Can Artificial Intelligence Ever Be Truly Creative?
When Jane Goodall discovered chimpanzees using tools — bending sticks into just the right shape to penetrate a termite mound, drawing out the insects like fish on a hook — she caused an outcry. “The Toolmaker” could no longer be our species’ special title and people had a collective identity crisis. The same is happening all over again with AI. Is creativity unique to humans?
Many people argue AI cannot be creative because humans produced and trained it on their own ideas. But humans also produced and trained Beethoven, Dickinson and Da Vinci. Geniuses don’t spring out of the Earth like Greek gods. Like it or not, everything you’ve ever created was inspired — at least in part — by something another person taught you.
Even geographically isolated cultures come up with the same artistic and literary themes over and over again — a great flood, talking animals, people with wings and personified planets. People’s brains are so similar across the board that no matter where they go, they write the same stories and share the same dreams. Just like AI, you’re hardwired to have certain thoughts.
The Lovelace test — named after Ada Lovelace, the original computer programmer — is one proposed attempt to understand whether AI can be creative. To pass the test, an artificial agent must make something so original or advanced that the programmer could not explain how the AI generated it.
But does AI have to break the boundaries of its own code to be original? Not even humans can do that — genetics, hormones and brain structure dictate your thoughts and actions, yet you still find ways to be exceptionally creative. This school of thought argues that just like people, AI is creating what it can with what it has.
Therefore, just because an AI’s neural networks limit what it can generate may not preclude its ability to create new ideas. Everyone’s thoughts have an invisible outer edge.
Where Does Creativity Come From?
Some say generative AI is simply rearranging the data people feed it. But everyone borrows bits and pieces from the books they read, the art they admire and the songs they listen to. Is that plagiarism? How do you draw the line?
In order to learn, both humans and machines need inputs. People learn to draw by interacting with other people’s artwork first — looking at picture books, coloring in the lines, tracing drawings and trying to replicate cartoon characters.
Similarly, machine learning enables software to devour millions of data points — far more than a person could experience in their lifetime — and rearrange them to create something new. A Generative Adversarial Network uses convolutional neural networks to replicate human creativity. Its outputs improve as it learns, leading many people to say AI is creative.
Others attest that creativity comes from having new experiences. But in some ways, writing about what happens to you or painting what you see is the opposite of creativity — the ability to come up with something new on your own sets creativity apart from recordkeeping.
Because AI will never have a brush with death or travel by rickshaw, any stories it writes are wholly fictional. Some would say that makes it more creative than, for example, a person who writes a dramatized account of an adventure they had.
But new experiences also prompt people to think about the world differently, not just forming the basis for stories. Visiting a monastery or caring for a sick spouse can trigger previously unknown emotions or thoughts that lead to self-expression, which is one definition of creativity.
Will AI Ever Be Creative?
It depends on how you define creativity. In many ways, neural networks function like a human brain, and you can draw parallels between how humans and AI programs generate ideas. But if creativity requires self-expression, then artificial intelligence is decidedly not creative because it neither experiences emotions nor feels the need to express itself. It simply does what you ask of it.
AI software has no internal prompts like sadness, joy or anger to inspire it to write songs. It has no religious beliefs, favorite flavors, desires, fears, hopes or dreams. It is like a brain in a jar on a long-forgotten shelf, perfectly preserved and unfeeling, peering ever outward through a cloud of formaldehyde. It always needs a human to guide it. Without human ideas, it isn’t going anywhere anytime soon.
Room at the Top
It’s possible that AI could one day surpass human creativity and intelligence. If people panicked when chimpanzees bent a few sticks, their collective egos might take a beating if computers started writing better poetry than them.
Maybe, though, it won’t be the end of the world. There’s something to be said for learning to accept your limitations. Whether or not AI can become truly creative, humans can ultimately fall back on the fact that they created AI in the first place — without our inputs, it would simply be a few lines of code.
- NVIDIA: From Chipmaker to Trillion-Dollar AI Powerhouse
- Laura Petrich, PhD Student in Robotics & Machine Learning – Interview Series
- Liquid Neural Networks: Definition, Applications, & Challenges
- Patrick M. Pilarski, Ph.D. Canada CIFAR AI Chair (Amii) – Interview Series
- AI Leaders Warn of ‘Risk of Extinction’