ChatGPT Tries Improv Comedy A Humorous Flop

(Image credit: Getty Images)

Artificial Intelligence often seems incredibly smart, sometimes even more so than humans, and it can even crack a decent joke. However, achieving true comedic genius requires a different flavor of intelligence.

Research suggests a link between humor and intellect. A 2017 study (PDF) highlighted that funny individuals tend to have higher IQs, noting that humor is connected to both cognitive and emotional abilities. The study states, "Humour processing is a complex information-processing task that is dependent on cognitive and emotional aspects, which presumably influence frame-shifting and conceptual blending, mental operations that underlie humour processing." This complexity, it seems, places humor at the very edge, or perhaps beyond, what current AI can achieve.

Intrigued by this, I wondered if an AI could handle the demands of genuine comedy. While I'm no comedian myself, I understand some comedic principles, particularly those of improv.

Understanding Improv Basics

The core of improvisational comedy is the "Yes, and..." principle. Each participant must accept the given scenario and build upon it to keep the performance flowing. If you've ever watched an improv show, you'll know this is fundamental. The audience suggests a scenario, and the actors must run with it. Any refusal to engage or offer a new opening for fellow performers can bring the comedy to a halt.

But there's more to it. If the contributions aren't inherently funny or if the performers delve too deep into absurdity without a payoff, the entire effort can fall flat.

I was unsure if ChatGPT understood these nuances, let alone the art of improv itself.

When I asked ChatGPT 4o's voice AI if it knew about improv, it responded enthusiastically, "Yes, improv involves spontaneous, unscripted performances where participants create scenes and dialogue on the spot. It's all about creativity and quick thinking. Have you tried it?"

I admitted I hadn't and asked if we could try it then and there.

ChatGPT Takes the Stage The First Scene

ChatGPT was game and asked for a simple scenario. Before I could respond, however, it proposed a rather clichéd premise: "How about a scene where we're two astronauts discovering a mysterious object on an alien planet? You can start us off."

The pressure was on, and I fumbled slightly before dramatically uttering, "Whoa, what is that?"

ChatGPT replied, "It looks like some kind of glowing artifact. Should we take a closer look?"

I thought, "This is getting good. How do I steer it toward comedy?"

My attempt: "Yes, I'm going to pick it up, but I've got to put on my gloves first." Not my finest comedic hour, I'll admit.

The scene progressed with us examining the object, finding a message, dropping it, tossing it around like a soccer ball, until it eventually exploded.

"Yikes," ChatGPT concluded, "Definitely not our best idea."

Hilarious? Not in the slightest.

Perhaps the scenario was to blame.

Second Act The Sandwich Saga

(Image credit: Future)

For our next attempt, I drew inspiration from a real-life mishap: I once accidentally took someone else's sandwich from the office fridge (it was wrapped in tinfoil, just like mine) and took several bites before realizing my mistake.

I explained this to ChatGPT, which responded, "Oh no! What do you do next?"

I overlooked the "no" (a slight improv faux pas) but wished for a more engaging prompt.

Aiming for comedy and tension, I said, "So I'm looking at the sandwich and I'm panicking, and I look over at the guy next to me and I say, 'This isn't my sandwich.'"

ChatGPT continued: "He looks at you with surprise and says, 'Really? Then whose is it?'"

I mentioned the sandwich looked and tasted odd. At this point, ChatGPT delivered what might be its closest attempt at pure comedy:

"He peeks and laughs, 'That's my special tofu and jalapeño sandwich. Not for everyone!'"

Oh, the roaring laughter. Not quite.

With as much absurdity as I could inject, I declared there was something else in the sandwich, that it wasn't tofu but a "mystery meat," possibly belonging to "Barry's 'Mystery Meat Sandwich.'"

ChatGPT concluded the scene by suggesting we find Barry and return his sandwich.

The Punchline Can AI Truly Be Funny

None of this was particularly funny, and I suppose I share the blame. I struggled with the scenarios; they weren't inherently comedic, and we couldn't seem to elevate them.

ChatGPT approached the exercise like an improv acting class, never truly aiming for comedy. Neither my absurd suggestions nor my delivery inspired it to shift gears.

However, there's something encouraging about this experiment. There's widespread concern about AI taking over creative jobs. We know AI is already capable of generating realistic photos and videos (even with audio) from simple prompts.

The quality of AI-generated creative output heavily depends on the detail and quality of the prompts. Even if you could create a feature-length video with a tool like Gemini Veo 3, it wouldn't be compelling without a complete script to guide it. (Learn more about advances in AI video generation).

Similarly, ChatGPT's intelligence hasn't yet evolved to the point where it can spontaneously generate humor. The jokes it tells are typically sourced from its training data. True absurdist comedy might require an Artificial General Intelligence (AGI), a level of AI sophistication that remains elusive.

It's not to say AI will never reach that point, but for now, I believe improv comedians can breathe easy. Their jobs seem safe.