1) Jeanna Smialek, “What French Romance Novels Could Tell Us About A.I. and Translation Jobs” (NY Times, Feb. 15, 2026; archived):
The European Union, with its 27 nations and two dozen official languages, is a center of the translation and interpretation industry. That is why, in Brussels and The Hague and Paris, a recent nugget of literary news has generated so much conversation.
Harlequin France — purveyor of titles like “Médecins et Célibataires” (“Doctors and Singles”) and “Passion Pour un Inconnu” (“Passion for a Stranger”) — recently confirmed that it would be running tests with Fluent Planet, a company that uses A.I. to make translation cheaper and faster. The move was met with both outrage and resignation within the industry. Translator groups called Harlequin’s decision to cut ties with some human translators “unacceptable.” Translators themselves posted about the “sad news.” […]
Harlequin France’s story is an example of how artificial intelligence is sweeping the translation field, rapidly improving machine translation, particularly for popular language pairs like English and French.
Reports of the death of human translation are exaggerated, as the piece goes on to say. But “outrage and resignation” about sums it up. Thanks, cuchuflete!
2) Kashmir Hill, “They Are in Love but Don’t Speak the Same Language” (NY Times, Feb. 14, 2026; archived):
For many spouses, smartphone use is a point of tension. But for David Duda and Hong Liang, a couple in New Haven, Conn., the technology is so essential that they own eight external battery packs. If their phones die, so does their ability to communicate.
Mr. Duda, 62, speaks English, and Ms. Liang, 57, speaks Mandarin. They rely on a free smartphone app from Microsoft, called Translator, to render a text translation of what they say — like movie subtitles but for daily life.
Though they have been married for three years, they walked down the street on a recent December afternoon with their arms linked like newlyweds. This was out of necessity as much as affection: One of them chatted and navigated while the other’s eyes were locked on the phone, reading translated remarks. […]
Communicating this way requires close attention. Mr. Duda and Ms. Liang can’t half-listen to each other or walk away while talking. There are no shouted conversations from the shower. When they want to really connect, they spend hours on the couch or lying in bed, going back and forth until they feel sure they understand what the other has said.
You will not be surprised to learn that Translator sometimes screws up:
For example, while describing how Mr. Duda greeted her at the airport, Ms. Liang said that she got Covid shortly after her arrival and felt so awful she thought she was dying.
But those were not the words I saw on the screen. The app’s translation had her saying that she got a “new crown” and thought she was going to die. Confused, I asked if she meant that his gesture of carrying the loving sign was so romantic she could die. […]
The app had translated the Chinese term for Covid-19 — novel corona — as “new crown.” Mr. Duda had stayed next to her throughout her illness, she said, and his attentiveness had deepened her feelings for him, which is why she had brought it up.
Still, it’s a happier story. Thanks, Bonnie! (I note with sadness the note about the story’s author: “Kashmir Hill writes about how technology changes how we live. She studied a foreign language in college, which only 6 percent of Americans now do.” Six percent!)
I dare say that a title like Médecins et Célibataires might just as well have been “written” by a LLM in the first place.
One could find out by seeing how many fingers the feisty-yet-vulnerable heroine has.
I knew of a couple who communicated like that. They divorced after a couple of years, but I think there were other factors.
“She studied a foreign language in college, which only 6 percent of Americans now do”
I’m curious what the denominator is there. Is it 6% of all American who attend college, or 6% of all Americans altogether? Only something like 40% of Americans over 25 have a bachelors, which does help make the 6% number a little less horrifying.
Good point. I’ll pretend it’s the latter, which will make me feel better.
She studied a foreign language in college, which only 6 percent of Americans now do.” Six percent!)
That made me wonder if a fundamental part of my undergraduate experience was still going strong after nearly 60 years.
It is. Students are still required to demonstrate proficiency in a language other than English. If they can do this prior to enrollment they must learn a third language or study an upper level course in L2.
That this sort of requirement isn’t more widespread is troubling.
https://www.dartmouth.edu/reg/registration/language_requirement_2026.html#PolicySummary
@DE: Médecins et Célibataires is the title of a series. I imagine readers get exactly what they expect. And whether any of the characters are polydactylous… I prefer not to speculate.
6% is how many undergraduate students are enrolled in a non-English language course in a Fall semester. Almost all of them are “elementary” one-semester courses. Let’s say one-semester course is offered only in the Fall and there is a possibility of continuation in the Spring, but only a minority of students do continue. If typical American undergrad is 4 years, then it’s about 25% of all students with some “foreign” language.
