Creativity & AI

Before the advent of generative AI, creativity was assumed to be solely in the domain of humans. But, now, machines are able to compose poetry, create beautiful images, and compose music. While these creations are exciting, they are justifiably worrying too. Hollywood writers and actors are striking, in part, because of fears that their creations will be replaced by those of machines. If some of the world’s top creatives are worried about generative AI, shouldn’t we all be?

We are obsessed with the question: What does this expansion of machine capabilities mean for human creativity? The truth is that we don’t know—indeed we would argue that we can’t know because of the very nature of creativity.

There are numerous pathways to creativity. A highly complex system—such as the human mind or a large language model—allows for many different routes to generate an idea by making new connections between new and old ideas in a form of idea or concept mixology. This aspect of both human and AI creativity suggests that generative AI can supercharge human creativity, especially as the way an AI system creates new concepts can be quite different from that of a human. An AI system can also tap into a wealth of ideas that an individual human–or even a collective–might not have access to.

We’ll return to the topic of creativity and AI in future essays as it is too important and complex of a topic to try to tackle all at once. Today, we’ll focus on a few concepts about creativity itself to help set the stage for how to think about creativity in a mixed world of humans and machines.

Creativity as Restless Selves

Our minds are intrinsically restless. When our brains confront the same thing over and over again we experience a phenomenon known as repetition suppression. Familiarity tends to reduce the energy we invest, a strategy that helps automate behavior and promotes energy-efficient predictions about our surroundings. However, the catch is that our attention starts to fade. While we like predictability, we also crave elements of surprise and novelty.

Curiosity steps in as the tool we use to balance the continuous tension between exploring the uncharted world and capitalizing on our known environment. While curiosity can stimulate a particular form of creativity, it doesn't single-handedly guarantee it. Curiosity is fundamentally about the pursuit; we forecast areas where our learning growth will be exponential. It thrives in an intermediate novelty environment—neither too predictable nor too unpredictable. Intriguingly, the advent of generative AI, especially sophisticated language models, expands this curiosity-driven 'Goldilocks Zone.' This expansion allows for a comfortable and natural exploration of vast, previously untapped associations within the realm of human knowledge.

Creativity keeps our minds engaged with the world. As the world is infinitely complex, humans will never cease to be creative, even in the face of AI's attempts to simplify and chart our course. There's also a cognitive bias at work, an “end times” bias. It makes it challenging for us to envision all potentialities, while it's straightforward to anticipate the ways AI might disrupt what we currently deem creative practices. We can easily perceive the threats to creative tasks but struggle to envision the new possibilities and innovative solutions to more advanced problems.

The human desire for continuously updating our brains is insatiable and lies at the heart of creativity. Whether it involves interpreting and reflecting on the present world or venturing to envision and explore new ones, this drive remains constant. Generative AI introduces an “alien” perspective to our search, offering an alternative, both curious and creative, to the human mind.

“Creativity is the unique and defining trait of our species; and its ultimate goal, self-understanding, remains tantalizingly within our grasp.”

—E. O. Wilson, The Origins of Creativity.

Creativity is not a solitary endeavor: it's an appraisal made by the society in which an invention takes root. Novelty is merely one facet of the creative process—the cultural evaluation of the creation is equally crucial. Consider images generated by Midjourney. They may exhibit an engaging novelty, but how truly creative are they? It's plausible that our perception of creativity in such AI-generated imagery stems more from our surprise at the capabilities of AI, rather than the inherent creativity in the images. If this holds true, our collective assessment of their creativity and their subsequent value may decrease as we become more accustomed to the endless stylistic combinations AI can produce.

To classify something as creative, it's not sufficient for it to merely possess novelty or embody an unusual blend of elements. Creative work should defy our expectations of what's possible, which are often confined by cultural norms and regulations. When judging creativity, there are as many prescriptive shoulds as there are permissive coulds. This dynamic operates in two directions: while creativity exists to shatter constraints, mere disruption doesn't guarantee creativity. If society isn't prepared, the creative effort might go unrecognized.

If ChatGPT assists a scientist in making an enlightening connection, linking today's inquiries to a previously overlooked discovery, where does the creativity lie? Is the machine creative for making the connection, or is the scientist creative for utilizing the tool effectively? Does the earlier scientist fail the creativity test because their discovery was ignored? Can today’s scientist claim complete creative credit for realizing the value of the insight down the line?

The complexities of defining and recognizing creativity suggest that it's more than just an attribute of an individual or an AI. It's a dynamic process influenced by societal and cultural perceptions and a shared understanding of novelty and value. As AI continues to evolve and integrate into our daily lives, our definition of creativity will likely need to evolve as well.

“Creativity is an inherently social act, an experiment in the laboratory of the public.”

—David Eagleman and Anthony Brandt in The Runaway Species: How Human Creativity Remakes the World.

Creativity as Agency

When you ask individuals about their experience of being in a creative state, they often liken it to the act of play. But it’s play which operates inside certain bounds. There are rules to any play and it is how these rules are considered that defines the type of creativity.

How people make choices about the type of creativity they want to engage in matters. If individuals choose to stay within the existing rules, they engage in exploratory creativity. An artist may choose to make more extreme use of color or a scientist may test a new hypothesis within an existing paradigm. Around 97% of human creativity is this kind of creativity. Even if generative AI only doubles this capability, we’re looking at a big increase in creativity.

Combinatorial creativity emerges when individuals decide to blend rules from diverse domains. Think of a sushi burrito or balloon animals. There's no doubt that generative AI triggers a significant surge in this type of creativity. Of course not all of it will be any good. The creator has to choose what works and what doesn’t. Surely we’d want to avoid the trap of creative automation bias where the mere possibility of AI-generated creations leads to unchecked production. Just because AI can make a movie, doesn’t mean it should.

Transformational creativity, the rarest form, occurs when the rules change, constraints are lifted, or occasionally imposed. This creativity is highly prized as it inherently ushers in a completely new state. It's a driving force for many scientists, resonating with Richard Feynman's observation, "Science is like sex: sometimes something useful comes out, but that is not the reason we are doing it."

Ultimately creativity occupies an open and fundamentally positive space. Any AI that diminishes opportunities (or livelihoods) for creativity is a huge problem. But it is a choice. We're currently navigating the infancy of generative AI, where creativity is artificially constrained by immature and rather deficient user design. Given the present state, one might presume that the most creative role in the 2020s would be a "prompt engineer." However, this will undoubtedly change, as creative designers will play a crucial role in the broader discovery of these tools.

Agency is the snake biting our creative tails, where our actions shape our future creativity. If today’s creativity becomes stilted by AI, if AI is an intrusion rather than an enabler, we will reject it. Creativity, by its very nature, is an expression of being human.

“Creativity is about humans asserting they are not machines.”

—Marcus du Sautoy in The Creativity Code, Art and Innovation in the Age of AI