gumgum

Artificial Creativity

How artificial intelligence is paving the way for the future of creativity

GUMGUM is AI & ART

ART OR NOT

That's what a team at the Rutgers University Art and Artificial Intelligence Lab set out to answer. They started by creating an Ai system called a generative adversarial network. Part of that network is the "critic": an algorithm that evaluates the work of a second, “generator” algorithm programmed to create imagery. This generator algorithm begins by producing imagery at random, none of which rises to the critic’s standard for art. The critic’s negative feedback slowly nudges the generator to produce images that move closer and closer to the specifications set by its creators. The result, they say, is an AI that can do something long considered the sole province of human beings: create an original work of art.

To demonstrate just how close artificial intelligence can come to the human creative process, GumGum devised a Turing Test. We commissioned five artists—and Pindar Van Arman’s Cloudpainter—to create a piece of art based on the same dataset, a collection of art by 20th-century American abstract expressionists. Then, we asked them to document the process, showing us their preferred tools and telling us how they came to their final work. Take a guess at the results below to see which were created by man, and which by machine.

Click on the canvas
  • GumGum
  • GumGum
  • GumGum
  • GumGum
  • GumGum
  • GumGum

art or not?

There is something about artificial intelligence that inspires a low-grade panic in most people. On the one hand, AI represents an existential threat to human supremacy.
Marina Esmeraldo

THE FUTURE OF AI AND CREATIVE INDUSTRY

There is something about artificial intelligence that inspires a low-grade panic in most people. On the one hand, AI represents an existential threat to human supremacy. Last year, Google’s AlphaGo beat the world’s best human player of Go, a notoriously complex strategy game with 300 times more possible plays than chess. On the other, AI represents a very practical threat to our economic well-being, eliminating jobs once performed by unskilled or junior workers. Already, AI has taken its place on the machine line (manufacturing robots), the customer service line (chatbots) and the design production line (facial recognition AI).

But pursuits that rely on and express our very humanity—art, literature, music—and the industries that harness that creativity were supposed to be safe. Machines that compose Irish folk Music, write Harry Potter fan fiction and edit science fiction movies all appear to be creeping into this most sacred realm.

If art is a process by which human beings express some idea or emotion, filter it through personal experience and set it against a broader cultural context, then by that measure, what AI generates at the behest of computer scientists is, well, something else. “At its root, art is one person communicating with another,” said Pindar Van Arman, a classically trained artist who has been coding art robots for 15 years. “I don't think that a machine will ever make art on its own until the machine is a person.”

AI IS YOUR NEXT CREATIVE PARTNER

Creatives who want to know how they can harness AI to chart a new course might look to the stars. In 2006, NASA engineers faced a dilemma: Aircraft was becoming so automated that pilots were spending too much energy inputting commands and managing automation sequences. At the same time, they were lulled into a false sense of security by imperfect monitoring systems. The team posed the following question, “How do we balance between exploiting increasingly powerful technologies and retaining authority, with clear roles between humans and automation?”

The answer was the H-metaphor, a model for interaction with intelligent machines that is more like horse and rider than master and servant. According to the H-metaphor, much like a rider who trusts his horse to negotiate the terrain, humans should rely on machines for set and forget processes. But they should also have the power to chart the course and tighten the reigns, retaining big picture decisions and refining results to get to a desired outcome.

Creative industries can use the H-metaphor too. Like computer scientists, they engineer their teams for maximum results. They hire copywriters, art directors, UX designers and technologists that fit their parameters for taste, ingenuity and personality. Then they weight their algorithm by doling out seniority, assign briefs and wait for the output.

Max Fresn, chief creative at Born AI, is experimenting with an AI system that he hopes will be able to do exactly what those teams do. Like NASA’s H-Metaphor, Fresn would set the course by programming the AI to solve a specific problem, then let the machine do its work.

“Instead of paying junior copywriters and art directors to give me a million bad ideas that I have to cherry pick and nurture, I can have an AI generate a billion shitty ideas that I can cherry pick and work on them myself,” Fresn said. Considering AI’s ability to search and process millions of visual data points, such an application would be imminently helpful when scanning the web for relevant reference, for example.

Fresn’s work, however nascent, is not unlike that of AI artist Pindar Van Arman. Ultimately, he’s creating a system in his own image, choosing the parameters and weights that suit his taste and priorities. These techniques are also being explored at creative.ai, where Pieters said his team is developing AIs modeled after particular team members.

“It’s never 100 percent correct, but it’s correct enough, because creativity is messy,” he said. Such models would be better put to use training young creatives than replacing them, because they would give them access to artificial guidance modeled on their boss.

Cloudpainter

Get the full guide

Artificial Guide GumGum

Computer Craphics, Inc.