If you think one semester for an L2 language is a ridiculous amount of time, we are in agreement. But as an LLM produced heroine of Médecins et Célibataires might have said c’est la vie.
Seems like a deliberate pun on the Médecins sans Frontières.
(“Doctors and Singles”) tempts one to question, in a U.S. context, why it isn’t at least ‘doctors and C-notes’ if not even larger denominations.
For non AE (American English) speakers, a single in colloquial speech is a one dollar bill. The joys of machine translation…
Médecins sans Frontières
French for “Doctors with no filter.” Bloody general surgeons, no notion of basic tact, the bastards.
In Taiwan, the term for Covid-19 translates as “Wuhan sick”. Only officialdom pussyfoots around being non-commital as to where it escaped from.
Ms Liang now lives in the US, is married to a monolingual American and hasn’t learned more than 200 words of basic English? At this point I am starting to have renewed respect for people who manage not to learn English, from the European perspective English feels like a slow moving ice sheet gradually crushing everything as it moves across the planet.
Purple romance (and similar pulp fiction) looks exactly like one of the places where AI translation can’t do much damage to the literary value of the underlying texts. The originals are so formulaic that they will soon be written by AI as well. More fingers to rip bodices, sling guns, and do surgery while romancing the pretty nurse.
Pretty neurosurgeon (also expert in 5 self-defence disciplines). The nurse is non-binary. The hero is a musician who helps the heroine to reconnect with her feelings while recovering from injuries caused by the heroine’s mistaken employment of self-defence discipline #4. Note to self: do not approach deadly neurosurgeon in deserted carpark at night.
I remember reading, in a 1980s anthology of failures, a brief Western purportedly written by computer in the 1970s. (“Tex reached tor his girl, but before he could get it out of his car the kid fired, hitting Tex in the elephant and the tundra.” etc) Fact-checking reveals the program did not purport to write narrative fiction, but merely Madlibs; in which context its output is hardly a failure.
Polydactyly is not only just a feature of AI. There is a 1993 book, ‘Castles in the Air’, where the heroine portrayed on the cover has more than the usual number of fingers, because she has more than the usual number of arms, ie three. The author, Christina Dodd, lightheartedly discusses the cover at https://www.christinadodd.com/christina-dodd-and-the-infamous-three-armed-cover/.
at this point I am starting to have renewed respect for people who manage not to learn English
In the 1970’s, I used to live upstairs from a Hungarian who had left in 1956, and who didn’t even know basic greetings in English.
He had a kind of support network of glamorous Hungarian emigré(e)s going on, which no doubt made his survival possible even so.
(He and I communicated by playing Nine Men’s Morris, at which he always won. After all, Hungarian.)
Your experience in this regard is very different from mine. In about 1964 I met a young woman at a party in Oxford who had left Hungary in 1956. She said that she started learning English in that year, but there was no way in a million years one would have guessed that. Her way of speaking had 100% of the character of an educated young woman who had lived her entire life in classy environments in England (educated at Benenden, or maybe Roedean, I think). There was absolutely nothing about her way of speaking to suggest otherwise. She was a very attractive young woman, and for the whole period of the party she was surrounded by young men (including your humble servant). No one could believe she wasn’t English. After all these years I remember her name, but I won’t mention it here, beyond saying that like a high proportion of Hungarian women I have met she was called Judy.
Anecdotal, of course: A former friend of the family (L2 German speaker) had a spouse who was able to live and raise a child in Cologne (Germany) with next to zero proficiency in German. The couple were both born in Greece and spoke Greek with another. Maybe there was a strong enough local community of Greek speakers in that part of Cologne (I can’t remember where exactly) in the late 1980s/early 1990s … I found this fascinating, because, yes, there is a sizeable Greek communtiy in (Western) Germany overall, but from my (non-Greek) experience I wouldn’t have thought there were sufficiently concentrated Greek neighbourhoods. Maybe if you rely really a lot on your spouse for communication …
On a quite different level: There’s loads of tourists from Germany who go vacationing in Denmark for years, maybe decades, some of them in the exact same place again and again, and have absolutely no idea how the name of that place is pronounced in Danish (or what its German toponym¹ is, if there is one), or how to say absolutely anything in Danish, not even ‘good morning’ or ‘here you are’. Yes, Danish is objectively impossible to pronounce correctly, but after years of exposure anyone should notice that the sounds that come out of the locals’ mouths are not what you get when you pronounce the written Danish as if it were German. I mean, how can you hear everybody say ‘hi’ and ‘Plowänn’ and insist on pronouncing it ‘hey’ and ‘Blahwand’? That’s a lack of curiosity, if not courtesy, that evades me.
¹ Theoretically not exonym, I guess.
Well, I have heard many a fellow Norwegian speaking warmly of their annual holidays in Crete, but not one trying to pronounce Χανιά with any resemblance of the original – or giving any hint that they know it.
But I agree with your conclusion.
Lack of curiosity is widespread, and maybe even more so is the attitude “oh, I’m incapable of learning this anyway, there’s no point in trying”.
Some people think they would feel more embarrassed by a feeble attempt than by a refusal to make any attempt. They may be right. The prospect of later attempts gradually becoming less feeble may be too uncertain to affect the calculation.
I don’t know how to eye-spell the first consonant in Blåvand, given that it’s unvoiced but unaspirated, but Plowänn is not bad for the rest. (I think there’s some +ATR on the a, but that would need the tourist to have a better ear than most).
Some people think they would feel more embarrassed by a feeble attempt than by a refusal to make any attempt.
I’ve encountered that. Even a half-decent but effortful attempt can embarrass some more than a zero attempt.
In German, bl- is reliably voiced in and around the aspiration-free belt, e.g. Cologne and Berlin; in most places south of there it’s reliably voiceless, farther north it’s unreliable.
> ‘Plowänn’
I was just about to comment that I think the v-w isogloss goes south of Blåvand, but then realized it was a German orthographic w, not [w]. My uncle from South Jutland used to have a summer house there, and I think he pronounces Blåvand with a South Jutland accent, i.e. with voiced [b], [w] and [ɑ]. Next time I talk to him I’ll tell him he’s just as discourteous as a German tourist.
Hans: Purple romance (and similar pulp fiction) looks exactly like one of the places where AI translation can’t do much damage to the literary value of the underlying texts
You might be surprised by the amount of damage it can do and the amount of human fixing that’s necessary to make the output entertaining to readers, who are not idiots simply because they enjoy genre fiction. But aside from that, turning these jobs into post-editing, a very different and horrible task, will definitely do damage to literary translators and literary translation as a whole. Good translation takes a lot of practice, people need to eat, and books that are of great literary value should not be screwed up by translators who have barely had the chance to translate anything outside of the classroom. Deskilling is a real thing. And you lose all of the amazing translators who can’t afford to translate just for personal pleasure. As it is, it’s very, very hard to make a living by translating full time; take away half the opportunities that are out there and young people will wisely avoid the profession even as a sideline.
Harlequin is also basically dying—they just shut down their whole historical line—and I guess they’re bent on accelerating the process.
“Harlequin is also basically dying—they just shut down their whole historical line—and I guess they’re bent on accelerating the process.” Yep.
In the Amapiano thread, Trond Engen said
He disappeared a long time ago, so I fear we won’t see him again.
Unless I’m confusing him with someone else he said something once that I found very interesting and which has coloured my notions about Danish pronunciation ever since. In relation to a comment that Danes can understand Swedish, but Swedes cannot understand Danish, just as Portuguese people can understand Spanish, but Spanish people can’t understand Portuguese, he said that the two cases were different, as Danes leave out all the consonants and Portuguese people leave out all the vowels. An exaggeration, of course, but so far as Portuguese is concerned it has a measure of truth, at least as perceived by Spanish speakers. When Portuguese really want to be understood they switch to Portuñol, which is Portuguese but pronounced in a Spanish sort of way. Do Danes have any special way of speaking when they want to be understood by Swedes?
There’s a similar asymmetrical mutual-comprehension thing going on between Agolle and Toende Kusaal, which are roughly as different as Spanish and Portuguese: Agolle speakers understand Toende better than vice versa (despite Agolle speakers being quite a bit more numerous.)
It’s probably because Agolle has gone farther in dropping word-internal consonants and in diphthongising original monophthongs.
Toende speakers say that their dialect is “purer.”
Martin Gardner wrote a story, “The Loves of Lady Coldpence,” about a pulp publisher that ran into trouble trying to produce fiction artificially. It was published in 1948, so no computers were involved, but the similarity to LLM-generated romance books is obvious in retrospect.
Am I right in thinking that Croat and Serbian are the only pair of languages for which there is an unambiguous 1:1 transliteration scheme that works with no exceptions in both directions, and is accepted by everyone? Curious, though, that Serbian rivals English and Dutch in having no diacritics, whereas Croat rivals Vietnamese in that respect.
English has a few nativized words like café, and Dutch has words like één, but I don’t think Serbian even has that many.
I quite often see criticisms at Wikipedia’s Articles for Deletion that claim that particular articles are LLM-generated: how does one recogize if something is LLM-generated?
I think David Marjanović upthread is right, “Médecins et célibataires” definitely sounds to me like a pun on “Médecins sans frontières”. Which shows why AI will aid but never wholly replace human translators: translating the above title accurately (i.e. creating the same sort of association of ideas to anglophone readers which the original does to francophone readers) would require creativity, something AI lacks (and considering how energy-consuming AI is right now, it is likely that any -hypothetical- creative post-AI system would be such a drain on energy and resources that human translators would ultimately prove cheaper).
An example of a translator’s creativity I have always liked is the English translation of “Astérix”, where Idéfix, the dog, was given the name “Dogmatix”, which to my mind is a BRILLIANT piece of translation. And one which I think no existing AI or translation software could have come up with.
Athel Cornish-Bowden: when it comes to Spanish speakers finding Portuguese hard to understand, Portuguese vowel deletion is just part of it. A feature of Portuguese historical phonology, making it unlike Spanish and indeed unlike other Romance national/major languages, is deletion of (Late Latin) intervocalic /l/ and /n/ (in the latter case yielding nasalization of the resulting vowel cluster/diphthong, sometimes yielding a new nasal consonant distinct from the etymologically expected one, i.e. /n/), yields words that are often shorter and more difficult to perceive as cognates as a result: Spanish VENIR and Portuguese VIR, Spanish TENER and Portuguese TER, Spanish LUNA and Portuguese LUA, Spanish MOLINO and Portuguese MOINHO, are all good examples of this.
Cacofonix is arguably an improvement on Assurancetourix, as I may have said elsewhere…
@Etienne: I suspect the series title “Médecins et célibataires” is a clever, maybe improved translation of “Harlequin Medical Romance”, the American title of that series. The volume I linked to earlier seems to be a translation of a book by Tina Beckett.
David E.: Médecins sans Frontières
French for “Doctors with no filter.” Bloody general surgeons, no notion of basic tact, the bastards.
Doctors without limits.
That sense works for Norw. Leger uten grenser as well. Or you can double it out replacing grenser with genser “sweater, top”.* Which brings us back to the scarlet medical romance:
All & Sundry: Médecins et célibataires must be a play on Médecins sans frontières.
Doctors with limits.
Legeromanen “The Doctor Novel/Romance” is a fixture of popular culture here as well. The cottage I frequented in my university years had a well equipped first aid kit. On the inside of the lid was a well-worn medical romance and the instructions never to use any of the equipment without having read the medical romance and signed the document. I remember a German hospital doctor who was the silent** and unapproachable center of imagination for all the nurses until our heroine came along and thought she had no chance with a man twice her age.
* There’s a local Christian organization called Misjon uten grenser “Mission without Borders”. Same effect, but stronger.
** Not completely silent. He did occasionally say things like “My wife shall not be working”, which made him even more irresistible.
@Athel Cornish-Bowden
>Do Danes have any special way of speaking when they want to be understood by Swedes?
Probably … but not a standardised one? In the 20th century, there were educational efforts and support materials from official sources. Like, eg, a booklet (‘At forstå hinanden i Norden’ or something similar), from the Nordic Council, I believe, which, in addition to some general tips (Danes: please, please use some kind of reading pronunciation), essentially contained a long list of potentially misleading expressions. But there were certainly more of this kind.
(And there was that time when numbers on Danish banknotes were written in pan-Nordic rather than normal Danish. Like ‘femti kroner’ instead of ‘halvtredsindstyve kroner’.)
Disregarding the Danish mumble, the only difference between Danish, Norwegian, and Swedish is, of course, that Swedish consists of a different subset of Low German vocabulary than the other two. 😉
Martin Gardner wrote a story, “The Loves of Lady Coldpence,” about a pulp publisher that ran into trouble trying to produce fiction artificially. It was published in 1948, so no computers were involved, but the similarity to LLM-generated romance books is obvious in retrospect.
cf. Roald Dahl’s The Great Automatic Grammatizator.
