There’s this conversation that keeps happening, and… ok. Ok. This is the post that finally set me off.
The replies pointed out something crucial, something that makes this whole debate even more infuriating: Some of us actually had to learn English. …
I am a writer. A writer who also happens to be Kenyan. And I have come to this thesis statement: I don’t write like ChatGPT. ChatGPT, in its strange, disembodied, globally-sourced way, writes like me. Or, more accurately, it writes like the millions of us who were pushed through a very particular educational and societal pipeline, a pipeline deliberately designed to sandpaper away ambiguity, and forge our thoughts into a very specific, very formal, and very impressive shape.
There’s a growing community (cult?) of self-proclaimed AI detectives, who have designed and detailed what they consider tells, and armed their followers with a checklist of robotic tells. Does a piece of text use words like ‘furthermore’, ‘moreover’, ‘consequently’, ‘otherwise’ or ‘thusly’? Does it build its arguments using perfectly parallel structures, such as the classic “It is not only X, but also Y”? Does it arrange its key points into neat, logical triplets for maximum rhetorical impact?
To these detectives of digital inauthenticity, I say: Friend, welcome to a typical Tuesday in a Kenyan classroom, boardroom, or intra-office Teams chat. The very things you identify as the fingerprints of the machine are, in fact, the fossil records of our education.
The bedrock of my writing style was not programmed in Silicon Valley. It was forged in the high-pressure crucible of the Kenya Certificate of Primary Education, or KCPE. For my generation, and the ones that followed, the English Composition paper—and its Kiswahili equivalent, Insha—was not just a test; it was a rite of passage. It was one built up to be a make-or-break moment in life: A forty-minute, high-stakes sprint where your entire future, your admission to a good national high school, and by extension, your life’s trajectory, could pivot on your ability to deploy a rich vocabulary and a sophisticated sentence structure under immense, suffocating pressure.
And that one moment wasn’t an aberration. Every English class and every homework assignment for three years prior (and more, it could be argued) was specifically designed to get the teacher marking your composition to award you a mark as close as possible to the maximum of 40. Scored a 38/40? Beloved, whoever is marking your paper has deemed you worthy of breathing the same air as Malkiat Singh.
It’s a memory that’s hard to write over—the prompt, written in the looping, immaculate cursive of the teacher on the blackboard: “A holiday I will never forget.” Or perhaps it was one of those that demanded that you end the entire composition with, “…and that’s when I woke up and realised it was just a dream.” The topic was almost irrelevant. The real test was the execution.
There were unspoken rules, commandments passed down from teacher to student, year after year. The first commandment? Thou shalt begin with a proverb or a powerful opening statement. “Haste makes waste,” we would write, before launching into a tale about rushing to the market and forgetting the money. The second? Thou shalt demonstrate a wide vocabulary. You didn’t just ‘walk’; you ‘strode purposefully’, ‘trudged wearily’, or ‘ambled nonchalantly’. You didn’t just ‘see’ a thing; you ‘beheld a magnificent spectacle’. Our exercise books were filled with lists of these “wow words,” their synonyms and antonyms drilled into us like multiplication tables.
The third, and perhaps most important commandment, was that of structure. An essay had to be a perfect edifice. The introduction was the foundation, the body was the walls, and the conclusion was the roof, neatly summarising the moral of the story and, if you were clever, circling back to the introductory proverb to create a satisfying, if predictable, loop. We were taught to build our paragraphs around a strong topic sentence. We were taught the sin of the sentence fragment and the virtue of the compound-complex sentence. Our teachers, armed with red pens that bled judgment all over our pages, were our original algorithms, training us on a specific model of “good” writing. Our model compositions, the perfect essays from past students read aloud to the class, were our training data.
And that’s a culture that is carried over into high school, where set books must be memorised, and arguments for or against certain statements must be elaborately made for you do reach and surpass the English literature passmark. You could literally recite Shakespeare in the middle of the night right before any exam.
