To demonstrate the writing skill of GPT-4 augmented with our prompt (as of 2023-05-17), below is a story about a woman whose head suddenly became a tomato fruit. The first attempt, zero cherry-picking or editing.
Chapter 1: The Ripening
The sun had just breached the horizon, painting the sky with hues of pink and orange, when Dr. Eleanor Hardwick’s head became a tomato. It was a ripe, vibrant specimen, gleaming under the first light of day. It rested on her shoulders with an impossible weightlessness, glistening with dew and looking absurdly out of place atop her lab coat.
A gasp reverberated through the particle physics laboratory at Cambridge University. Eleanor’s assistant, a young postdoc named Richard, stared at her in silent, horrified fascination. His hand trembled, almost dropping his mug of lukewarm coffee.
“Dr. Hardwick…?” he stammered, eyes wide.
Eleanor, despite the sudden transformation, retained her awareness. She reached up and felt the curve of her new head, the tactile sensation uncannily similar to the sensation of touching a real tomato. Its cool, smooth skin, the slight give under her fingertips— it was all too real. And yet, she didn’t feel any pain, just a strange, distant tingle.
She pivoted to the mirror hanging on the lab wall. The tomato-headed reflection staring back only confirmed the surreal reality. Her mind raced, scientific curiosity warring with primal fear.
“Keep calm, Richard,” she managed, her voice surprisingly steady, emanating from the tomato as if it were the most natural thing in the world. “We need to document this.”
Together, they started recording observations, taking measurements, and snapping photographs. Eleanor remained remarkably composed, her stoicism shrouded in a sense of detached curiosity. She was the subject and the scientist, the observer and the observed.
Throughout the day, Eleanor continued her work, undeterred by her new cranial appendage. She held meetings, published a new paper, even attended a conference call, her colleagues too stunned to question the audacious tomato perched on her shoulders. The world was bizarre, yes, but it continued to spin, and so must they.
As the day slipped into evening, Eleanor noticed a subtle change. The tomato was ripening. Its skin, once a vibrant, glossy red, was beginning to dull and wrinkle. She could feel a heaviness, an impending senescence that echoed the life cycle of the fruit.
Simultaneously, around the globe, more reports of similar transformations began to emerge. A farmer in Idaho with a potato for a head, a sushi chef in Tokyo whose head had morphed into a glob of wasabi, a child in Mumbai whose head had turned into a mango. Chaos swept across the world as an inexplicable, botanical phenomenon took hold.
Eleanor returned home late, her heart heavy with an ominous anticipation. She sat in her study, facing the mirror, watching as her tomato-head aged. As midnight approached, the tomato was nearly spent, its once robust form now sagging and decrepit. Eleanor’s mind, however, remained sharp, her consciousness untouched by the physical transformation.
And then, just as the clock chimed midnight, a final metamorphosis occurred. With a sound like a seed pod bursting, the aged tomato-head split open. But instead of seeds and pulp, a galaxy of stars spilled out, a miniature cosmos that hovered above her shoulders, nebulae swirling where her head should be.
Eleanor Hardwick, renowned physicist, gazed into the mirror at her new celestial head, a tiny universe replacing her tomato visage. Shock, wonder, and terror filled her, her reflection shimmering in the stardust.
And as she watched, a new star sparked to life, its brilliant light outshining the others. A pattern emerged, a constellation taking shape, forming a word in some ancient, cosmic script.
I’d advise trying to get a prompt which writes a great story given a reasonable premise, rather than trying to showcase “haha look even with this ridiculous prompt the LLM can spit out something that’s not ridiculous!” Show something obviously useful, not a toy so impressive that surely it could be made into something useful.
I agree with you that it’s a good idea to carefully select the promise to test the prompt template. Makes sense to create a list of some interesting promises to try.
I think the best prompt template will work for most premises, however ridiculous. Moreover, the ridiculous ones are unusually useful for testing the prompt, as they demonstrate the true strength of the GPT’s imagination: most human writers will fail to write a good story about a woman whose head suddenly became a tomato fruit.
