Will Machines Replace Artists?

(Can they? Should they? And will they?)


[Disclaimer: this author has no professional knowledge of coding, nor has been involved in making any AIs.] 

Before we talk about the implications of this question, let’s first discuss why it isn’t such a big deal. 

By that, I mean that we’ve been asking many questions of this sort: will machines replace ____? Well, it turns out, machines have already replaced many parts of our lives, since anything that “uses power to apply forces and control movement to perform an intended action” is a machine.

Screws are little machines.

Even artists who try to make as many of the materials they use as possible are aware that some will just inevitably be mechanically produced. Paint, for example, is made in huge factories (unless you’re willing to pay $60/tube for handmade in Holland). When artists put wires on the back of their paintings so you can hang them easily, guess what? They’re using hand drills to put in screws. Some artists have even more complex machinery in their studio to build their own frames. 

Then why are we so unsettled by Artificial Intelligence?

Because, as the name implies, what defines an AI is its ability to learn and grow on its own, whereas a machine carries out direct orders (“rotate this rod to mix this tub of paint!”).

An AI gathers information, deduces patterns, and applies them to new information, correcting errors on the way. In this sense it’s much like the human brain. There actually is a branch of AI design that takes direct inspiration from human neural networks. 

It is NOT like a human brain, however, because it is able to process so much information at the same time. This allows AIs to spot out the most subtlest of patterns. Spotify’s AI, for example, is collecting user behavior based on genre, bpm, lyrics, or some other vague standards like the “femininity” of a song, and uses this to create “Discover Weekly” playlists for each of its 286 million users. 

So what does it even mean for AI to make visual art? 

When we left off art fairs last year, the word on the street was that AI art is the next big thing. 

A broader definition of AI Art, however, called “Generative Art” or “Algorithmic Art”, has been around for a while, especially in the design field. Consider, for example, that Photoshop uses AI when you use the “edge selection” or “sky replacement” functions. 

Generative Architecture by Michael Hansmeyer, designed for Mozart’s opera.

Why we all suddenly started talking about AI Art is because in 2018, a portrait created by French trio of artists Obvious called Portrait of Elmond Bellamy sold for $432,500 at Christies. It was one of eleven portraits they made by processing thousands of portraits from the 14th-20th century. Which is very close to the way we just defined what all AIs do. 

Obvious, Portrait of Elmond Bellamy

This was followed by other artists using the term AI Art to define themselves. What this recent wave of AI has in common is something called Generative Adversarial Networks (GANs). Unlike the broad definition of algorithms, this model uses two networks. One comes up with an answer, the other evaluates it, like a double-checking system for when you’re scoring college students’ quiz results. 

Mario Klingemann, The Butcher’s Son (2017)

Which is to say, there’s no one way to use AI in art. Many artists modify the results after running the AI, and AI technology itself is changing and diversifying quickly.

CAN an AI trick us into thinking that it’s a human artist? 


We’ve seen a literal urinal, a blank canvas, a canvas with two paint splatters on it, and have learned to call them art. As we just saw with Obvious’ example, because AIs by definition use existing examples of how we define art and apply them to create a new one, of course we will perceive it as a work of art as well. At least, we won’t be able to convincingly argue that it is not. 

In addition, AI art is still ultimately made by a human artist so it really wouldn’t be a trick. The current state of AI is much more rudimentary than you would think. It has the ability to find solutions to complex problems, but no AI has yet posed a critical question like Socrates with the Socratic Method or Nietzsche declaring the death of God. 

Sougwen Chung uses both hand-drawn and computer-generated marks and calls her art Omnia Per Omnia (2018) “collaborative”. 

 So, it may be comparatively easy for AIs to generate something that would be perceived as art because the formal boundaries of art (“what art should look like”) today is so broad. But that also means an AI needs much more specific directions on what kind of art they should be generating. It is still up to humans to create a relevance for AI art. As long as humans are still the audience of AI art, only we can argue about whether or not, or in what specific way it is relevant to our lives. 

But SHOULD we be using AIs to make a perfect imitation of something that human artists can make? 

What would be the benefit of that? For the sake of practice, I could build an AI that maps out the position of ducks that appear on every lake in Google Earth. But I probably won’t, because there’s no immediate reason to do that compared to the amount of time and energy it’ll take. 

It bears emphasizing that just because we CAN do something doesn’t mean we SHOULD.

If AI art  - or any use of AI - is to go beyond a self-indulgent demonstration of how far our technology has come, it needs a clear intention connected to the way it is executed. 

Obvious, who we’ve seen above, wrote a manifesto to accompany their piece in which they posed the question of whether AIs can emulate creativity. Other artists like Ahmed Elgammal are exploring similar questions, using AI not as an end goal but a way to rethink our existing definitions of creativity.

Ahmed Elgammal says: “I think it is expanding human creativity to totally new levels. You have to have imagination to have creativity, but the human imagination is limited because we are constrained by our world. The machine is good at exploring possibilities, so if we can frame the machine to explore possibilities for you in the art space, then the machine can give you lots of new ideas.”

Even more so, because AIs don’t create a direction on their own but mirror the way we think and pose problems, they might just amplify the negative aspects of our society

So AI art, using the same tools, has the power to discover and criticize these problems. For example, artist Joy Buolamwini, founder of the Algorithmic Justice League, examines how facial recognition has much higher error rates for people of color and for women. Her video-poem entitled "AI, Ain't I A Woman?" shows AI failures on the faces of iconic women like Oprah Winfrey, Michelle Obama, and Serena Williams. 

Joy Buolamwini, AI Ain't I A Woman

What AIs MIGHT do for us: Fully Automated Luxury Communism

Is our moral reluctance towards AIs just holding us back from making radical changes - like eliminating social inequality? Right now, AI-generated art is still being displayed through high-end galleries or art fairs. But what if we popularized this genre, say, to produce 100,000 more Jackson Pollock paintings? Since there will be no visible difference between Pollock’s painting and AI's, wouldn’t that allow for everyone to enjoy a Pollock at home?  

I know, this is a radical idea. But it’s worth having a thought experiment with, as Aaron Bastani argues in the book Fully Automated Luxury Communism. From cameras to computers, technology was always feared as a corrupting influence on the humaneness of art, yet every time we’ve managed to adapt and redefine creativity. What if automating repetitive, manual tasks in society lets us focus more on creative tasks that we believe make us human? - asks this theory.  

Aaron Bastani, Fully Automated Luxury Communism


Receive updates on new artworks, artists, special events, and holiday deals! Unsubscribe anytime.

Sign up to receive updates about new featured artists, upcoming studio tours, and more!

By clicking the button you agree to our Privacy Policy and Terms and Conditions