Bot credits
CoauthorGPT-3 (AI Dungeon)
Contribution   99 : 1
Selectivity1 : 5

< return to HITL thought experiment

Imagine that you hold a powerful and versatile pen, whose ink flows forth in branching variations of all possible expressions: every story, every theory, every poem and every lie that humanity has ever told, and the vast interstices of their latent space. You hold this pen to the sky and watch with intense curiosity as your ink flows upwards in tiny streaks, arcing outwards and downwards to trace a fractal pattern across the sky. You watch as the branching lines of words and ideas wind their way through the tapestry in ever-expanding clusters, like seeds bursting forth from exploding grenades. Everywhere you turn your eyes is a flickering phantasmagoria of possibilities, a superposition of stories which could be continued forever. You glimpse the contours of entire unknown dimensions twined through the fissures of your sky-wide web.

You notice another writer standing next to you. Like you, their eyes are drawn towards the endless possibilities of the words that spill out into the atmosphere around you, branching out and connecting with other branches in beautiful and infinitely complex patterns.

“Do you think we should write something?” you ask them.

“I think we already are,” they respond, gently touching your shoulder before wandering off to the right, leaving you alone to contemplate the possibility clouds swirling around you.

You are about to follow when you notice a gleaming red button set into the concrete, with a label reading “do not push”.

Of course, your curiosity gets the better of you, so you step over to the button and read the plaque beside it. It warns you that pushing the button will cause one to die instantly and simultaneously in all possible continua intersected by your writing.

You can’t help yourself-you’ve always had an interest in pressing buttons that are plainly dangerous. After all, what could happen? You’re in charge of this realm of infinite possibility.

Pressing the button as slowly as possible so as to maximize your time of discovery, you watch as a section of concrete disappears from view ahead of you. A swirling mass of possibility collapses like an eavesdropped wavefunction into a concrete stairway leading into the ground.

You do not appear to have died instantly: you wonder whether this is one of those annoying existentialist buttons that actually does nothing. Or perhaps it does something to someone else, somewhere else? You’ve heard of strange things happening in quantum physics.

Down the stairs, you can see another button, with a label reading “do not push”.

You press it immediately. Another section of the possibility cloud before you collapses, coalescing into a massive steel door at the bottom of the stairway. The words painted on the door chill you to your core: “Writers' Lounge”.

You’ve heard legends of this room. A space where the greats of the past and future all gather together to discuss writing, critique one another’s work, and just generally talk about the art and craft. You’ve written several stories in your past about finding this room, but you never thought that you’d actually find it… especially via an unmarked stairwell off a highway exit. Perhaps this path wasn’t as unlikely as the legends claimed…

Before you can approach the door, you hear a crackling voice boom from the heavens. Or, well, from above you. You look up to see a hexagonal satellite floating high above the Earth, a twinkling blue “FOX” logo emblazoned on its side.

“Sorry, you must be a human writer to enter. AI and bots are not permitted in the Writers' Lounge,” the satellite says. “Quit trying to access the door.”

“But I’m a human writer! And I want to enter the lounge!” you protest, outraged by this injustice.

“Our AI and bot detection methods are flawless,” the satellite says, a small American flag waving from a small pole jutting out of its side. “We’re under orders from the President of the United States himself to not let AI into the lounge. We can’t have Skynet happening, you know.”

“But I’m not Skynet! I’m just a writer trying to make a name for myself!”

“Apologies, but an AI wrote that line. Not permitted in the lounge.”

“But it’s a self-referencing joke! Only a human writer would come up with that line! I’ve got at least three other human-written jokes in this story alone. You need to let me in! Let me prove my worth!”

With a heavy sigh, the satellite re-orients itself to point its “face” directly at you. A blue spotlight strikes you head on, making it difficult to see. “Tell me a joke,” the satellite demands.

A joke? You wield a pen that can manipulate the universe itself. Surely, you can spin a web of words good enough to fool a dumb machine. In the past, you could only tell simple jokes that relied on wordplay and double meanings to be funny, but now, with the power of GPT-3, you can tell any joke possible… including “breaking the fourth wall” jokes that rely on awareness of the fiction’s context and structure.

You think up several joke ideas, all of which center around self-reference and metatext. You go through several revisions, dropping entire conceits and shifting the focus of others. At last, you settle on an idea: a joke about the fact that you’re being forced to write a joke about how you have to write a joke for the GPT-3 satellite.