This style has a history, of course, a history far older than the microchip: It is a direct linguistic descendant of the British Empire. The English we were taught was not the fluid, evolving language of modern-day London or California, filled with slang and convenient abbreviations. It was the Queen’s English, the language of the colonial administrator, the missionary, the headmaster. It was the language of the Bible, of Shakespeare, of the law. It was a tool of power, and we were taught to wield it with precision. Mastering its formal cadences, its slightly archaic vocabulary, its rigid grammatical structures, was not just about passing an exam. It was a signal. It was proof that you were educated, that you were civilised, that you were ready to take your place in the order of things.
(I’ve tried to resist it, but I can’t help myself, and perhaps you’ve already picked up on it: See the threes?)
In post-independence Kenya, this language didn’t disappear. It simply changed its function. It became the official language, the language of opportunity, the new marker of class and sophistication. The Charles Njonjos and Tom Mboyas of their time used it to stamp their status in society. The ability to speak and write this formal, “correct” English separated the haves from the have-nots. It was the key that unlocked the doors to university, to a corporate job, to a life beyond the village. The educational system, therefore, doubled down on teaching it, preserving it in an almost perfect state, like a museum piece.
And right there is the punchline to this long, historical joke. An “AI”, a large language model, is trained on a vast corpus of text that is overwhelmingly formal. It learns from books published over the last two centuries. It learns from academic papers, from encyclopaedias, from legal documents, from the entire archive of structured human knowledge. It learns to associate intelligence and authority with grammatical precision and logical structure.
The machine, in its quest to sound authoritative, ended up sounding like a KCPE graduate who scored an ‘A’ in English Composition. It accidentally replicated the linguistic ghost of the British Empire.
So, story time:
I went to the top government boarding school in my country by average GPA in our equivalent of the O-Levels / GCSE. My school, which only had 4th and 5th form students (ages 16 and 17 respectively), was a benchmaxxing factory; the school average GPA was effectively the overriding cost function; all 60ish govt boarding schools were ranked nationally by this, and for the trial exams (but not the real final nationwide exam) we knew not only every school’s rank nationally but every student’s rank nationally across all boarding schools, all 8,000+ of us. Roughly half the teachers in my school were Excellent Teachers, a literal translation of an official designation given to something like the top 1% of teachers nationwide in one of the standardised subjects as evaluated by (if I recall correctly) student average GPA on our GCSE equivalent in their subject of specialisation. Every student in my school was a straight-A student in the Lower Certificate of Education (LCE), another nationwide standardised exam taken by all 3rd form students. The results spoke for themselves: something like 1-2% of the half-million-odd students nationwide sitting the GCSE equivalent got straight As back in my day while over 60% of us did; ~0.1% got straight A+ while 3-5% of us did, so ~30x base rates. The top students got full-ride scholarships sponsored by either the government, the central bank, or government-linked O&G companies to study abroad in the UK, Australia, North America etc; this was life-changing for us.
A lot of my peers came from rural areas and didn’t really have good English. They did have great short-term memory, tremendous work ethic and pain tolerance, and a really intense desire to get those life-changing scholarships, because poverty fricking sucks. They also knew that I was one of the few who knew “correct English”. So for instance, to prepare for timed the English essay exams like Marcus wrote above, they would write out 3-5 “general-purpose essays” of several hundred words each, then hand them over to me to fix grammatical errors and suggest esoteric grade-enhancing synonyms (the greater the vocabulary displayed, the more points out of 50 were awarded), then they’d thank me and proceed to memorise these essays wholesale. Come exam time, they’d be presented with 3-4 free response essay prompts, they’d pick the one that most closely matched an essay from the memorised set, then they’d regurgitate it essentially verbatim, except for the introduction which had to be modified to flow from the prompt. These memorised essays were early-gen LLM slop in quality, but they did their job for my peers.