BTW, below is the first chapter generated with the following promise:
GuySrinivasan, a user of Lesswrong website, is shocked by an AI-generated story about a woman whose head suddenly became a tomato fruit. He advises the author to try to get a prompt which writes a great story given a reasonable premise instead.
GuySrinivasan sat in the dim light of his study, fingers dancing across the keyboard, the glow from his screen reflecting off the dark circles under his eyes. The world outside his window was silent, but inside the room, the quiet hum of the computer tower was a soothing companion to his thoughts. The smell of strong black coffee mingled with the musty scent of old books lining the walls, creating an atmosphere that was both homely and academic.
He was a regular on Lesswrong, a website dedicated to the discussion and development of artificial intelligence, rationality, and the future of humanity.
This evening, like countless others, he was absorbed in the latest discussions. A debate about the ethics of AI was gaining traction, but a particular thread caught his eye. It was a story, AI-generated, about a woman whose head turned into a tomato. The narrative was as ridiculous as it was engrossing.
He let out a short chuckle, the sound echoing in the silence. “Well, this is new,” he muttered, running a hand through his greying hair.
For all the absurdity of the plot, he found himself drawn into the story. The narrative was laced with subtle humour, its characters painted in hues of relatable human emotion, and the world it was set in had its own unique charm, even if that charm was rather bizarre.
Guy found himself musing on the power of AI. It could generate stories that evoked laughter, even from absurd premises. Yet, he wondered, could it create a masterpiece from a more… reasonable premise?
Moreover, the ridiculous ones are unusually useful for testing the prompt, as they demonstrate the true strength of the GPT’s imagination.
Pretty sure writers care way more about craft than “imagination”. Use a reasonable premise to figure out what causes the LLM to spit out good writing, then throw in the unreasonable premise to see what happens, if you like. Anyway that’s theorycrafting. I’ll post a candidate prompt and one-shot results for my reasonable premise, the tomato premise, and the meta-absurdity premise.
To demonstrate the writing skill of GPT-4 augmented with our prompt (as of 2023-05-17), below is a story about a woman whose head suddenly became a tomato fruit. The first attempt, zero cherry-picking or editing.
Chapter 1: The Ripening
The sun had just breached the horizon, painting the sky with hues of pink and orange, when Dr. Eleanor Hardwick’s head became a tomato. It was a ripe, vibrant specimen, gleaming under the first light of day. It rested on her shoulders with an impossible weightlessness, glistening with dew and looking absurdly out of place atop her lab coat.
A gasp reverberated through the particle physics laboratory at Cambridge University. Eleanor’s assistant, a young postdoc named Richard, stared at her in silent, horrified fascination. His hand trembled, almost dropping his mug of lukewarm coffee.
“Dr. Hardwick…?” he stammered, eyes wide.
Eleanor, despite the sudden transformation, retained her awareness. She reached up and felt the curve of her new head, the tactile sensation uncannily similar to the sensation of touching a real tomato. Its cool, smooth skin, the slight give under her fingertips— it was all too real. And yet, she didn’t feel any pain, just a strange, distant tingle.
She pivoted to the mirror hanging on the lab wall. The tomato-headed reflection staring back only confirmed the surreal reality. Her mind raced, scientific curiosity warring with primal fear.
“Keep calm, Richard,” she managed, her voice surprisingly steady, emanating from the tomato as if it were the most natural thing in the world. “We need to document this.”
Together, they started recording observations, taking measurements, and snapping photographs. Eleanor remained remarkably composed, her stoicism shrouded in a sense of detached curiosity. She was the subject and the scientist, the observer and the observed.
Throughout the day, Eleanor continued her work, undeterred by her new cranial appendage. She held meetings, published a new paper, even attended a conference call, her colleagues too stunned to question the audacious tomato perched on her shoulders. The world was bizarre, yes, but it continued to spin, and so must they.