Brooklyn, ny

We started with playing around with some eye-catchy 3D imagery but then quickly realized that what we tried to create looked like it was designed by human artists for a human audience. When humans create, we’re naturally drawn to pay attention to composition, a harmony of shapes, colors, etc. We thought computers might have a different idea.

We did research to find out where computer-generated art and data visualization are at this point in time. Then we looked for creative ways to integrate random and procedural objects and effects into our workflow. We reduced the amount we spent giving our own creative input and let computer algorithms help us make this final piece.

“We looked for creative ways to integrate random and procedural objects and effects into our workflow.”
Computer Craphics, Inc.
GumGumThis is a  human's  artwork

Marina Esmeraldo

barcelona, spain

This piece is part of an ongoing experiment called bossa landscapes, abstract compositions based on the observation of real things and landscapes, working from the principle that reality adds a layer of meaning and beauty i can’t find otherwise.

The name of the series is a reference to the musicality and cadence the works impressed upon me, and draws inspiration from my native brazil’s musical genre bossa nova, where the term “bossa” referred to an aesthetic reformulation and a new way of doing things, based on the balance of simplicity and dissonance.

“Reality adds a layer of meaning and beauty i can’t find otherwise.”
Marina Esmeraldo
GumGumThis is a  human's  artwork

Leandro Castelao

brooklyn, ny

Envision the future. our eyes as the most powerful tool we have. our thoughts interacting with the piece of art itself, transforming and recreating it. like a non-stop looped creative dialogue.

Technology will take abstraction to a totally new level and i wanted to talk about that with my piece. possibilities will expand and create new art forms, things we’ve never imagined.

The piece highlights the relationship between an evolved human being trying to understand and learn from a piece of art that looks like a perfect organized chaos. at the same time, there’s an active role between the piece of art and the viewer. in the end, the viewer is the co-creator, transforming, re-creating and changing.

I believe taste will evolve and there will be multiple trends going on at the same time. feedback loop. the bot operates largely as a human artist would. it paints, pauses, and considers its progress before applying its next stroke.

“There’s an active role between the piece of art and the viewer.”
Leandro Castelao
GumGumThis is a  human's  artwork

Briahna Wenke

charleston, sc

This project is so fascinating to me because it touches on this underlying fear of mine—and many others—that technology will ultimately leave no room for us to simply be human. Painting taps into something primal, something most people let lie dormant for most of their lives. It reminds us of just how human we really are.

My main goal is to work through the mental restrictions we develop as humans. A robot doesn’t struggle with self-doubt, or a lifetime of experience to work through. AI is given algorithms, structured guidelines and data and just needs to move its arm. It must be nice and easy.

When looking at the reference, O’Keefe’s description of what she saw in nature helped guide my interpretation of Dreher Island, where I was an artist in residence. I tried to allow the space I was in, and its movement, to guide my hand. Likewise, Elaine de Kooning, who referred to painting as a verb rather than a noun, struck me. Because it’s the process that feeds me, rarely the outcome.

“It’s fully the process that feeds me, only rarely the outcome. and the outcome is what feeds the viewer, ideally.”
Briahna Wenke
GumGumThis is a  human's  artwork

Cam Floyd

los angeles, ca

I really responded to Elaine de Kooning’s unique balance of pure abstraction and use of observation in her work. Like de Kooning, who used a statue of Bacchus for her model, I decided to draw on Greek mythology too, using Bernini’s sculpture, “Apollo and Daphne.” I liked how de Kooning merged the figure with the trees in the background to become one surface. This reminded me of Daphne, who was turned into a laurel tree. By emulating de Kooning’s signature contour lines, I hope to subtly reference Daphne’s tree in my own work.

“I chose to place my imagined de kooning-esque creation in a museum gallery to reference the challenge of honestly interpreting another artist’s work.”
Cam Floyd
GumGumThis is a  human's  artwork

Pindar Van Armin

WASHINGTON, DC

Like any artist, the Cloudpainter robot has evolved over time. In its first incarnation, the Cloudpainter was simply an automated brush following paint-by-number instructions. Now it is a complicated system informed by a variety of style transfer algorithms programmed by its artist, Pindar Van Arman, along with a rotating brush head and a feedback loop.

The work begins with a series of photographs of the portrait subject. Using facial recognition AI, the computer selects which picture it will use as its source material. Then, Van Arman programs his bot with a selection of algorithms that will work together to create an original piece of art.

The Cloudpainter uses a variety of mechanisms to paint. The robot arm pictured dips its brush into a palette of paint pots that sits beside it. Another, more complicated mechanism (not pictured) allows the brush to travel along an X/Y axis while applying colors.

Pindar Van Armin
GumGumThis is a  ai's  artwork