Can AI be Creative?
My Conversation with, and about, AI
Welcome to Polymathic Being, a place to explore counterintuitive insights across multiple domains. These essays take common topics and explore them from different perspectives and disciplines and, in doing so, come up with unique insights and solutions. Fundamentally, a Polymath is a type of thinker who spans diverse specialties and weaves together insights that the domain experts often don’t see.
Today's topic tackles the recent discussion around whether AI can be creative when it comes to art, writing, and other content creation. We’ll dig into what being creative really means, and how creative we really are. Throughout this essay, we’ll actually have a conversation with ChatGPT to discuss and test whether it can be creative.
(For more fun, in the voiceover recording of this essay, I use AI, via Play.ht, to perform text-to-voice on the ChatGPT responses and the three quotes I reference. You can listen to it here on Substack, or any major podcast player (← click this link))
New AI engines are being deployed such as OpenAI’s DALL-E 2 graphic art generator and their ChatGPT natural language model. Their output is incredible to see and is raising concerns across a wide swath of the public with issues like:
This sparked my interest, specifically around whether these AI engines can be creative as well as spawned some introspection on what creative actually means. I decided to test a few things out and so I figured the fastest way to get started was to ask the AI itself. So I logged into ChatGPT and asked it the following question:
Me: How do I prove that AI is creative?
ChatGPT: Proving that AI is creative can be a challenging task, as creativity is a complex and subjective concept that can be difficult to define and measure. However, there are a few ways in which one might approach this task:
One way to demonstrate that AI is creative is to show that it can generate original and novel ideas.
Another way is to show that AI can adapt and improvise in response to new situations or challenges.
Another approach is to compare the creative abilities of AI to those of humans, and to show that AI is capable of producing work that is similar in quality or novelty to that produced by humans.
Ultimately, the question of whether AI is creative is a complex and multifaceted one that may depend on how we define and measure creativity. There is ongoing debate in the field about the extent to which AI is capable of true creativity, and it is likely that this debate will continue as AI technology continues to evolve.
I think ChatGPT answered well which forces us to dig into understanding what we even mean when we use the term “creative.”
What is creativity?
Let’s start with the Dictionary.com definition:
the state or quality of being creative.
the ability to transcend traditional ideas, rules, patterns, relationships, or the like, and to create meaningful new ideas, forms, methods, interpretations, etc.; originality, progressiveness, or imagination.
Where creative is defined as:
having the quality or power of creating.
resulting from originality of thought, expression, etc.; imaginative:
So to summarize, creativity is the ability to create something that transcends the traditional. I’m pretty sure this is something most of us have in mind when discussing the topic but when you dig in, this form of creativity is actually kind of rare.
How creative are we?
“It is a mistake to think that the practice of my art has become easy to me. I assure you, dear friend, no one has given so much care to the study of composition as I. There is scarcely a famous master in music whose works I have not frequently and diligently studied.”
- Wolfgang Amadeus Mozart
Fundamentally, the vast majority of our life is spent in mimicry and building on the work around us. If we want to become a painter and go to art school, we study the great painters and styles in history. If we want to become a musician, we learn to perform the music of the past. If we want to be an Engineer or an Architect, we study the sciences and designs across the spectrum. No one is 100% originally creative. We are incrementally creative by building off of others.
“If I have seen further, it is by standing on the shoulders of giants.”
– Sir Isaac Newton
So creativity requires transcending the traditional, and yet we also use the traditional to assist us in how we operate in life. In fact, how do we know what to create if we don't know what exists? This is why artists study, musicians practice, and polymaths explore. In learning from others we can find the seams and confluences of past creativity and identify unique opportunities for the future. While this might sound less creative, it does put us in good company with Mozart and Newton.
How much creativity can we handle?
An interesting observation has emerged for me thus far in this exploration in that investigating creativity opens an entirely new Pandora’s Box if we apply these requirements back to our own human outputs. As we consider whether AI can be creative, especially when compared to humans, I think we need to pause and consider how much creativity humans actually tolerate.
Many of our essays thus far highlight the difficulty of solving our increasingly complex problems within focused disciplines. Numerous books have also been written on the topic such as Thomas Kuhn’s The Structure of Scientific Revolutions which tries to push us out of our paradigms. The main thread of frustration is that so many groups are anchored to traditional ideas, rules, patterns, and relationships. This investigation into creativity forces us to consider that the vast majority of people don’t transcend the traditional!
So what causes humans to resist the very creativity we worry AI will achieve? We explored some of this in the essay Embrace the Divergents where we referred to a personality type that was willing to challenge the status quo as divergents, rebels, and pirates. These types aren’t psychologically safe and they certainly don’t think about things the same way. In fact, our negative reaction to pure creativity is rooted in what we explored in Face the Fear regarding innovation. Fundamentally, we have a positive bias toward the known and comfortable.