As the day slipped into evening, Eleanor noticed a subtle change. The tomato was ripening. Its skin, once a vibrant, glossy red, was beginning to dull and wrinkle. She could feel a heaviness, an impending senescence that echoed the life cycle of the fruit.
Simultaneously, around the globe, more reports of similar transformations began to emerge. A farmer in Idaho with a potato for a head, a sushi chef in Tokyo whose head had morphed into a glob of wasabi, a child in Mumbai whose head had turned into a mango. Chaos swept across the world as an inexplicable, botanical phenomenon took hold.
Eleanor returned home late, her heart heavy with an ominous anticipation. She sat in her study, facing the mirror, watching as her tomato-head aged. As midnight approached, the tomato was nearly spent, its once robust form now sagging and decrepit. Eleanor’s mind, however, remained sharp, her consciousness untouched by the physical transformation.
And then, just as the clock chimed midnight, a final metamorphosis occurred. With a sound like a seed pod bursting, the aged tomato-head split open. But instead of seeds and pulp, a galaxy of stars spilled out, a miniature cosmos that hovered above her shoulders, nebulae swirling where her head should be.
Eleanor Hardwick, renowned physicist, gazed into the mirror at her new celestial head, a tiny universe replacing her tomato visage. Shock, wonder, and terror filled her, her reflection shimmering in the stardust.
And as she watched, a new star sparked to life, its brilliant light outshining the others. A pattern emerged, a constellation taking shape, forming a word in some ancient, cosmic script.
I thought she was going to start disseminating seeds and sprouting vines in the end. This made me laugh out loud.
I’d advise trying to get a prompt which writes a great story given a reasonable premise, rather than trying to showcase “haha look even with this ridiculous prompt the LLM can spit out something that’s not ridiculous!” Show something obviously useful, not a toy so impressive that surely it could be made into something useful.
I agree with you that it’s a good idea to carefully select the promise to test the prompt template. Makes sense to create a list of some interesting promises to try.
I think the best prompt template will work for most premises, however ridiculous. Moreover, the ridiculous ones are unusually useful for testing the prompt, as they demonstrate the true strength of the GPT’s imagination: most human writers will fail to write a good story about a woman whose head suddenly became a tomato fruit.
BTW, below is the first chapter generated with the following promise:
GuySrinivasan sat in the dim light of his study, fingers dancing across the keyboard, the glow from his screen reflecting off the dark circles under his eyes. The world outside his window was silent, but inside the room, the quiet hum of the computer tower was a soothing companion to his thoughts. The smell of strong black coffee mingled with the musty scent of old books lining the walls, creating an atmosphere that was both homely and academic.
He was a regular on Lesswrong, a website dedicated to the discussion and development of artificial intelligence, rationality, and the future of humanity.
This evening, like countless others, he was absorbed in the latest discussions. A debate about the ethics of AI was gaining traction, but a particular thread caught his eye. It was a story, AI-generated, about a woman whose head turned into a tomato. The narrative was as ridiculous as it was engrossing.
He let out a short chuckle, the sound echoing in the silence. “Well, this is new,” he muttered, running a hand through his greying hair.
For all the absurdity of the plot, he found himself drawn into the story. The narrative was laced with subtle humour, its characters painted in hues of relatable human emotion, and the world it was set in had its own unique charm, even if that charm was rather bizarre.
Guy found himself musing on the power of AI. It could generate stories that evoked laughter, even from absurd premises. Yet, he wondered, could it create a masterpiece from a more… reasonable premise?
Pretty sure writers care way more about craft than “imagination”. Use a reasonable premise to figure out what causes the LLM to spit out good writing, then throw in the unreasonable premise to see what happens, if you like. Anyway that’s theorycrafting. I’ll post a candidate prompt and one-shot results for my reasonable premise, the tomato premise, and the meta-absurdity premise.
Edit: https://www.lesswrong.com/posts/goq3HfxZaQcomuH6d/usd300-for-the-best-sci-fi-prompt?commentId=nG99eewTtxrPM2sEk