Does AI threaten creativity or is AI threatened by creativity?
Generative artificial intelligence certainly likes to show its hand to creatives as algorithms push to learn and disseminate mock craft. Great art and great writing are formed by human emotion, built from personal experience and blended with layers of nuance. AI cannot replicate such inquisitive thought patterns. But it will try.
It will surely get better as time moves on, and the reason it will improve is because of us. Tech companies analyse our likes and dislikes, our interactions and connections, right down to our lines and words to find what works with an aim to create something cheaper and quicker, but inherently of poorer quality.
Humour has not got away with it either, as AI studies what makes a joke work and attempts to replicate the punchline.
READ MORE
“Humour can be understood as a type of psychological play,” says Sarah Crosby, author of Five Minute Therapy. “Like play in childhood, it gives us a way to test boundaries, experiment with ideas and process complex or even threatening material in a safe way. From a therapeutic perspective, humour often provides a release of tension and creates the conditions for flexibility of thought.
“We can see something from a new angle or reframe something painful with distance. It’s also deeply relational. Shared laughter strengthens bonds, helps regulate emotion and fosters a sense of belonging.”
With this in mind, what is the benefit for AI in learning humour?
Tech companies are analysing how and why we laugh in an attempt to replicate laughter’s ability to soothe. Humour is a stress reliever so perhaps it isn’t such a bad thing having additional ways to create laughter and ultimately boost our mood. But does this mean AI humour will take away from how we define humour in the human context, morphing it into something else entirely?
[ Stand-up comedy: I never knew that five minutes could feel like a lifetimeOpens in new window ]
There is a mistrust of algorithms as it is, and AI seems to tell a very poor joke.
However, as Janet M Gibson, professor emerita of psychology at Grinnell College in Iowa, says: “Humour varies by culture, in what they prefer and when they use it. But the internet allows for more cultural mixing, so today’s users are getting exposure to other culture’s humour more than yesterday’s users. This melting pot will likely cause some shifts in cultural values or attitudes and preferences regarding humour.”
Writing jokes and sketches and engaging with audiences is not easy and comedy is often written and rewritten until the punchline lands with a hard-hitting laugh-out-loud, gasping-for-breath laugh or that appreciative, well-earned laugh and nod that says, “I see what you did there”. It bonds people together, creates shared understanding and questions society in a lighthearted way.
How can an artificial intelligence predict, infer and create comedy gold that requires such a high level of human perception?
Can AI replicate the same unequivocal belly laugh some jokes deliver, jokes that have taken an age to perfect by the best comedians?
Again, it will try, but does it work?
“AI has come a long way in producing humour, in detecting it, and in explaining why something is funny,” says Gibson. “It’s not quite at the level of human behaviour, but it’s getting there.”
Similarly, Crosby highlights that “memes, gifs and shared in-jokes online are a contemporary form of humour that rely on speed, intertextuality and collective participation. While some nuance and tone can be lost in digital formats, the connective function of humour remains. What we may lose in subtlety, we gain in accessibility.”
As Crosby suggests, humour works best in social situations and shared experiences. It is this collective context that helps us understand why something is funny. The humour derived from AI falls somewhat short of the mark, having not yet mastered this layer of nuance just yet. However, Gibson highlights that we have gotten to a point in AI history where the lines between human-generated and AI-generated content is becoming blurred.
“Humans can no longer tell with high accuracy whether verbal or pictorial humorous examples were generated by a human or a computer,” says Gibson. “But it does have its flaws.” She uses the example of asking a friend to tell a couple of jokes. The first joke may garner a smile, but as the friend tells more gags they tend to get funnier and funnier. AI, however, often tells the same joke over and over again with minimal changes from pun to pun.
So what makes good comedy? Humour is obviously subjective, which means we don’t always know what will make someone laugh. However, we do know that anticipation and surprise are two defining factors that can lead to a brain stimulus reacting with laughter. Sounds simple, but it is far from easy. An audience predicting the punchline can lead to a whimper. Shocking the audience’s expectation with a satisfying build and strong punchline is key, which AI still seems to struggle to recreate along with the soothing nature of humour and laughter.
“A well-timed joke can communicate truth in a way that might otherwise be difficult to hear, and it can disarm conflict or ease tension,” says Crosby. “There’s a saying that there’s nothing as truthful as a half-joke, and that captures how humour often allows truths to slip through the cracks, sometimes playfully, and sometimes with a passive-aggressive edge. But humour is also deeply contextual.
“What lands as funny depends on timing, the people in the room, the relationship between them, and even cultural or generational taste. Like art, humour isn’t universal. It resonates differently depending on who is receiving it. In therapy, humour can provide an entry point into sensitive material, signalling safety and humanity.”
[ Donald Clarke: Comedy has to be funny? Don’t make me laughOpens in new window ]
While AI can certainly produce humour, it will always lack that lived experience, and as a result it requires a set of formulas to create something resulting in any kind of laugh. Timing, context and shared humanity are not formula based, but rather spontaneous and circumstantial. Perhaps it is not how clever the joke is that makes it work but rather its timing which AI is lacking in.
“Our responses to it [AI] though, can be revealing,” says Crosby. “It prompts us to consider what we truly value in humour be it spontaneity, vulnerability, or the sense that someone gets us. In that way, the fleeting, slightly hollow quality of AI humour might actually sharpen our awareness of humour’s role in genuine human connection.”
Will AI ever reproduce this element of community with its jokes?
In just a few short years generative AI has delivered to the masses a significant tool that is still developing, preferable without manipulating copyrighted work. And it also needs monitoring. But for now, as the debate rages on as to whether AI can truly be creative, it tells the same repetitive jokes as a five-year-old.
“Although some may believe AI can never be funny like a human, there’s a sunny side for getting AI to produce humour today,” says Gibson. “Just like calculators help humans who are bad at math, many humans are bad at humour production, and yet humans need or want humour in their lives. Not everyone has immediate access to human humour.
“Nowadays you can ask your computer to produce jokes, and it can provide them 1-2-3. Humans can’t do that.”