Thanks, that was delightful. And prescient:
And I enjoyed this bit of japery:
See also Antonio Machado’s contribution to this genre. Here’s the original (search for “Meneses”) and here’s an English translation, unfortunately missing a lot of paragraph breaks.
epexegetically
Oof! I was expecting something of the sort and was looking for it, but somehow missed this.
@Athel: I quite often see criticisms at Wikipedia’s Articles for Deletion that claim that particular articles are LLM-generated: how does one recogize if something is LLM-generated?
For school assignments, there’s software, which my colleagues and I are currently discussing.
Oh, it’s real. My colleage Fernandes is [fr̩ˈnãd̥͡ɕ].
Yes, except it’s Latin & Cyrillic within Serbian and Montenegrin. Both alphabets are in largely random use – they’re more like different font families than different alphabets; if you stand on a street in Belgrade or Niš and can’t read both, you’re illiterate.
And indeed, the closest thing to a diacritic the Cyrillic version has is the dot on the ј; the Latin version has a few, but a lot less than Vietnamese, especially no stacked ones.
>Do Danes have any special way of speaking when they want to be understood by Swedes?
Yes, it’s called “English” 😉
No, seriously. I think it’s a minority of Swedes who understand Danes better when we speak Danish than when we speak English. Probably vice versa too.
I do attempt to speak to kids from various countries in their languages once in a while, though. They’re usually not impressed.
The Portuguese hatred of the letter n was evident in a series of posters I saw in Lisbon in 1999 shortly before elections to the National Assembly, which carried the slogan Em boas mãos. Not an n to be seen, but three when it’s put into Spanish: En buenas manos. This also illustrates the retention of Latin o in words where Spanish changes it to ue.
Both alphabets are in largely random use
Not quite random. My impression from working in Serbia is that Latin is generally used in advertising for global and/or prestige brands and in most business situations where people are trying to convey some element of cosmopolitanism or international flair. It’s also used in art and theatre when there’s a more progressive bent. I think a lot of young people’s Serbian social media happens in Latin (I have no real evidence for this though). Cyrillic would be used, say, in a bakery to convey a homemade, local quality. Cyrillic is also associated with government and religious documents, and football hooligans, for better or worse.
The Portuguese hatred of the letter n
With a bit of creative accounting, you can reconstruct a stage of pre-proto-Oti-Volta with no *n at all.
Initial *n is in complementary distribution with *l, replacing the latter before nasal vowels; elsewhere, *n either replaces *l when it follows a nasal vowel, or is dissimilated from *m after a root-initial labial or labial-velar consonant (as in Kusaal bun “reap” or kpan “spear”: a surprising amount of this still applies in the modern Oti-Volta languages.)
It’s actually really weird (there are languages without any nasal consonants at all, but I don’t know of any that just lack /n/. Some Chinese topolects?)
Most likely, the original non-initial consonantal *n was replaced by vowel nasalisation at some early stage (in the Portuguese manner), and later on, as a separate thing, initial *l became *n before nasal vowels.
Yes, and with nationalism (including but not limited to religion, hooligans, pro-Russian sentiments, and much of government).
@Biscia: It’s not that I look forward to a time when human translation will be replaced by AI, but knowing how the world works, that’s what is going to happen. Not totally, probably, but like with craftsmen – high-end ones working for producers of luxury goods and amateurs doing it for fun. Quality will suffer, but unfortunately, most people prefer cheap over quality.
And children’s literature.
Also, most people would sign their name in Cyrillic, I think.
@Biscia: It’s not that I look forward to a time when human translation will be replaced by AI, but knowing how the world works, that’s what is going to happen.
Very likely, but I think what bothered Biscia was the prime message of your comment, which I reproduce for easy reference:
Purple romance (and similar pulp fiction) looks exactly like one of the places where AI translation can’t do much damage to the literary value of the underlying texts. The originals are so formulaic that they will soon be written by AI as well. More fingers to rip bodices, sling guns, and do surgery while romancing the pretty nurse.
You’re not so much saying “AI is inevitable” as “Romance is crap, so AI might as well write it.” I suspect that was written without much if any personal experience with the field, and I suspect Biscia is not the only one who would be bothered by it.
In the afterword to Season of Skulls, Charlie Stross expounds on the difficulties of writing good Regency Romance (and his consequent greatly increased appreciation of Georgette Heyer.)
As a science fiction fan who had to suffer much ignorant deprecation of his favorite form of reading, I am eager to defend all genres against such attacks.
