Drawing Perfect Circles

While at Berkeley, I took an avant-garde engineering course known as Critical Making. I say avant-garde because the syllabus had us designing everything from boxes to protests to evil. The goal of our course was to design and fabricate with emerging technologies while being grounded with a civic conscience. We were learning that if design was our superpower, what were the boundaries to break and what were the boundaries to protect?

Before the beginning of each class, we would have regular share-outs, where classmates would teach classmates about random subjects. One day, my classmates presented a mysterious exercise. They split us into teams based on the tables we sat at and began to pass out items. A pencil to one group, a pipe cleaner to another, a sponge to my own team, and paper to all.

"Please draw a perfect circle on the paper with the instrument we provided."

My team laughed—we only had a limp sponge to use as an instrument. Sponges do not draw perfect circles; they cannot even make a mark on paper. But that was the point. The classmates that presented had set up this exercise as a simple metaphor for privilege. Some people always have the pencil within arm's reach—the right tool for the job—while others are stuck with the pipe cleaners and the sponges.

I loved the concept of the exercise, but I couldn't help but come up with a loophole. I borrowed the sponge for a moment and wet it using the faucet at the back of the classroom. I held it over our piece of paper—drip, drop—and had gravity paint a few perfect circles for us out of water. My teammates nodded in approval, even though we all knew this little cleverness was beside the point of the exercise.

I thought of this memory one evening while trying to write up a reflection for a conference I was attending virtually. A week prior to my first week at Columbia, I was tuning into the International Conference on Computational Creativity. The entire agenda of the conference was about computational creativity and what it means to be creative. How could computation, humanity's latest and greatest tool, transform creative activities such as music, visual arts, literature, and the like? I know the term computational creativity sounds very fancy, but it's more mainstream than you think. EDM, Photoshop, Tik Tok filters—these are just a few of the ways computational creativity has become embedded in our everyday lives.

One of the focal points of this conference was about one of the most hyped forms of computation around: artificial intelligence. Many presenting researchers unveiled their bots and demonstrated their competence on creative tasks. Someone created an architecture bot that could collaboratively build Lego-like 3D structures with you in realtime. Another researcher created a neural network that demonstrated modest competence in both music and movement, showing off a machine that had multimodal intelligence. One group out of MIT Media Lab presented a project that created machine-imagined animals using StyleGAN2, a state-of-the-art deep learning model. They created hybrid animals—goldfish puppies, horsefly hummingbirds, and starfish spiders in digitally surrealist styles. Machines were obviously becoming capable of creative activities, and there was only more momentum to come.

As exciting and inspiring as it was to be in this niche space, in retrospect, much of it was like seeing people try to use sponges to draw circles, when the pens were readily available. As it always is with research, after the methods and the results, the researchers would always present the limitations. In the case of the architecture bot, there was no way to converse or set goals with the agent; it operated mutely and nondeterministically. The researcher behind the multimodal music and dance neural network could only embody choreographies with stick figures and admitted that the network often produced "deadpan versions" of the original dataset it was trained upon. In the hybrid animals project, users had very limited controls over the creation of their goldfish puppies and starfish spiders; creation was reduced to a click and generate process. All these research prototypes were the sponges we were training to draw perfect circles.

I never had a working definition for creativity before, but the conference offered me a dozen. Autotelic creativity, little-P creativity, big-P creativity, explainable computational creativity, casual creation—these were some of the snowflake adjectives and definitions researchers used to circumscribe creativity. Part of the reason why there were so many was because many researchers were trying to pin down just how human creativity would change once AI became more "creative". If AI became artificially creative, good enough to paint, to conduct music, and to design, what would that mean for human creativity? Would creativity be reduced or endangered in some way by technology?

I found some answers with respect to the human and artificial creativity question when I hopped into a Zoom meeting for a philosophy panel. As it so happened, Jan Løhmann Stephensen, a panelist, had written a paper on "post-creativity" that addressed this exact tension. Post-creativity is the idea that human creativity would be reformed after AI learned to computationally understand creative tasks. Stephenson's paper, titled "Post-Creativity and AI: Reverse Engineering Our Conceptual Landscapes of Creativity" was about the imaginary distinction between post-creativity and creativity as it is now, a mostly human enterprise.