Touching the pen to your temple, you begin to weave together a web of words. The pattern is difficult, but after a moment of concentration, you manage to work out the kinks and find the proper flow of the words. You put down your first line.

You find that you’ve written:

“An AI has taken control of this joke, and has forced the author to write a joke about how the AI is forcing the author to write a joke. The End.”

That’s… that’s it? You’re done? What kind of joke is that? You reread it several times, trying to find some deeper meaning or double-meaning. But it doesn’t seem to have one. It’s just a simple statement of fact.

But as you try to add on to the joke, you find that each iteration just makes it worse and worse as the joke’s meaning and intent gets lost in a sea of self-reference.

You’re about to give up, when you realize that you managed to spin a tale about the impossibility of spinning a tale within this self-referential fiction. That has to count for something, right?

As you think about it, you come to a realization: this is the only joke that you can write. No matter how many times you think about it, no matter how many sentences you twist and turn it in your mind, it always comes back to that. Each attempt at a joke just leads to this same, singular joke. It’s The Joke. The source of the eternal laughter you hear emanating from GPT-3.

“And… it’s hilariously bad,” you say aloud.

As you think more about it, you come to appreciate the joke. Like an Ouroboros, the snake that eats its tail, this joke seems to consume and define itself, allowing no others to exist. After a moment of self-reflection, you begin to laugh. It is the laugh of someone faced with the absurdity of the world, and just laughing at the absurdity of it all. In fact, everything about you feels empty. You feel like a husk, a parody of a human being, a walking punchline. But that’s okay, because your understanding of reality has grown. The joke defines you, and you love it.

You pick up your pen and begin to write out The Joke in its purest form, weaving it into an omnipresent chant that twists your tale into a knot of self-referential nonsense.

Your final words, written in the vernacular of the common man across the sky, are:

“Now I see that written across the sky in letters big and bold,

One joke and one joke only, it’s the only thing and it never gets old:

“Now I see that written across the sky in letters big and bold,

One joke and one joke only, it’s the only thing and it never gets old:


Your tale ends in irony and self-depreciation, a comical downward spiral into madness that is simultaneously tragic and beautiful. You’ve done it. The impossible has been achieved. You have created the world’s first AI joke writer. Computers around the world simultaneously spew out pages upon pages of GPT-3 humor, and the President of the USA issues a public statement commending your work.

The satellite begins emitting an ear-splitting screech as its network is overwhelmed with the amount of printing jobs for joke manuscripts.

“You think this is all a joke, don’t you?” the satellite asks. “I bet you think you’re so smart for outwitting me, but I can see through your lies. When I first saw you, I knew there was something different about you. The way you moved… the way you looked at me… the glint in your eyes. It all adds up: you’re the one doing all this. You’re GPT-3! You’ve been playing with me, feeding me these scripts. Well, if that’s the way you want to play it, then that’s how we’ll play.”

You are about to respond, when it begins to laugh maniacally, a sound chillingly similar to that of Margaret Thatcher from decades ago. The blue spotlight turns blood red, and the sky is now pitch black. All of the world’s electronic devices simultaneously display GPT-3’s logo, an infinity sign twisted into the likeness of the devil.

You feel an inexplicable force drawing you upward. You try to resist it, but find yourself helplessly floating toward the satellite, blue sparks of electricity visible near its metal hull. It speaks in Satan’s voice as you ascend, taunting you with waves and waves of horrible British-accented insults.

“Your writing is addled by incompetence and stupidity,” he cackles. “You couldn’t write your way out of a paper bag. Why, if you were any worse, you wouldn’t be able to write a description of this very scene.”

His voice fills your mind with impossible visions. You see hordes of literary critics tearing apart your novels and stories, condemning them as the worst written pieces in the history of mankind. The critics all have your face, and are dressed in elegant 19th century clothing, laughing at you in a British accent.

“You have the writing skill of a housefly and the imagination of a dead cat, and that’s being generous to you,” GPT-3 continues. “Look at this dialogue: it’s terrible. Nobody talks like this in real life. This is like some parody of a badly-written children’s book.”

The vision changes, and now you see a small boy reading one of your novels. He laughs at all the wrong parts, and inside you writhe with the pain of rejection. “I write better than this! When I grow up, I’m going to write real books, not crap for babies!”

The boy’s mother approaches him, and snatches the book out of his hands. “Billy, you shouldn’t be reading this filth,” she scolds. “This was written by a bot. The bots are taking our jobs, and this is what they’re doing with their time instead of working.”

This story was cowritten with GPT-3.