AI won’t replace artists- instead, it will augment them
In an unsurprising yet exhausting twist, the art world is obsessed with works created by artificial intelligence. If you’ve been keeping up with art news, you’ll know that last October a very generic AI portrait sold for a cool $432,000. It was made by an open-source program called GAN. Christie’s claimed it’s the “first portrait generated by an algorithm to come up for auction.” Of course, that’s not really true- artists like Thomas Ruff have been making compilation artworks like this for decades. The piece that sold at Christie’s wasn’t even an advanced or inspired use of AI image creation. So why was it sold? I believe it’s because we’re at the logical end-point of a consistent trend in the art world: rejecting itself. For centuries now, art has been in a perpetual freefall, fragmenting and challenging the very definition of its form.
This isn’t new. Master painter Caravaggio rejected the light and idealism of painting in the late 1500s by depicting incredibly dark, morbid, dramatic scenes. Picasso rejected the human form in favor of a more intuitive, emotional figure. Duchamp rejected the idea of art as a sacred creation and did think like calling a urinal art- to the shock and awe of audiences. Rothko rejected figure and representation entirely and made a fabulous living selling blurry blocks of color on canvas. Andy Warhol rejected whatever meaning was left in art with his irreverent representations of kitsch pop culture.
Warhol’s spiritual descendant Jeff Koons has devolved further into the art of meaningless shamelessness by making even uglier, kitschier, more repetitive pieces, and regularly stealing ideas from lesser-known (and arguably more talented) artists. The fine art world doesn’t care if Koons, or anyone else, is unethical. Anything’s ok as long as it’s a rejection of old standards. First was the rejection of perfection, then a rejection of realism, then a rejection of form, then of meaning. And now, it seems, we’ve reached a rejection of any purpose whatsoever. AI art is just the concept of interpretation boiled down to its nuts and bolts. Though that may have some philosophical interest and aesthetic appeal, I can’t make the case for anything broader.
When will we realize that rejecting standards is the new standard? When will we reject rejection itself?
I wonder if the conceptual de-evolution of the definition of art has reached its lowest point in AI art. Here’s how AI artworks: programs are trained to recognize patterns in art and structure, then use those patterns to create their own works. It’ll be a while before they can start searching for ways to create meaning and react to social trends the way real artists do. For now, they’ll be- by nature- 100% derivative.
In fact, this is the one place where the art world’s rejection obsession could still produce something interesting. Why don’t we ask our AI to create works that analyze then reject trends, instead of emulating them? If they can identify the rules of art, we might as well ask them to break those rules.
However, I feel like that’s a lot less interesting than letting artists use AI to create meaningful works. In a previous article, I talked about ways that technology could reinvent art as we know it. I believe it will rely heavily on massive open-source asset libraries. I also think that those libraries or databases will incorporate a great deal of information that we upload. How so? Consider the following technologies.
This Person Does Not Exist creates faces that look realistic, but are actually a conglomeration of different faces.
Promethean AI takes assets created by others and puts them in an AI-generated scene. It takes care of the lighting/ shadows & is easy to modify.
Nvidia’s GauGAN allows you to draw a shape and fill it with an object/ material type. Then, AI turns it into a photorealistic 2D scene.
WordsEye lets you describe a scene in writing, then AI creates that scene. It has a small asset library for each object so you can choose the right one.
Promethean and WordsEye both work within a 3d space, while the others are 2d. The thing that connects these all is their use of AI and asset libraries to reach a specific goal. With the exception of This Person Does Not Exist, they all allow a user to input specific requests or constraints. (This Person Does Not Exist could be modified to allow user constraints, since it’s made with the open source GAN system). In terms of art, these are all more like a collage than a painting. But the technologies are still in their infancy. The more customizable the systems become, the more creative and individualized they’ll be.
In this tech, we can begin to see a specific potential application of personal asset libraries. Most people of our era already create rich asset libraries on their phones through phone, video, and audio recordings. We use these to create our own identities, both personally and socially. But what if we could use these images to create things inspired by our personal worlds? What if the photos we take could be tools instead of end results?