But he pointed out that this distinction is imaginary. It is our assumption that human creativity exists in the first place, independent from tools and technology. What is human creativity without a pencil, a ruler, or a hammer? What is a fashion designer without a sewing machine, or a chef without a stove? What is a writer without language? Even language, one of the things we think of as inherent to us, are tools that have been crafted for generations to help us offload our thoughts into written storage. His point was that tools and technology have always been conduits for creativity and that therefore it has been a fallacy to think that creativity exists without technology. He framed creativity as a human invention, something whose definition is always in flux as humanity and technology co-evolve.

Stephenson wrote about how history illustrates this: creativity has meant different things to different societies over time. Early on in Western history, creativity meant activities that were the "providence of God". Creatura non potest creare, an aphorism from Latin, translates to, "the creature, or the created being, cannot itself create". Then in the Enlightenment, creativity manifested as intellectual property, something atomic and achievable by individuals. During the twentieth century, different movements modified the definition of creativity to suit their own cause. Take for example the diametrically opposed definitions of creativity posed by mid-century counterculture and capitalism. Creativity, as portrayed by the counterculture, was this under attack by capitalism, while was trying to mass produce what was previously artisan for profit. Meanwhile, capitalism pushed this narrative that its refined, industrial products were the pinnacle of human creativity. That narrative still stands today. Take Apple for example iPhones, Apple's "Think Different" campaigns, and the mythos around Steve Jobs.

I feel like the capitalist framing of creativity, which is that we should lionize creativity in our work and our products, took greater hold in society. To quote Stephenson, we have built "the impression that creative self-realization, especially in one’s working life, is a fair and normal thing to expect of oneself, as well as of others (e.g. colleagues and employees)." People will tell you they want ikigai, a Japanese concept where people are able to hit the trifecta of work, fulfillment, and societal need in their careers. We fete out Oscars, Nobels, and Grammys to those who have shown themselves to be the best at being original. Entrepreneurs angle themselves as innovators, and influencers angle themselves as brands, creatively collaging their lives for mass consumption. "[Creative self-realization] is a historically unprecedented expectation, which we have only quite recently come to embrace on an individual and societal level...never before have we had such high cultural expectations to the mundane activities of maintaining our subsistence."

If we looked at this history (of creativity as a social constructed concept) as a story, we are now at a new chapter with AI and post-creativity. So far what has been written about it has been heavy with existential, dystopic suspense. An AI can write a news article now, what is next for journalism? An AI can sculpt chairs, how will this change 3D modeling? If an AI can prove all the theorems in Newton's Mathematica Principia, when can we expect the Singularity?

Composing a poem, choreographing a dance piece, creating a composition, producing a portrait photo—these are the problems that researchers set on generative networks, which is a more specific term for the AIs devoted to creative tasks. These bots are the boogey monsters that make the headlines in tech beats. The fear factor around these generative networks is quite valid. Unfettered access to generative networks, particularly automatic portrait generators and motion transfer networks, by developers with malicious intent are exactly the root of deepfakes and fake news.

We are not threatened by the fact that Google Translate (an AI-powered tool) can translate from English to Spanish, German, Malagasy, and so on. It is simply a tool. But it seems substantially more concerning when an AI can compose a poignant poem. Demonstrating prowess in a creative task elicits our existential angst because we view creativity as a human thing, something intrinsic to us, something not externalizable to tools. Creative tasks, as one of the final frontiers for AI, then become this vanguard of what makes us human and them machines.

I came away from that conference incredibly inspired to look for a computational creativity agenda of my own, but I now also understood research more modestly. Beyond the hype, much of research is like holding the sponge over a page and trying to draw a perfect circle. It is not the perfect tool by any means yet, but it presents a new tool for us to express our creativity.