time. Other authors have also suggested that a piece of software should “surprise” its inventors in order to be called an artist.

The AARON bot posts posthumous new art from Harold Cohen, generated daily. The author of the bot itself appears to be anonymous.

Each of these criteria sound like they should be enough. But they have already been met. For example, the algorithm that produces Mandelbrot set images can be specified in a single sentence, yet produces dazzling animations of infinite complexity. These results are very surprising and original, but we do not call its iteration equation creative, or an artist; the same can be said for many other complex systems. Scott Draves’ Electric Sheep evolves over time, producing new imagery far beyond anything that Draves originally conceived.

The Mandelbrot set algorithm produces endlessly surprising and dazzling visuals. This does not make it an artist; the algorithm can be described in a single short sentence.
Scott Draves’ continually-evolving Electric Sheep

General Observations

Many of the theories in this section have the following recipe: identify some attributes of human artists, and then hypothesize that AIs with these attributes will be considered artists. Artists make high-quality work; artists supply intent; artists are creative; artists grow. Many of these attributes sound important — but turn out to be easy to implement in a real system, once you try. Furthermore, amateur human artists can make art even when they aren’t skilled enough to possess these attributes.

One could object that my definitions are too shallow; maybe “intent” is more than just some short sentence definition. I suspect that any attempt to define “intent” or “expression” that is not shallow is going to end up in the same place: the “intent” or “expression” has to come from a person (or social agent).

For example, perhaps an AI artist would have to be autonomous or have “agency.” It is easy to devise a piece of software that runs indepedently and takes actions without further human intervention; hard to define these terms in a meaningful way beyond that.

Will an AI Ever Be An Artist?

With this background, I now turn to speculating about the future. As we have seen, authorship of all current algorithmic art is assigned to the human author behind the algorithms. Will 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?

This is science fiction

Human-Level AI. If we ever develop AI with human-level intelligence and consciousness, by definition, it would be able to create art. But, this scenario is science fiction and we have no idea if this is possible or how it would be achieved. Making meaningful predictions about a world with “true AI” is impossible, 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 is made by aliens from outer-space — if do we ever meet them, we will have more pressing questions than what kind of music they like.

Social AI. If, as I have argued, creating art is a fundamentally social act of expression and communication, then it follows that AI could be granted authorship when we view the AI as a social agent.

Paro therapeutic robot seals

There are many examples of shallow chatbots and robots that people treat in somewhat social ways, or even fooled by. The classic example is Eliza, a simple text-based “psychiatrist” program developed in 1964, based on simple pattern-matching and repetition of what the user types. 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, there are many anecdotes of people being fooled by “chatbots” in online settings, including the recent plague of Twitter bots. But, once the veil is lifted, it is clear that these chatbots do not exhibit real intelligence, and we feel cheated if we’d thought they were “real.”

Perhaps, for many users, the system does not need to be truly intelligent, it just has to be perceived as a social agent. Maybe someday agents like a Siri or Alexa will be treated like junior members of the family, who can answer questions or make artwork for you. I do not believe this will happen, but it could.

Non-social AI. A few algorithms have been promoted as artists. For the reasons given above, I do not believe that such methods will be accepted as true artists. It is possible to imagine the artworld changing this, say, a curator at a well-known museum forms a show in which software algorithms are credited as the artists. There would be controversy, and discussion in newspapers and journals. Perhaps other curators and galleries would follow suit. This sort of process has happened for things like abstract expressionism, and not for chimpanzee art. So far, however, gallery curators are crediting the human artists, not the software that they author and run.

Dangers. A continual danger of new AI technology is that human users misunderstand the nature of the AI. When we call a shallow AI an “artist”, we risk misleading people. If you convince people that an AI is an artist, then they will also attribute emotions, feelings, and ethical weight to that AI. Consequently, I would argue that calling such AIs “artists” is unethical. It leads to all sorts of problems, including overselling the competence and abilities of the AI, to misleading people about the nature of art. We also deprive the AI’s designers of authorship credit.


I do not believe that any software system in our current understanding could be called an “artist”. Art is a social activity.

I mean this as a warning against misleading oneself and others about the nature of art. Creative and intelligent people write software that creates art; the software itself is not intelligent or creative.

Of course, 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 do not 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. 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.

Art maintains its vitality through continual innovation, and technology is one of the main engines of that innovation. 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.