Let’s explore that idea a little. You probably have a specific digital aesthetic, built from your entire digital footprint: pictures saved on your phone, things you liked on Pinterest, what you post on Instagram, etc. An AI could easily analyze all your posts and recreate new scenes and objects that fit your aesthetic. You could ask it to create a room inspired by this; or by specific constraints that you outline. You could start by drawing the room’s shape, and objects within it, like you, would in GauGAN. Then, it could then be built the same way that Promethean AI creates spaces (it would pull from images of rooms you’ve posted and liked). Finally, the room could be filled with objects from your Amazon and IKEA wishlist.
Let’s say you want to go further, and add your own objects to the scene you’re creating. One way you could do this is by including 3d assets you create out of real-life objects. This could be done easily with apps that are available on your smartphone. Or, just use your old 2D photos. Free online platforms like Smoothie 3D are able to turn any 2d photo into a 3d object. New startups like Threedy.ai aim to make this process even faster and less manual using AI. And once you’ve got a rough 3d scene, you could use a GauGAN-like drawing feature to sketch in some shapes- then fill those with an object.
And let’s push it further still. Let’s say you want to add a sci-fi gadget to the scene you’re creating. You pull on your virtual headset and draw a rough cylinder. Then, you have to fill the shape with a texture- so you ask the AI to pull from a moodboard you’ve created. It creates a cylindrical object using inspiration from these images. Using this same technique, an entire room full of objects could be created.
So wait- how does this fit into the world of fine art?
Clearly, these methods (and endless more that haven’t been imagined yet) could propel the art world to new heights. Creation and self-expression will become increasingly accessible and increasingly multi-media. This technology would also allow artists to create far more immersive artistic experiences than are currently available.
One final application I imagine is a type of art that doesn’t just use AI to incorporate the artist’s own digital information. Instead, I think the next logical step is to and expand art into the world of our minds and social systems.
AI art could have a truly personal relationship with the viewer.
Imagine that you’re walking into a gallery to see a new art exhibit. At the door, you sign into your google account, then slip a virtual reality headset over your eyes. Suddenly, you’re transported into a room. The walls are a dreamy conglomeration of scenes that look like they could be from your own life. An animation of your cat, created from hundreds of your cat photos and videos, scampers across the floor. The plants around the room are like nothing you’ve ever seen before- they don’t exist in the real world- but they look like they could be real. And they’re displayed as aesthetically as the botanical images you regularly like on Instagram. You see yourself in a mirror- a 3d conglomerate of every selfie attached to your Google account. The outfit you walked into the room wearing has been scanned by cameras and replicated in VR. You can switch through different outfits- both ones you’ve worn in photos, ones from videos you’ve liked, and ones AI has created by compiling styles you like.
If a stranger walks into the exhibit, the AI combines your tastes with theirs. A window opens up on the wall and you see scenes from your childhood mixing with scenes from theirs. The floral carpets that covered the ground in your world now morph into tiles with a flower pattern. The music shifts from your personal Spotify playlist to a remix of your songs and theirs, created by AI. This world could never exist between anyone else. You feel intimately connected with this person in a brand new way.
While the art world is built on rejection, technological and virtual art is a world of opportunity. Anyone can jump in and make the next wild new thing, hypnotizing audiences the way new paradigms in art have amazing cultures throughout history. This act of creation, freedom, and self-exploration is art. If you think that computer programs could replace artists, you have a fundamental misunderstanding of why art is created.
Art has always just been a way to share human perception.
Artists have always taken inspiration from the world around them, the same way AI does. It’s much more interesting and creative to use AI as a new medium or tool rather than the end result. AI and digital art is a new direction in which we can explore our perception and inspiration.
If we’re going to continue rejecting art norms, it’s time to boycott the norm of art that is simplistic, meaningless, and boring. Of all of the ways creativity will evolve, AI art is sure to be one of the most interesting avenues- and, of course, the most intelligent.
A text by Tara Bingham for towardsdatascience.com .
Photo Illustration by Joe Webb. Glasses courtesy of © 2016 Sony Computer Entertainment inc.