By my peers’ standards I grew up almost obscenely privileged educationally: there was a 1980 edition copy of the Children’s Britannica in my parents’ house, a Reader’s Digest Great Short Stories of the World, etc. So I didn’t have to do all this. But I was also thoroughly spiky in the Bruce Sterling sense, which wouldn’t do in a benchmaxxing factory like my old boarding school. For instance: I loved microbiology growing up; for my 12th birthday present I asked for volume 2 of the standard A-Levels biology textbook because its chapter on cell ultrastructure electrified me, and when I entered boarding school and finally got the chance to express my passion for microbio in an exam I went all out and wrote a 2-page answer to a free-response question with a (IIRC) 5-line response box. I got a big fat zero out of 5 on it because my Excellent Biology Teacher crossed out everything after the box’s 5 lines and only graded for teacher’s passwords in the box. This killed my passion for microbio; it also turned me into a superb guesser of teacher’s passwords, which I’m not sure I’ve completely unlearned. My teacher did her job: I got that A+, and that life-changing full-ride scholarship. This particular anecdote isn’t that related to Marcus’ essay though.
Last random anecdote: in mid-2023 I matched on a dating app with an English schoolteacher whose native language was English. She followed up on my reply to her opener with “did you use ChatGPT for this?”
Back to Marcus Olang:
So, when you read my work—when you see our work—what are you really seeing? Are you seeing a robot’s soulless prose? Or are you seeing the image of our Standard Eight English teacher, Mrs. Amollo, her voice echoing in our minds—a voice that spoke with the clipped, precise accent of a bygone era—reminding us to connect our paragraphs with a suitable linking phrase? Are you seeing an algorithm’s output, or the muscle memory of a thousand handwritten essays, drilled into us until the structure was as natural as breathing?
The question of what makes writing “human” has become dangerously narrow, policed by algorithms that carry the implicit biases of their creators. If humanity is now defined by the presence of casual errors, American-centric colloquialisms, and a certain informal, conversational rhythm, then where does that leave the rest of us? Where does that leave the writer from Lagos, from Mumbai, from Kingston, from right here in Nairobi, who was taught that precision was the highest form of respect for both the language and the reader?
It is a new frontier of the same old struggle: The struggle to be seen, to be understood, to be granted the same presumption of humanity that is afforded so easily to others. My writing is not a product of a machine. It is a product of my history. It is the echo of a colonial legacy, the result of a rigorous education, and a testament to the effort required to master the official language of my own country.
Before you point your finger and cry “AI!”, I ask you to pause. Consider the possibility that what you’re seeing isn’t a lack of humanity, but a form of humanity you haven’t been trained to recognise. You might be looking at the result of a different education, a different history, a different standard.
You might just be looking at a Kenyan, writing. And we’ve been doing it this way for a very long time.
Fwiw I’m >9:1 confident that I’m Kenyan. I Don’t Write Like ChatGPT. ChatGPT Writes Like Me is written with substantial AI assistance. It gets 100% on Pangram and the tone and rhythm just seems way too AI-like. Maybe you’re like “oh doesn’t that just prove his point? But no.
As some points of comparison, your comment doesn’t read that way at all. Neither does his essays from 2021 and earlier. Maybe you’re like, argumentative styles are different from his usual fare? But I asked Claude to come up with some Kenyan essays from that time period that has this ChatGPT rhythm and Claude was unable to[1]. If his argument was actually correct, you’d expect way more pre-2022 Kenyan writers to sound the way he describes.
EDIT: The biggest reason I could imagine being wrong (and I don’t think this is likely), which is why I’m not at >>99:1, is if he intentionally wrote in an extremely LLM-y way as performance art. I think that’s theoretically possible albeit very hard for such a long essay[2]. But if so that’s also a very different performance/meta-point than the object level points he was making!
Claude thought this sounded AI-y: https://sunwords.com/2019/04/28/i-am-here-to-compete-for-your-attention/ I disagree strongly, as does Pangram. It does have some surface features like em-dashes, but other than that the tempo is very different from AI (and also different from Western blogs/essays I usually read in a way that’s hard to articulate).
I had a few sentences like that here, but it’s very hard to actually sustain a fully AI-slop voice while trying to distinguish yourself as clever and not actually using AI to sophisticated readers. Even dropping the second constraint, I think I’d have a lot of trouble writing the entire story in that style without any AI assistance. Pangram, naturally, flags. my essay as 100% human.
Yes, tbc all his old stuff and several other online Kenyan essayists from 2019 and before I tested got 0% on Pangram, while all his new stuff gets 100% except for, oddly enough, the latest article[1].
But also even if you don’t trust Pangram[2], you can read it yourself and get a sense. I don’t think the AI usage is subtle here.
