The Rise of The Art Machines

Technology continues to grow, change, and touch every facet of our lives and as it does we seek out new ways to put ourselves out of work. In some ways it’s for the best, saving lives from dangerous work. But it has created many hardships as well. There has always been one bastion of hope for humanity — creativity. Robots will never be able to do creative tasks. They can’t paint, they can’t write amazing articles, they can’t invent amazing products, services, technologies, or memes. Creativity is so fleeting, so fluid, and above all so subjective that humans themselves cannot even agree upon what it is or what it means. Which inherently poses problems for rational machines. Creativity, some argue, is exactly what makes us human. Creativity is the soul of humanity.

One of the other things that makes humans human, is the desire, the perpetual compulsion to overcome a challenge. We see something that cannot be chopped up, distilled into data points, fed into a machine, and then replicated, duplicated and mass produced, and we say to ourselves — I could do that! Enter Artificial Intelligence. The idea of AI has been around for decades, popping up in Sci-Fi making life great and easy. The impetus of AI is in our desire to create life and to be the masters of all things, in the most noble of ways of course. Part of me is aghast that someone out there is trying their damndest to figure out some way to take my job! But another part of me is curious if it can be done and if so, how well? Let’s take a look!

This lovely line comes from Google Arts and Culture’s poem portrait project. You enter a word and a photo of yourself and you get back a line of poetry that is part of a collaborative poem generated from your word, words from other users, and from the machine learning program that was trained on 20,000,000 words of 19th century poetry.

Here are a few other examples of lines from the collaborative poem generated by other users.

Most of what we see in the way of computationally-derived creativity falls into what’s known as the uncanny valley. That weird, weird space where something feels or appears close to human but not quite close enough to actually be human. In the case of AI poetry it usually feels like it’s making sense and almost beautifully done but something… is not quite… right? There are many examples of AI poets all over the web. A favorite of mine is poem.exe, and as we move on from poetry I’ll leave you with a “haiku” from poem.exe:

Words are one thing, they are a system of sounds represented by a system of characters strung together into a cascade of systems that create language. (Look how many times I got “system” in there!) This seems full with opportunity for AI, and it is. But computers could never take on images, they’re much too complex, with far too many un-systemetizable variables, right? Remember the uncanny valley? Buckle up, and feast your eyes on Deep Dream

“Initially it was invented to help scientists and engineers to see what a deep neural network is seeing when it is looking in a given image. Later the algorithm has become a new form of psychedelic and abstract art.”

While a lot of what you see from Deep Dream is the stuff of nightmares, it is interesting to imagine what AI “sees” in its “brain.” While Deep Dream stumbled into creating art, The Painting Fool set out with the express intent of becoming an artist. The project’s aim according to Simon Colton, the creator of The Painting Fool, is “To build a software system that is one day taken seriously as a creative artist in its own right.” When you look at a lot of what the project has produced, it is much more refined and intentional than a lot of other AI works, some of it even feels like “real” art.

There are still pieces that dip into the uncanny valley, pieces that feel so surreal in a way that is clearly not human. But, that’s not necessarily a bad thing, you could imagine viewing art as a being from another planet, and feeling similarly disconnected from the mind of the artist. But it’s that closeness to humanity, the echoes that make us uncomfortable. The echoes are there because the apple doesn’t fall far from the tree. AI is not yet able to come to their art on their own, it must be guided. The Painting Fool has been guided by Colton with certain principals in mind: good art makes you think, beauty is in the mind of the beholder, the meta-mountain, the whole is more than a sum of the parts. Which begs the question, where does the creativity lie?

Mike Cook is an AI researcher with the Royal Academy of Engineering with a focus on automated game design and computational creativity. He is most well-known for his AI ANGELINA (A Novel Game-Evolving Lab Rat I’ve Named Angelina) that creates video games. “The aim is to develop an AI system that can intelligently design videogames, as part of an investigation into the ways in which software can design creatively.” ANGELINA in essence starts with a seed, and then creates from that, tries its own creation and tests the results against goals or rules for “goodness” set by Mike. This is commonly referred to as computational evolution. An interesting idea that mirrors a lot of the way the creative process works. While ANGELINA is able to create games all on its own, even submitting some of them to game design competitions, the challenge is still how can AI decide what makes good stuff, “good.” Mike struggles to explain to ANGELINA even simple concepts like “why walls are so good!” He also admits that behind the scenes it may have created games so great, that they would make developers weep. Many of the games made are very simplistic, and quite weird but definitely worth your time to check out.

When you think about what creative work looks like, what image would you draw or describe? Probably the familiar trope of the messy desk strewn with paper and the most important part… the overflowing trash can. As the character you’ve imagined furiously draws, or writes, or invents, or whatever, she grabs her hair in frustration, crumples up the paper, and adds to the can. The true creativity is not in the creation but the curation of the work. The theme we begin to see emerge when we follow the rabbit hole of what each of these projects are doing seems to be that in the end humans must make the decision as to whether or not something is good.

While creative AI may not be this fool’s errand that at best will make a copy of humans (we have plenty of those as it is) and at worst will destroy us all and turn us into dirt and or batteries, it is not entirely a fruitless endeavor. If you set the fear aside for a second, you realize it can also be an opportunity to learn something about ourselves and further expand what we know about the labyrinth that is the human brain. By being forced to crystalize what it is we do when we create, we can better understand ourselves and the philosophical question of what makes us, us. What’s more, by using systems that recreate some of the leg-work creativity, brainstorming, iteration, execution) creative computation may be able to harness the power of AI to further boost our own creativity, freeing us to do more of the curation of ideas and executions. Will we finally fulfill the shining vision of automation: that manual labor would be abolished and we would all be free to think, and manage, and lounge in leisure while robots built our cars and cleaned our houses and made our food and so on. In the future we will no longer need writers or art directors, or illustrators. Computers will be doing all of that work while we sit in our cushy creative director chairs and do what humans do best of all — Judge other people and other things.

I leave you with one more of Google’s AI poem, to ponder and judge: