What Does It Actually Mean to Write ‘Like a Human’?
By Marcus Olang’
Suddenly, everyone’s become an A.I. detective.
We now have checklists of the supposed telltale signs that something was written by a chatbot. There are companies that claim their detection software can catch it for you. Paranoia. Finger-pointing. Witch hunts.
Then last year, that cloud of suspicion fell on Marcus Olang’.
After Olang’, a writer in Nairobi, Kenya, was accused of sounding like ChatGPT, he wondered: What does it mean to write like a robot in the first place?
In the Opinion Video above, Olang’ argues that the frenzy over catching A.I. writing tells us a deeper, older story. It’s not actually about technology, but about what it means to sound, and be, human — how that definition has changed throughout history, and who it leaves behind.
Below is an edited transcript of this video. We recommend watching it in its original form for the full effect. You can do so using the video player above.
Marcus Olang’: Last year, I sent off a project proposal I’d worked on for days. Proper days. And then the reply came back: Could you do a rewrite with a more human touch? It sounds a little like it was written by ChatGPT.
A more human touch. A more human touch.
It’s a sentence that got me wondering: What does it actually mean to sound like a robot? And what does it mean to sound like a human?
Here’s the thing, though: I don’t write like ChatGPT — ChatGPT writes like me. Or more accurately, it writes like the millions of us around the world who were taught a very specific kind of English — taught to polish our thoughts into this exact same shape. Let me explain.
My formal writing journey started in Mrs. Amollo’s English classroom in Nairobi, Kenya, when I was about 9. Mrs. Amollo spoke in a careful, considered accent: the Queen’s English, a colonial relic, an inheritance of sorts. And this English, we learned, had rules.
Show off your vocabulary: You never just walked; you strode purposefully. You didn’t just see a thing; you beheld a magnificent spectacle.
Reach for a simile in every paragraph: Your heart never raced; it pounded like a drum. Purity, we were told, was “as white as snow” — this, in a country that only ever sees snow atop Mt. Kenya. And not many people see the top of Mt. Kenya.
Make your points in threes — always in threes. A beginning, a conflict, a moral. The character, the motivation, the growth. Balanced, weighted, complete.
Additionally, you never let a thought merely stand on its own. Therefore, you brace it with a transition. Consequently, your reader is never left wondering where you’re headed.
And the highest art of them all: the contrasting parallel. Say a thing by naming its opposite. It’s not the size of the dog in the fight; it’s the size of the fight in the dog. It’s not what you know; it’s who you know. It is not this; it’s that.
And you bring it all together in the long sentence — the longer, the better. Every thought telegraphed, announced and delivered, the whole procession rolling on in one measured breath until it arrives — finally, gracefully, completely — at its full stop.
This was the formula to become a proper writer — a proper human writer. If you studied English in a former British colony or territory, all of this may sound familiar to you.
All those lessons built to one moment: the national exam at the end of primary school. Forty minutes to write a single original story by hand on a double-sided A4 paper. An examinable demonstration of your understanding of all of those rules. And riding on it: your chance at a place in a good high school, then a good university, then a good career. Which is to say, the rest of your life.
And then ——
Archival news clip: A new artificial intelligence program called ChatGPT.Archival news clip: ChatGPT.Archival news clip: Humanlike writing ability.Archival news clip: Humanlike text.Archival news clip: But with more humanlike features.
Somebody feeds this machine what exactly? A vast archive of human language including, yes, centuries of formal, published English. The books, the law, the scientific journals, the Shakespeare, the Scripture — the same texts that built the register Mrs. Amollo drilled into us. And this machine suddenly comes out the other side sounding like the best student Mrs. Amollo ever had. A machine that suddenly sounds like me.
Social media clip: Ten obvious signs you’re using A.I. for writing.Social media clip: Very easy telltale signs that something was written by ChatGPT.Podcast clip: Longer sentences, paragraphs are rarely interrupted with short, punchy statements.Social media clip: The rule of three, or listicles when it doesn’t need to be there.Social media clip: Weird transitions like “moreover.”Social media clip: Instead of saying, “This building is a gallery,” it says, “This building serves as a gallery.”Social media clip: The classic “It’s not about X, it’s about Y” line.Podcast clip: I just saw this referred to as “The DiGiorno Construct.”Clip of archival DiGiorno Pizza ad: It’s not delivery. It’s DiGiorno.Social media clip: Most people just don’t speak or write like that.Social media clip: If this seems like you, it might be time to put down the A.I. and read a book.