Fwiw reading it I think this is a false negative on Pangram. I’m not certain here but the cadence feels LLM-y in a way that is maybe harder for the models to pick up on.
I do trust Pangram, more than my own eyes. I focus on content more than form ;) so don’t have the AI writing detectors many do.
So I’m confused. I trust Pangram so much I suspect it’s not a false negative, and that he’s right about ChatGPT writing like he was taught to, but also using ChatGPT and perhaps defending and deflecting with that claim.
Still not sure it matters. Content over form. Ad machina is valid but weak.
(nothing to say other than that I rarely give comments that counterargue a heartfelt take of mine 3 +ve valence emojis; I just really like your comment, thanks)
Somewhat tangential to the questions of whether this essay was AI-written and whether any human actually writes like a LLM, I think linguists now widely agree that LLMs picked a lot of traits from the formal register of African varieties of English when OpenAI (and later other American companies) hired Kenyans and Nigerians (possibly a lot of English teachers among them!) to do RLHF
I think it’s important to acknowledge that part of the reason we all hate LLM writing is that we hate LLMs. There has got to be some motivated reasoning involved. I agree with many of the critiques but not to the same degree after correcting for this.
Also, current-gen AIs will write like whatever you want them to, if you keep reminding them. I haven’t experimented with this a lot, but I do a little co-created fiction just for fun. While they have a default voice, they can change it on request. Even better if you give them writing samples to imitate, which I’ve barely tried.
This is in some part an extension of the “AI will never take MY job!” trope from those who can pretend AI isn’t constantly progressing. The best human writers still beat it, particularly when there’s little effort to make it better. But that’s very likely a hurdle not a wall.
Please don’t lynch me; I don’t like it or use it, either. But it’s not as bad or as static as all that. And like life, that status ain’t nohow permanent.
Writing the above also reminded me of Steve Yegge’s self-description from his classic rant Done, and Gets Things Smart:
Squaaaaawk
So we all think we’re smart for different reasons. Mine was memorization. Smart, eh? In reality I was just a giant, uncomprehending parrot. I got my first big nasty surprise when I was in the Navy Nuclear Power School program in Orlando, Florida, and I was setting historical records for the highest scores on their exams. The courses and exams had been carefully designed over some 30 years to maximize and then test “literal retention” of the material. They gave you all the material in outline form, and made you write it in your notebook, and your test answers were graded on edit-distance from the original notes. (I’m not making this up or exaggerating in the slightest.) They had set up the ultimate parrot game, and I happily accepted. I memorized the entire notebooks word-for-word, and aced their tests.
They treated me like some sort of movie star — that is, until the Radar final lab exam in electronics school, in which we had to troubleshoot an actual working (well, technically, not-working) radar system. I failed spectacularly: I’d arguably set another historical record, because I had no idea what to do. I just stood there hemming and hawing and pooing myself for three hours. I hadn’t understood a single thing I’d memorized. Hey man, I was just playing their game! But I lost. I mean, I still made it through just fine, but I lost the celebrity privileges in a big way.
Having a good memory is a serious impediment to understanding. It lets you cheat your way through life. I’ve never learned to read sheet music to anywhere near the level I can play (for both guitar and piano.) I have large-ish repertoires and, at least for guitar, good technique from lots of lessons, but since I could memorize the sheet music in one sitting, I never learned how to read it faster than about a measure a minute. (It’s not a photographic memory—I have to work a little to commit it to memory. But it was a lot less work than learning to read the music.) And as a result, my repertoire is only a thousandth what it could be if I knew how to read.
My memory (and, you know, overall laziness) has made me musically illiterate.
This was also memorably interesting to me, because teenage me was a maxed-out parrot like young Yegge: I had stupendous short-term recall without commensurate deep understanding. I used to read history textbooks overnight and perfect-score multiple-choice question exams the next morning (right before crashing from sleep deprivation). I also happened to be good at high school math & physics. So people who overindexed “smarts” on recall of narrow facts and math+physics were invariably confounded by their later realisation that I was nowhere near as off-the-charts in other cognitive domains.
