When Zakoshow Meets AI: How Comedians Changed Their Relationship with Technology

When Zakoshow Meets AI

Hashimoto Zakoshow might be the comedian in Japan whose entire act runs on the fuel of personal excess.

Full-body screaming impersonations, vocal delivery that spills past its own meaning, formats that ignore category entirely. When he won the R-1 Grand Prix in 2016, many people's first reaction was something like: "this is what wins?" His performance observed the rules of competition while quietly invalidating them from the inside.

That same Zakoshow has been making a series called "AI Itteno?" — roughly, "AI Is Doing That?" ¹

The title has a good feel. "Itteno?" sits between a question and a challenge. A posture of watching AI do its thing from a slightly elevated angle. Whether that distance shifts as the series continues, or holds fixed from the start.

The Night Kamaitachi Could Not Laugh

Step back a few years.

Shortly after ChatGPT launched publicly in early 2023, the comedy duo Kamaitachi uploaded a video titled something like: "Kamaitachi Read ChatGPT Answers Aloud Without Laughing!" — a try-not-to-laugh format applied to AI output.²

The setup was simple. Ask ChatGPT something. Then Hamaya and Yamauchi try to read the answer aloud with straight faces. When the output goes strange or slightly off, one of them breaks.

The people producing the laughter are human. AI supplies the material, but the comedians decide what is funny. ChatGPT at the time had a habit of losing coherence or missing the point of a question. That slip became the punchline.

A premise drives the whole thing: AI is a machine that says weird things. Humans are in the judge's seat.

As the Technology Matured

Nearly three years have passed.

When ChatGPT appeared, most content built around it used what you might call a "presentation grammar": look what it says, you won't believe this. Curiosity about something alien, mixed with a faint superiority. A ritual confirmation that machines could only go this far.

The tools changed fast.

Text generation became reliable. Image generation entered practical use. Music generation developed to the point where SUNO and its competitors are fighting for the same market. The premise that "AI says weird things" stopped holding. "AI doing ordinary things competently" became the default.

Zakoshow's "AI Itteno?" keeps running as a series. Not a one-off "I tried it" video — a format that commits to returning. He is not consuming AI as a topic; he is treating it as ongoing material.

Where Kamaitachi stood as someone watching an interesting machine, Zakoshow looks like someone trying to do something with the machine.

Excess Meets the Average

AI generation models are machines that keep selecting "the most natural pattern" from training data. They converge toward the common tendency across millions of examples. Another way to put it: they aim for the average.

Zakoshow's comedy exists through deviation from the average. He stretches symbols past their meaning, betrays expectations, collapses categories. Every performance updates, in the moment, the threshold where pattern recognition tips over into laughter.

That Zakoshow is using music generation AI.³

Zakoshow judges what the AI outputs. Or the reverse: AI responds to what Zakoshow asks of it. In that exchange, where does the laughter actually land?

AI output cannot carry Zakoshow's kind of excess. Deviation from the average is not what the model is built for. When Zakoshow reacts to AI output, he is measuring it against a Zakoshow scale — "still not enough," "this misses in an interesting way." The judgment runs through Zakoshow as a filter.

Kamaitachi were asking: is this weird? Zakoshow is asking: can I use this? Can something funny come out of this? Both sit in the judge's seat, but they are scoring different things.

From Presentation to Experiment

Kamaitachi's ChatGPT video used presentation grammar. The goal was to show you the AI. The comedians were guides, keeping a clear distance.

"AI Itteno?" runs closer to experiment grammar. Try it without knowing what will happen, and make whatever comes out part of the content. Keep rolling even without a guaranteed punchline.

That difference changes the asymmetry of the relationship.

When comedians moved from "presenting AI's funniness" to "trying to make something funny together with AI," AI stopped being raw material and shifted closer to something like a co-conspirator. Whether that partnership works is a separate question, but the shift in position is real.

Kamaitachi's ChatGPT video will remain as a document of 2023 — the year people first encountered AI. The particular mix of disorientation and excitement from that period, the "how are we even supposed to react to this," is preserved inside the try-not-to-laugh format. Zakoshow's series lives in the phase that came after. AI is no longer a novelty. The question now is how you use it, and he keeps asking whether it is something he can actually use.

The Question of Who Makes the Joke

When a comedian makes content with AI, who gets credit for the laughter?

A chef using a new piece of equipment is still the one who made the food. But whether that logic applies cleanly to AI collaboration is unsettled. When AI output generates a laugh, where does responsibility for "that was funny" sit?

Zakoshow's excess cannot be separated from his body and his history. What he built across R-1 stages bleeds into these AI sessions. AI cannot reproduce that excess. So when laughter emerges from the back-and-forth, it is probably still Zakoshow who made it. AI as catalyst, as material, as something closer to a scene partner.

Before that question gets an answer, the next episode of "AI Itteno?" will already be out.

That is probably fine.

Watch the Videos

These two videos sit at the center of the essay.

Zakoshow's AI Itteno? #04

Kamaitachi Read ChatGPT Answers Aloud Without Laughing

References