Pixar Animation Studios is a great example as they were considered incredibly innovative with movies like Toy Story and Monsters, Inc. Yet these two were only transformative in their setting and characters, breaking away from historic fairy tales. Yet both of these movies follow the archetypical “Hero’s Journey.” This is because pure creativity is uncomfortable. If this seems surprising, I’ll just recommend the Children’s book Carla’s Sandwich which perfectly captures the challenge of the highly creative transcending the traditional:
There is a limit to how much pure creativity we can handle. Our creativity also must be grounded in and founded on the learning and creativity of the past. Lastly, there is the difficult acceptance of where we fall individually on the creativity and/or acceptance of creativity scale.
So, can AI be creative?
AI is coded on a similar foundation as our own social learning. The coders have merely compiled millions of data points across an exceptionally large variety of disciplines. It has an advantage in that it doesn’t have the human restrictions on memory recall, doesn’t have to sleep, and doesn’t have an ego about what it knows, or doesn’t know. It has a head start by having the consolidated learning of centuries on hand to build off of in mimicry, as do we, but faster.
In continuation of the ChatGPT conversation from above, I decided to test the AI’s creativity with the following:
Me: Please write a sonnet on a topic for which there are no sonnets already
ChatGPT: Sure, here is a sonnet on the topic of quantum computing:
Invisible to the naked eye, yet real,
Quantum computing lies beyond our grasp,
A world of particles that we can feel,
But never see, no matter how we grasp.
Its qubits dance and spin,
defying laws Of classical physics that we know,
And yet it holds the key to so much more
Than we could ever dream of, don't you know.
So let us delve into this realm unknown,
Where things behave in ways that we don't see,
And find the answers to the questions shown
To us by science, but as yet unseen.
For in this world of qubits and uncertainty,
We'll find the answers to our greatest quest.
So how well did it do? I’m not an expert in poetic composition so I did some research and found a great article by cognitive psychologist and poet Keith Holyoak who wrote in The MIT Press Reader:
To evaluate computer-generated poetry, let’s pause to remind ourselves what makes a text work as a poem. A successful poem combines compelling content (what Coleridge called “good sense”) with aesthetically pleasing wordplay (metaphor and other varieties of symbolism), coupled with the various types of sound similarities and constraints of form.
Interestingly enough, I think it meets the fundamentals of these requirements. So I asked ChatGPT what they thought:
Me: Do you think that was creative and unique?
ChatGPT: Creativity and uniqueness are subjective qualities, so it is ultimately up to you to decide whether you think the sonnet I wrote was creative and unique. However, I tried to write a sonnet on a topic (quantum computing) that is not often explored in poetry, and to use language and imagery that is evocative and engaging. Whether or not these efforts were successful in creating a creative and unique piece of poetry is up to the reader to decide.
Since it is up to me to decide I will propose my conclusion:
AI can have elements of what we’d call creativity. Between the AI art at the top of this essay and the quantum physic sonnet from ChatGPT, AI has demonstrated a unique and novel output. However, I don’t believe AI is at the point where it fits the full definition of creativity in that it does not have the ability to transcend traditional ideas, rules, patterns, or relationships. AI is just rules based on top of known and therefore, traditional data.
What I think is the most insightful and counterintuitive insight from this whole investigation is that I’m also not sure many humans meet this requirement either... I think this reinforces why the Polymathic mindset is so critical to foster because they strive for broad learning across multiple topics. It’s here where they can learn what’s been created in order to identify collaborative, cross-disciplinary confluences and fusions that transcend the traditional and comes up with new and unique insights and solutions. I personally view AI like ChatGPT as a method to investigate new areas of knowledge faster than I have before.
Fundamentally, these AI engines are going to get better and are going to challenge humans to recognize where true creativity is. I don’t see them as an existential threat but as a tool to use just as we leverage our strengths and how we build our capabilities through the mimicry of great thinkers, artists, and writers of the past.
As an interesting application, I’ve begun using ChatGPT to summarize my essays to see if my main points are clearly coming through. And just for fun, sometimes I like to add a creative twist:
Me: Please summarize this essay so my mother could understand it as if you were a 1920’s Gangster:
ChatGPT: Hey, boss. I read this article about how people are asking if these new AI engines can be creative, ya know, like us humans. The writer was trying to figure out how to prove it, so they asked one of these AI engines how to do it. The AI said there are a few ways, like showing it can come up with original ideas or adapt to new situations. But then the writer got all philosophical about what it means to be creative, like how we humans mostly just build on the work of others and don't come up with entirely new stuff all the time. Anyways, the writer didn't give a straight answer about if AI can be creative or not. They just said it's a complicated question and people will probably keep arguing about it.
I mean, it’s not wrong and I love the snarky attitude. If you listen to it, I just wish the AI at Play.ht had the learning material to pull off the right voice but that might have to wait for another time.