A while back, a big Silicon Valley investor announced that if your email included the word “delve,” then a robot wrote it. Nobody real talks or writes like that, was basically his point.
And to that, the whole British Commonwealth — from Lagos to Nairobi to Mumbai — looked over the tops of their glasses and went: My brother in Christ, “delve” is just English. It’s the English we were taught. We shan’t be judged because we read more English than you, please.
But while we’re here, let’s talk about “delve.” Or, in the original Old English, “delfan,” which means “to dig.” Centuries later, Shakespeare put it in “Hamlet”: “But I will delve one yard below their mines.”
Richard Burton as Hamlet on Broadway, 1964: But I will delve one yard below their mines.David Tennant as Hamlet, Royal Shakespeare Company, 2008: Delve one yard below their mines / And blow them at the moon.
And that very English made its way onto a ship to my part of the world on the tongue of the missionary. And that missionary passed it on to the head teacher — part of the proper English we were told to master if we wanted to be considered educated, to be considered civilized, to be considered human.
And while written English in America or Britain kept evolving, what we were taught in our classrooms didn’t. Even after Kenya’s independence, these lessons were cyclically imposed, repeatedly drilled, then calcified. They remained entrenched, frozen in time.
We internalized them so thoroughly that “delve” — a word that now sounds archaic to American or British ears — stayed perfectly ordinary in Lagos and Nairobi. Boring, even; a word you’d casually, nonchalantly throw into Pages 2, 12 or 22 of a report.
So how did this everyday word in Nairobi become a chatbot’s biggest tell? Well, maybe it’s because when A.I. companies needed cheap labor to help teach their new machines how to distinguish between “good” and “bad” writing, guess who they hired? Data workers in Kenya, for a couple of dollars an hour.
Picture it: Someone in Nairobi sits in front of two sample answers and picks the one that looks like the good writing they were taught — the one that uses words like “delve” — a couple of hundred times a day. Then, a couple of years later, a man in California looks at it and says: Yep, that’s for sure a robot word. And it’s not just some tech bro posting on Twitter. (I refuse to call it X.)
When researchers tested A.I.-detection software on essays written by nonnative English speakers, the software incorrectly flagged more than half as A.I.-generated. Now, guess how many essays by American students got the this-is-human stamp of approval? Go on, guess. Almost all of them. And the irony is there’s now a whole industry built on selling us a solution.
Social media clip: Just come over to StealthGPT and go to their A.I. humanizer.Social media clip: Click “humanize” and Humbot will do all the work for you.
That’s right. A website — another A.I. tool — that says it will help you “humanize” your writing and pass A.I. detection.
Social media clip: Just paste in your text, click “humanize,” and it rewrites your text with more natural words and sentences.
So that’s the whole loop. A classroom the British Empire built teaches and conditions us to write a certain way. A machine learns that way — partly taught by us. A second machine gets built to catch the first one, often flagging the exact features that were drilled into us.
And now we’re sold a new machine — a “humanizer” machine — to rough our English back up. Chopping up a long sentence here, dropping a “furthermore” there, so the previous machine and its believers can finally accept that we’re human.
Do you see how the goal posts keep moving?
We spend a decade and a half mastering formal English to be taken seriously in a world that needs convincing. And now that we’ve succeeded? The rules change — again. Now “human” means the slang — the loose, the casual, the imperfect. The fluid, evolving language of modern-day London and Los Angeles. The perfect is no longer “human” — the error is.
So “human” was never a real place we were meant to reach. It’s just the word for wherever we happen not to be. Which tells us that this — the whole panic over catching A.I. — was never just about technology. Perhaps it’s about how we define what “human” is in the first place.
No robots were used in the making of this video. The New York Times has sued OpenAI and Microsoft claiming copyright infringement. The companies have denied those claims.
darkzip•1m ago
By Marcus Olang’
Suddenly, everyone’s become an A.I. detective.
We now have checklists of the supposed telltale signs that something was written by a chatbot. There are companies that claim their detection software can catch it for you. Paranoia. Finger-pointing. Witch hunts.
Then last year, that cloud of suspicion fell on Marcus Olang’.
After Olang’, a writer in Nairobi, Kenya, was accused of sounding like ChatGPT, he wondered: What does it mean to write like a robot in the first place?
In the Opinion Video above, Olang’ argues that the frenzy over catching A.I. writing tells us a deeper, older story. It’s not actually about technology, but about what it means to sound, and be, human — how that definition has changed throughout history, and who it leaves behind.
Below is an edited transcript of this video. We recommend watching it in its original form for the full effect. You can do so using the video player above.
Marcus Olang’: Last year, I sent off a project proposal I’d worked on for days. Proper days. And then the reply came back: Could you do a rewrite with a more human touch? It sounds a little like it was written by ChatGPT.
A more human touch. A more human touch.
It’s a sentence that got me wondering: What does it actually mean to sound like a robot? And what does it mean to sound like a human?
Here’s the thing, though: I don’t write like ChatGPT — ChatGPT writes like me. Or more accurately, it writes like the millions of us around the world who were taught a very specific kind of English — taught to polish our thoughts into this exact same shape. Let me explain.
My formal writing journey started in Mrs. Amollo’s English classroom in Nairobi, Kenya, when I was about 9. Mrs. Amollo spoke in a careful, considered accent: the Queen’s English, a colonial relic, an inheritance of sorts. And this English, we learned, had rules.
Show off your vocabulary: You never just walked; you strode purposefully. You didn’t just see a thing; you beheld a magnificent spectacle.
Reach for a simile in every paragraph: Your heart never raced; it pounded like a drum. Purity, we were told, was “as white as snow” — this, in a country that only ever sees snow atop Mt. Kenya. And not many people see the top of Mt. Kenya.
Make your points in threes — always in threes. A beginning, a conflict, a moral. The character, the motivation, the growth. Balanced, weighted, complete.
Additionally, you never let a thought merely stand on its own. Therefore, you brace it with a transition. Consequently, your reader is never left wondering where you’re headed.
And the highest art of them all: the contrasting parallel. Say a thing by naming its opposite. It’s not the size of the dog in the fight; it’s the size of the fight in the dog. It’s not what you know; it’s who you know. It is not this; it’s that.
And you bring it all together in the long sentence — the longer, the better. Every thought telegraphed, announced and delivered, the whole procession rolling on in one measured breath until it arrives — finally, gracefully, completely — at its full stop.
This was the formula to become a proper writer — a proper human writer. If you studied English in a former British colony or territory, all of this may sound familiar to you.
All those lessons built to one moment: the national exam at the end of primary school. Forty minutes to write a single original story by hand on a double-sided A4 paper. An examinable demonstration of your understanding of all of those rules. And riding on it: your chance at a place in a good high school, then a good university, then a good career. Which is to say, the rest of your life.
And then ——
Archival news clip: A new artificial intelligence program called ChatGPT.Archival news clip: ChatGPT.Archival news clip: Humanlike writing ability.Archival news clip: Humanlike text.Archival news clip: But with more humanlike features.
Somebody feeds this machine what exactly? A vast archive of human language including, yes, centuries of formal, published English. The books, the law, the scientific journals, the Shakespeare, the Scripture — the same texts that built the register Mrs. Amollo drilled into us. And this machine suddenly comes out the other side sounding like the best student Mrs. Amollo ever had. A machine that suddenly sounds like me.
Social media clip: Ten obvious signs you’re using A.I. for writing.Social media clip: Very easy telltale signs that something was written by ChatGPT.Podcast clip: Longer sentences, paragraphs are rarely interrupted with short, punchy statements.Social media clip: The rule of three, or listicles when it doesn’t need to be there.Social media clip: Weird transitions like “moreover.”Social media clip: Instead of saying, “This building is a gallery,” it says, “This building serves as a gallery.”Social media clip: The classic “It’s not about X, it’s about Y” line.Podcast clip: I just saw this referred to as “The DiGiorno Construct.”Clip of archival DiGiorno Pizza ad: It’s not delivery. It’s DiGiorno.Social media clip: Most people just don’t speak or write like that.Social media clip: If this seems like you, it might be time to put down the A.I. and read a book.
A while back, a big Silicon Valley investor announced that if your email included the word “delve,” then a robot wrote it. Nobody real talks or writes like that, was basically his point.
And to that, the whole British Commonwealth — from Lagos to Nairobi to Mumbai — looked over the tops of their glasses and went: My brother in Christ, “delve” is just English. It’s the English we were taught. We shan’t be judged because we read more English than you, please.
But while we’re here, let’s talk about “delve.” Or, in the original Old English, “delfan,” which means “to dig.” Centuries later, Shakespeare put it in “Hamlet”: “But I will delve one yard below their mines.”
Richard Burton as Hamlet on Broadway, 1964: But I will delve one yard below their mines.David Tennant as Hamlet, Royal Shakespeare Company, 2008: Delve one yard below their mines / And blow them at the moon.
And that very English made its way onto a ship to my part of the world on the tongue of the missionary. And that missionary passed it on to the head teacher — part of the proper English we were told to master if we wanted to be considered educated, to be considered civilized, to be considered human.
And while written English in America or Britain kept evolving, what we were taught in our classrooms didn’t. Even after Kenya’s independence, these lessons were cyclically imposed, repeatedly drilled, then calcified. They remained entrenched, frozen in time.
We internalized them so thoroughly that “delve” — a word that now sounds archaic to American or British ears — stayed perfectly ordinary in Lagos and Nairobi. Boring, even; a word you’d casually, nonchalantly throw into Pages 2, 12 or 22 of a report.
So how did this everyday word in Nairobi become a chatbot’s biggest tell? Well, maybe it’s because when A.I. companies needed cheap labor to help teach their new machines how to distinguish between “good” and “bad” writing, guess who they hired? Data workers in Kenya, for a couple of dollars an hour.
Picture it: Someone in Nairobi sits in front of two sample answers and picks the one that looks like the good writing they were taught — the one that uses words like “delve” — a couple of hundred times a day. Then, a couple of years later, a man in California looks at it and says: Yep, that’s for sure a robot word. And it’s not just some tech bro posting on Twitter. (I refuse to call it X.)
When researchers tested A.I.-detection software on essays written by nonnative English speakers, the software incorrectly flagged more than half as A.I.-generated. Now, guess how many essays by American students got the this-is-human stamp of approval? Go on, guess. Almost all of them. And the irony is there’s now a whole industry built on selling us a solution.
Social media clip: Just come over to StealthGPT and go to their A.I. humanizer.Social media clip: Click “humanize” and Humbot will do all the work for you.
That’s right. A website — another A.I. tool — that says it will help you “humanize” your writing and pass A.I. detection.
Social media clip: Just paste in your text, click “humanize,” and it rewrites your text with more natural words and sentences.
So that’s the whole loop. A classroom the British Empire built teaches and conditions us to write a certain way. A machine learns that way — partly taught by us. A second machine gets built to catch the first one, often flagging the exact features that were drilled into us.
And now we’re sold a new machine — a “humanizer” machine — to rough our English back up. Chopping up a long sentence here, dropping a “furthermore” there, so the previous machine and its believers can finally accept that we’re human.
Do you see how the goal posts keep moving?
We spend a decade and a half mastering formal English to be taken seriously in a world that needs convincing. And now that we’ve succeeded? The rules change — again. Now “human” means the slang — the loose, the casual, the imperfect. The fluid, evolving language of modern-day London and Los Angeles. The perfect is no longer “human” — the error is.
So “human” was never a real place we were meant to reach. It’s just the word for wherever we happen not to be. Which tells us that this — the whole panic over catching A.I. — was never just about technology. Perhaps it’s about how we define what “human” is in the first place.
No robots were used in the making of this video. The New York Times has sued OpenAI and Microsoft claiming copyright infringement. The companies have denied those claims.