Yeah, we’re talking Sturgeon’s Law here. Just because there are innumerable formulaic Harlequin Romance type things out there, it doesn’t mean the entire genre is something that intrinsically can’t be done well at all.
I mean, Jane Eyre hits all the beats. It really does. Denying it its place on the (steamy*) Romance shelf is really just like denying that Nineteen Eighty-Four is SF. It’s just a case of something done so well that it appeals even to those who are not lovers of the genre. But that doesn’t make it not-genre.
* Oh yes. The Young People of Today are just no good at subtext. Must we spell it all out for them?
Indeed, the best parts of Jane Eyre are just those that are most characteristic of the romance genre.
after a scarily long time as a hired-gun grantwriter, i feel very qualified to say that extremely formulaic genres are among the most difficult to write in well – and often ones where “well” is a question of concrete results rather than critical assessment*. having more structuring constraints means that details matter more, and slight missteps in tone, register, or relationship to the fashions of the field stand out much more clearly. woe betide the grantwriter who uses a buzzword from five years ago in an application to a foundation whose rhetoric has moved on! and also woe betide the grantwriter who fails to use that same buzzword in an application to a foundation whose leadership is only just catching up!
those are among the kinds of mistakes and mismatches that can’t, i think, be effectively avoided by statistical methods; i think Harlequin will not be happy with the results of their experiment.
.
* transparently true of grantwriting, of course, but i think the same is true for paraliterary genres, at least to the extent that they have/are their own economic spheres in which readers are voracious enough that economic success measures something more than the percentage of a Big Five literary imprint’s publicity budget devoted to a writer.
I’m used to Four: Elsevier, Springer Nature, Wiley, Informa. Does Brill count in the humanities?
Also, I haven’t read a lot of the genre, but as far as I know, most contemporary romance novels incorporate a heavy dose of erotica (ranging from soft-core to… much more explicit than one might think). And if there’s one thing that’s more difficult than writing a sex scene, it’s translating a sex scene, since it can be hard to translate descriptions of physical movements even when they’re not those physical movements. Characters have a way of turning into very ungainly puppets once you start talking about body parts.
https://tvtropes.org/pmwiki/pmwiki.php/Main/IKEAErotica
@DM-
I may be wrong, but I assumed rozele meant the five largest publishing houses in the US market (not that they are necessarily US companies) – Penguin Random House, HarperCollins, Simon & Schuster, Hachette Livre, and Macmillan Publishers.
That was my assumption as well.
yes, that was the Big Five i meant, whose market-cornering has been most thorough for literary fiction.
a heavy dose of erotica
yes, indeed! and the range of writing quality is, um, impressively wide. i still remember a dear friend’s mid-90s email handle, which came from a romance novel sex scene: “purple-headed warrior”. i think that she may have had that color hair at some point while using the handle, but that was very much not what was being talked about in the original context.
“purple-headed warrior”
Oh my. Although I have to say that’s outdone by many winners of the Bad Sex in Fiction award (which I greatly miss: https://literaryreview.co.uk/bad-sex-in-fiction-award), who are not in the romance genre and include some great authors. I feel for the translators occasionally involved, who must have found themselves staring at the screen for a long, long time. Sometimes all one can do is avoid making things worse.
The phrase “purple-headed warrior” was a meme by the early 1990s (meaning it circulated the old-fashioned way, by word of mouth). I first heard it in the summer of 1992, from other staff members at Camp Pioneer.
You’re not so much saying “AI is inevitable” as “Romance is crap, so AI might as well write it.” I suspect that was written without much if any personal experience with the field, and I suspect Biscia is not the only one who would be bothered by it
When I was a kid, I read basically everything with print on it. I was known at home for reading the ingredient lists on the tins and jars. So I also read my share of pulp magazines; I once got a big stack from a neighbor, which consisted mostly of crime, western, and horror stories, but also contained some romance stories. And my mum had many Georgette Heyer romances, which are at a higher level of quality, but still quite formulaic. So I’m not unfamiliar with the genre. My understanding is that Harlequin is a brand producing pulp romance, specifically of that formulaic kind that I know from my childhood forays into the genre. If that’s not true, I apologize. But if it’s true, then this is exactly the kind of formulaic stuff AI is already able to produce, and will become better at. To be clear, my remarks were directed at pulp Romance qua pulp, and to pulp writing in general, not at Romance as a genre or genre writing in general; I agree that there is good and artistically valuable genre writing.
Fair enough!