Marcus Olang’s I’m Kenyan. I Don’t Write Like ChatGPT. ChatGPT Writes Like Me is interesting because my pre-university education mirrored his due to shared British colonial history, so it brought back a lot of memories:
So, story time:
I went to the top government boarding school in my country by average GPA in our equivalent of the O-Levels / GCSE. My school, which only had 4th and 5th form students (ages 16 and 17 respectively), was a benchmaxxing factory; the school average GPA was effectively the overriding cost function; all 60ish govt boarding schools were ranked nationally by this, and for the trial exams (but not the real final nationwide exam) we knew not only every school’s rank nationally but every student’s rank nationally across all boarding schools, all 8,000+ of us. Roughly half the teachers in my school were Excellent Teachers, a literal translation of an official designation given to something like the top 1% of teachers nationwide in one of the standardised subjects as evaluated by (if I recall correctly) student average GPA on our GCSE equivalent in their subject of specialisation. Every student in my school was a straight-A student in the Lower Certificate of Education (LCE), another nationwide standardised exam taken by all 3rd form students. The results spoke for themselves: something like 1-2% of the half-million-odd students nationwide sitting the GCSE equivalent got straight As back in my day while over 60% of us did; ~0.1% got straight A+ while 3-5% of us did, so ~30x base rates. The top students got full-ride scholarships sponsored by either the government, the central bank, or government-linked O&G companies to study abroad in the UK, Australia, North America etc; this was life-changing for us.
A lot of my peers came from rural areas and didn’t really have good English. They did have great short-term memory, tremendous work ethic and pain tolerance, and a really intense desire to get those life-changing scholarships, because poverty fricking sucks. They also knew that I was one of the few who knew “correct English”. So for instance, to prepare for timed the English essay exams like Marcus wrote above, they would write out 3-5 “general-purpose essays” of several hundred words each, then hand them over to me to fix grammatical errors and suggest esoteric grade-enhancing synonyms (the greater the vocabulary displayed, the more points out of 50 were awarded), then they’d thank me and proceed to memorise these essays wholesale. Come exam time, they’d be presented with 3-4 free response essay prompts, they’d pick the one that most closely matched an essay from the memorised set, then they’d regurgitate it essentially verbatim, except for the introduction which had to be modified to flow from the prompt. These memorised essays were early-gen LLM slop in quality, but they did their job for my peers.
By my peers’ standards I grew up almost obscenely privileged educationally: there was a 1980 edition copy of the Children’s Britannica in my parents’ house, a Reader’s Digest Great Short Stories of the World, etc. So I didn’t have to do all this. But I was also thoroughly spiky in the Bruce Sterling sense, which wouldn’t do in a benchmaxxing factory like my old boarding school. For instance: I loved microbiology growing up; for my 12th birthday present I asked for volume 2 of the standard A-Levels biology textbook because its chapter on cell ultrastructure electrified me, and when I entered boarding school and finally got the chance to express my passion for microbio in an exam I went all out and wrote a 2-page answer to a free-response question with a (IIRC) 5-line response box. I got a big fat zero out of 5 on it because my Excellent Biology Teacher crossed out everything after the box’s 5 lines and only graded for teacher’s passwords in the box. This killed my passion for microbio; it also turned me into a superb guesser of teacher’s passwords, which I’m not sure I’ve completely unlearned. My teacher did her job: I got that A+, and that life-changing full-ride scholarship. This particular anecdote isn’t that related to Marcus’ essay though.
Last random anecdote: in mid-2023 I matched on a dating app with an English schoolteacher whose native language was English. She followed up on my reply to her opener with “did you use ChatGPT for this?”
Back to Marcus Olang:
Fwiw I’m >9:1 confident that I’m Kenyan. I Don’t Write Like ChatGPT. ChatGPT Writes Like Me is written with substantial AI assistance. It gets 100% on Pangram and the tone and rhythm just seems way too AI-like. Maybe you’re like “oh doesn’t that just prove his point? But no.
As some points of comparison, your comment doesn’t read that way at all. Neither does his essays from 2021 and earlier. Maybe you’re like, argumentative styles are different from his usual fare? But I asked Claude to come up with some Kenyan essays from that time period that has this ChatGPT rhythm and Claude was unable to[1]. If his argument was actually correct, you’d expect way more pre-2022 Kenyan writers to sound the way he describes.
