Humour, inherently human in its nuances, is a tough nut for AI to crack. ChatGPT, OpenAI’s language model, stands at the forefront of this challenge. But the question lingers: Is ChatGPT merely fun to use, or does it actually cross the threshold into being funny?
Before diving into its comedic chops, let’s quickly recap what ChatGPT is. It’s a conversational AI developed by OpenAI, based on the GPT-3 model, with GPT-4 being available only to premium users. Known for its ability to generate text that’s almost indistinguishable from human writing, ChatGPT has been a game changer in various fields, including customer service, content creation, and beyond.
Recent studies, like the one conducted by Sophie Jentzsch and Kristian Kersting, have put ChatGPT’s humour to the test. Through a series of experiments, they explored its capabilities in joke generation, explanation, and detection. The findings? Intriguing, to say the least.
When asked to generate jokes, ChatGPT often recycles. Over 90% of the jokes it generated were reruns of the same 25 punchlines. It seems like ChatGPT has a handful of jokes up its sleeve and loves to replay them, like a favourite comedy sketch that never gets old – or does it?
ChatGPT shows flair in explaining jokes, especially those involving wordplay and personifications. However, it fumbles when the jokes don’t fit into its learned patterns. Instead of admitting to not getting the joke, it weaves a convincing, albeit fictional, explanation. It’s like having a friend who pretends to understand a joke but actually doesn’t, just to make the conversation a tiny bit awkward.
In detecting humour, ChatGPT identifies characteristics like structure and wordplay. It doesn’t get easily fooled by surface-level elements, indicating a deeper understanding of what makes a joke. However, this doesn’t necessarily mean it grasps the full spectrum of humour.
Here’s the crux: ChatGPT can mimic humour, but it’s more akin to a parrot reciting a joke book. It understands certain joke formats well, especially those involving linguistic play, but it doesn’t quite capture the breadth and depth of human humour. It’s fun, yes, but its attempts at being genuinely funny are akin to a comedian with a very limited script, so far from a Kevin Hart or Dave Chappelle!
As you may know, many online platforms use AI language models as the first line of customer service, and then later, if there’s a need, they redirect you to a real human. Talking to this AI can be dry and impersonal. Still, once the language models evolve, the possibilities are endless, as we can expect more sophisticated humour capabilities adding a more human touch. Online services, especially those prioritising customer service like leading bingo games platforms, are closely watching this space, impatiently looking forward to taking their offerings to the next level.
ChatGPT is a testament to the remarkable strides in AI, but when it comes to humour, it’s still learning the ropes. It’s fun and can occasionally elicit a chuckle, but it’s not about to headline a comedy club anytime soon. For now, the nuanced art of making humans laugh remains just that – an art, uniquely human.