The conceptual artist Marcel Duchamp is often seen as a key founder of conceptual art. His 1917 piece 'Fountain' is often cited as the first conceptual artwork. Spoiler: it's not a fountain. Duchamp spurred an early 20th-century conceptual art movement that pushed the boundaries of the definition of art. "It's a urinal." "I don't get it." "But is it art?" Soon you will see, that this is the wrong question.
Consider a few cases to grasp the complexity of this question.
- I draw a picture [ed. note: despite my artistic challenges]. If I hang it on the refrigerator at home, is it art? What if I'm a talented artist with a picture in a museum? Does the place where I hang it or the quality of the work determine if it is art?
- A programmer uses code to make art. Is this art? An entire field of generative artists would argue yes. Others, however, argue it is not. Following this line of thinking, a photographer uses a camera to make art. Is photography art? Is beautiful photography art? I think most would agree, yes, beautiful photography is art. But perhaps not all photography is art — we all have plenty of photos on our smartphones that are not art. Further, just like a photographer uses a camera as a tool to make an image, a generative artist uses code to make an image. Does the tool to make art matter?
- A computer uses an algorithm to make art. When we take the person out of the process, and the computer writes the code (like machine learning does), is the output of the code art? This is the challenge that many are grappling with today with the recent release of DALL·E 2.
DALL·E is an AI system from OpenAI that can generate photo-realistic images and artwork from simple text descriptions in natural language. It was created using a neural network on images and their text descriptions. A picture of a cat has the description, "cat sitting on the windowsill" or the like. Then, through deep learning, the DALL·E model learns from pictures and their descriptions to generate true-to-life images or artwork that's never existed before, all from a sentence or two of plain text. The DALL·E model "knows" what a cat on a windowsill looks like because it has "seen" many cats on windowsills and can recreate a representation that best fits that description.
To be clear: this is no Google image search trickery. There are no searches or filters that are being applied to previously captured images. Because a computer was trained on millions of images, it has representations of images linked to text. So when you ask it to draw a cat, it recreates a pixel representation of the cat. When you train an image on a bunch of Picassos, it recreates the pixel representation of Picasso's style. Thus, it can generatively build an image based on what images it was trained on. DALL·E 2 is a public beta, and I just received access. Naturally, I didn’t get anything done that day and started testing the model.
Candidly: it feels like magic. I started testing a simple idea of a scientist looking through a microscope (gallery above). First, "scientist looking through a microscope in the style of Picasso"...nailed it. Then, "scientist looking through a microscope in the style of Matisse...plenty of bright colors, as expected. Followed by, "Scientist looking through a microscope in the style of Marcel Duchamp", which shows a man (sigh) looking through a strange microscope-looking contraption. I next tested some other styles, like "abstract Pencil sketch and watercolor of a scientist looking through a microscope"...solid. And finally, "comic book hero scientist using a microscope to save the day", which shows a white lab-coat/cape-wearing superhero in spandex, just like my typical day.
Consider the first prompt "Scientist looking through a microscope in the style of Picasso." Impressive. But is it art? If I had painted the image by hand, and hung it on a wall, many would agree it is art. However a computer generated it, without me, and so we question whether or not it is art. If I trained a robotic arm to paint the output of the DALL·E 2 image? Is that art? Does the physical medium matter?
Artists often create something from nothing. In our society, creating things, the basis for creativity is elevated. When a composer makes a new piece of music, it is celebrated. When a founder starts a new company, it is celebrated. When a scientist makes a new discovery, it is celebrated. Creating something from nothing is hard, and perhaps this is part of the reason why it is celebrated. What happens when creating things is easy? Does the amount of effort make something art?
As technology becomes more sophisticated, earlier technology becomes more commonplace. When technology begins to fade into the background, how the technology can be used and what is made from it comes to the forefront. Ultimately, technology becomes infrastructure — simply the tools used to make something. Remember when Word, Excel, or PowerPoint seemed like radical innovations? Today, they are infrastructure that supports people to make things. Perhaps a better example is Adobe Photoshop and Illustrator, or more recently Procreate or Descript. Descript is text-based audio and video editing software. Using it feels a bit magical, but again the technology fades into the background and lets the creator create.
As technologies become more sophisticated, and AI models integrate into more fields, a collaboration between creators and technology will continue. In the end, technology sits in the back seat while the creator drives. It matters little what tools were used, how much effort was invested, or whether the output is analog or digital. Instead, what matters can be summarized in an editorial that defended the Fountain:
A slightly cropped version of the photograph was published in the Blind Man to illustrate an anonymous editorial that defended the urinal in clear – and, in their implications, revolutionary – terms: ‘Mr Mutt’s fountain is not immoral, that is absurd, no more than a bathtub is immoral. It is a fixture that you see every day in plumbers’ shop windows. Whether Mr Mutt with his own hands made the fountain has no importance. He CHOSE it. He took an ordinary article of life, placed it so that its useful significance disappeared under the new title and point of view – created a new thought for that object.’ (Anon., ‘The Richard Mutt Case’, Blind Man, New York, no.2, May 1917, p.5; note that the second issue formulated the journal’s title as separate words.)
Whether something classifies as art is the wrong question. Art, like music, makes you feel. Duchamp's art makes you think. Instead of asking whether something classifies as art, a better question is whether it activates emotion, logic, sensation. When we see an impressive technology, like DALL·E 2, we can both appreciate the technology, and eagerly await the creations made from using it. More people are going to be able to take ordinary objects or extraordinary technologies and create new points of view and new thoughts. Duchamp would be proud.
Further Exploration
- Description of DALL·E 2 from OpenAI
- Excellent twitter thread of DALL·E 2 examples
- #DALLE art hashtag thread from Twitter
- DALL·E 2 explainer with examples
- Some limitations of DALL·E 2
- Amazing DALL·E 2 prompt book from Guy Parsons