EDIT: The biggest reason I could imagine being wrong (and I don’t think this is likely), which is why I’m not at >>99:1, is if he intentionally wrote in an extremely LLM-y way as performance art. I think that’s theoretically possible albeit very hard for such a long essay[2]. But if so that’s also a very different performance/meta-point than the object level points he was making!
Claude thought this sounded AI-y: https://sunwords.com/2019/04/28/i-am-here-to-compete-for-your-attention/ I disagree strongly, as does Pangram. It does have some surface features like em-dashes, but other than that the tempo is very different from AI (and also different from Western blogs/essays I usually read in a way that’s hard to articulate).
I had a few sentences like that here, but it’s very hard to actually sustain a fully AI-slop voice while trying to distinguish yourself as clever and not actually using AI to sophisticated readers. Even dropping the second constraint, I think I’d have a lot of trouble writing the entire story in that style without any AI assistance. Pangram, naturally, flags. my essay as 100% human.
So you ran Pangram on his old stuff and other writing from similarly educated people and it didn’t read as mostly AI?
I’m sure he wrote this more like an LLM; this would happen accidentally if that’s the topic and claim on his mind, and why not do it on purpose?
Yes, tbc all his old stuff and several other online Kenyan essayists from 2019 and before I tested got 0% on Pangram, while all his new stuff gets 100% except for, oddly enough, the latest article[1].
But also even if you don’t trust Pangram[2], you can read it yourself and get a sense. I don’t think the AI usage is subtle here.
Fwiw reading it I think this is a false negative on Pangram. I’m not certain here but the cadence feels LLM-y in a way that is maybe harder for the models to pick up on.
Which is a mistake
I do trust Pangram, more than my own eyes. I focus on content more than form ;) so don’t have the AI writing detectors many do.
So I’m confused. I trust Pangram so much I suspect it’s not a false negative, and that he’s right about ChatGPT writing like he was taught to, but also using ChatGPT and perhaps defending and deflecting with that claim.
Still not sure it matters. Content over form. Ad machina is valid but weak.
(nothing to say other than that I rarely give comments that counterargue a heartfelt take of mine 3 +ve valence emojis; I just really like your comment, thanks)
Somewhat tangential to the questions of whether this essay was AI-written and whether any human actually writes like a LLM, I think linguists now widely agree that LLMs picked a lot of traits from the formal register of African varieties of English when OpenAI (and later other American companies) hired Kenyans and Nigerians (possibly a lot of English teachers among them!) to do RLHF
I think it’s important to acknowledge that part of the reason we all hate LLM writing is that we hate LLMs. There has got to be some motivated reasoning involved. I agree with many of the critiques but not to the same degree after correcting for this.
Also, current-gen AIs will write like whatever you want them to, if you keep reminding them. I haven’t experimented with this a lot, but I do a little co-created fiction just for fun. While they have a default voice, they can change it on request. Even better if you give them writing samples to imitate, which I’ve barely tried.
This is in some part an extension of the “AI will never take MY job!” trope from those who can pretend AI isn’t constantly progressing. The best human writers still beat it, particularly when there’s little effort to make it better. But that’s very likely a hurdle not a wall.
Please don’t lynch me; I don’t like it or use it, either. But it’s not as bad or as static as all that. And like life, that status ain’t nohow permanent.
On this note I’ve been eagerly awaiting the announcement of the winners of the recently-concluded Unslop Prize.
Writing the above also reminded me of Steve Yegge’s self-description from his classic rant Done, and Gets Things Smart:
This was also memorably interesting to me, because teenage me was a maxed-out parrot like young Yegge: I had stupendous short-term recall without commensurate deep understanding. I used to read history textbooks overnight and perfect-score multiple-choice question exams the next morning (right before crashing from sleep deprivation). I also happened to be good at high school math & physics. So people who overindexed “smarts” on recall of narrow facts and math+physics were invariably confounded by their later realisation that I was nowhere near as off-the-charts in other cognitive domains.