Just one,
but it’ll need countless training examples not to screw it up!
Humour is arguably one of the most defining and revealing aspects of the human condition. It is broadly similar across cultural divides, but at the same time dependent for its success upon an individual’s contextual awareness and social intelligence. As artificial intelligence becomes more integrated into our creative endeavours, its abilities and limitations with respect to generating humour deserves analysis. The on-point response I got when I mischievously prompted GPT-4 for a solution to the standard lightbulb conundrum is actually the perfect starting point to discuss how AI constructs humour, to think about its implications for AI-assisted copywriting and to identify its potential for producing content that raises a laugh.
The lightbulb joke exemplifies how GPT-4 can handle a classic joke format, output an appropriate response and even inject a meta-humorous twist that plays on the AI’s own learning process. In a display of wit that would fool even the staunchest AI sceptic, it has delivered an entirely unexpected yet logical punchline, which is a key ingredient in many successful jokes.
How does an AI keep its data cool? It stores it in the cloud!
Like anything else it learns, GPT-4 relies on vast datasets of human language to recognise patterns that are commonly found in humour. But while some jokes work because they hinge on their linguistic construction, others are doomed to fail because, ultimately, AI does not truly grasp anything at all.
Implications for AI-assisted copywriting
Understanding the hit-and-miss nature of GPT’s outputs puts the onus yet again on the copywriter’s diligent supervision. Only she can select and hone jocular, engaging content that captures attention and entertains the audience. And only she can discard the chaff, which is very important if you’re looking to use humour in social media campaigns, advertisements or any other content to engage users casually and memorably.
Why don’t AIs have a good sense of humour? Because they take things too literally!
So, while AI can generate humour, there are some challenges to bear in mind. What is funny to one audience may be offensive or confusing to another, and AI is bound to struggle with such subtleties. Furthermore, brands must always consider the appropriateness of humour in different contexts; an AI’s joke might sometimes miss the mark in terms of tone or relevance and have a correspondingly detrimental effect on a brand’s image. And the copywriter should always beware lest AI-generated jokes become formulaic and predictable, as machine learning models tend to overly rely on patterns that have previously led to successful outcomes.
What did the AI say to the algorithm? “I think we need some space, I keep running into your loops.”
Is there a future for AI wit?
It’s safe to say that we can expect more reliably humorous output from AIs as developers design more sophisticated neural network models that better understand and replicate the intricacies of human humour. For now, though, we can take advantage of hybrid systems, whereby AI suggests funny ideas that human copywriters refine and contextualise.
So to wrap up, while the lightbulb joke might seem simple, it proves that AI-generated humour can indeed change not just lightbulbs, but also the way businesses illuminate their connections with audiences. Most important is to do precisely what we do at Big Black Point: combine AI’s capabilities with the human factor to ensure that every punchline lands just right and every message shines brightly.
Why did the AI refuse to write a lightbulb joke? Because it was still trying to figure out how many humans it takes to perfect one!