Humanity lost in traslation?

devtek

Slow-in-law
Joined
Feb 17, 2013
Posts
28
I'm Italian, and I've been reading stories on Literotica since 2014, especially the Loving Wives section.
I've written many stories, in Italian, and I decided to try publishing some here.
After uploading them, they were rejected for suspected AI.
Okay, I thought, I used Grammarly to try to write in the best possible English, and it's possible that this caused problems.
I rewrote the story, changing some passages, and submitted it, in Italian, to ZeroGPT's AI verification. The result: the text is 100% human.
Since I can't translate properly myself, nor do I want to pay a professional translator, I have Google Translate translate the text, which should be less invasive.
The result: the exact same text, but in English, with a 34% AI score.
Another rejection would be frustrating. Do you have any suggestions?

ScreenHunter 2203.jpgScreenHunter 2204.jpg
 
All the major translation apps and websites use AI, and that's against the rules here. Sorry. You could try submitting your story in the original Italian, but I might as well warn you about two problems with that:
  1. You might have the same problem. The AI detector seems to have a lot of false positives for stories in languages that aren't English.
  2. Italian doesn't seem like a popular language here. I see only 39 Loving Wives stories here in that language total.
I don't want to keep anyone out, I like the fact that Literotica has an international community of readers and writers, but the site should really offer more and better support for the non-English side or be honest about how little it is.
 
This won’t be helpful to you necessarily but I wonder if it might be in the interests of this website to accept submissions in their original language, put them through whatever AI tests they find satisfactory and then, if they believe the text to be human, put it through their own preferred translation tool in order to then publish in English.

Just an idea. More work of course. But enables more sharing of stories. Which is the whole point of this site, no?

Personally, I’d be interested in more erotica from people who didn’t grow up in the English language world.
 
This won’t be helpful to you necessarily but I wonder if it might be in the interests of this website to accept submissions in their original language, put them through whatever AI tests they find satisfactory and then, if they believe the text to be human, put it through their own preferred translation tool in order to then publish in English.

Just an idea. More work of course. But enables more sharing of stories. Which is the whole point of this site, no?

Personally, I’d be interested in more erotica from people who didn’t grow up in the English language world.
I like your suggestion. I just tried submitting my story in Italian, using the text turned out to be 100% human.
If it's accepted, I'll try submitting it in English, pointing out that it's exactly the same story.
If the Italian one is accepted, but the English one is rejected, I'll still have made a step forward in expanding LIT to Italians.
 
Which would result in AI-written text, sadly.

Human translation is the only option.
Sure. But the point is that it would be the site’s own translation - insofar as they’d be completely confident in the provenance of the translation. And confident that the original text was not AI too.

The ideal of course is that there is a small army of human translators performing the service. I’m not suggesting AI translation is equivalent.

But perhaps in this scenario it’s a worthwhile compromise.

(Though I suppose there’s nothing stopping readers from copying and pasting an Italian text from here into an online translation tool anyway).

Anyway, just an idea. My other idea is we all learn to read Italian…
 
Sure. But the point is that it would be the site’s own translation - insofar as they’d be completely confident in the provenance of the translation. And confident that the original text was not AI too.
The translated text would be AI-generated (all translation tools now use LLMs sadly) and thus in breech of the site’s own rules. More pertinently it would expose the site to potential legal liability if there is a general finding for copyright theft against the genAI companies. Your solution isn’t a solution.
 
Sure. But the point is that it would be the site’s own translation - insofar as they’d be completely confident in the provenance of the translation. And confident that the original text was not AI too.

The ideal of course is that there is a small army of human translators performing the service. I’m not suggesting AI translation is equivalent.

But perhaps in this scenario it’s a worthwhile compromise.

(Though I suppose there’s nothing stopping readers from copying and pasting an Italian text from here into an online translation tool anyway).

Anyway, just an idea. My other idea is we all learn to read Italian…

This is still a problem, by the Site's own logic, which I think is conceptually sound despite the obvious difficulties of administering it.

The Site wants its readers to enjoy human-generated stories, not AI stories. An AI translation into English of a foreign language story, even if done internally, is still an AI-generated story, and the Site violates its own announced concept of integrity by allowing those stories to be published alongside real, human-generated English stories.

Seems like publishing in Italian is the right way to go here. It might not get as many reads, but my guess is it will get more than if published at another site.
 
Yes, that’s understood. Which is why I describe it as a compromise on their own rules.
Which doesn’t address the legal liability point, which I believe is at least coequal to the site’s principles. The legal peril is real, and porn sites are already beset by alligators.
 
Which doesn’t address the legal liability point, which I believe is at least coequal to the site’s principles. The legal peril is real, and porn sites are already beset by alligators.
You mean, if the site were to use a translation tool that was developed on stolen copyrighted material then the site would open itself up to some kind of liability?
 
The translated text would be AI-generated (all translation tools now use LLMs sadly) and thus in breech of the site’s own rules. More pertinently it would expose the site to potential legal liability if there is a general finding for copyright theft against the genAI companies. Your solution isn’t a solution.
I don't know how much of a problem this actually is if you translate with basic tools. Google Translate doesn't claim any rights to what's translated. And many stories by non-English authors clearly state that they were translated with Google Translate.
A good solution might be to try translating my stories myself, with all the errors that might be present, and then ask for help from one of the volunteer proofreaders on LIT.
 
Anyway, what makes the verification software think that the sentence "When friends joked with Alessandro, they said he was the best male lawyer in Milan and the second-best lawyer in his own home" (which was entirely thought up by me) could be flagged as AI?
That's what drives me crazy.
 
I don't know how much of a problem this actually is if you translate with basic tools. Google Translate doesn't claim any rights to what's translated. And many stories by non-English authors clearly state that they were translated with Google Translate.
A good solution might be to try translating my stories myself, with all the errors that might be present, and then ask for help from one of the volunteer proofreaders on LIT.
The issue is any translation tool, Google Translate included, will introduce text sourced from LLMs. Such text has been stolen from its copyright holders and many of them are sueing to be made whole. And it has been established that the genAI companies are lying wholesale about how they use copyrighted material.

Knowingly publishing such text potentially exposes Literotica to legal peril.

I use GT to check my fumbling French or German. The odd sentence isn’t the issue. The issue is a software-based translation from - say - a French novel to English will result in a text laden with AI slop. And one for which the publishers could potently be sued under breech of copyright.
 
Anyway, what makes the verification software think that the sentence "When friends joked with Alessandro, they said he was the best male lawyer in Milan and the second-best lawyer in his own home" (which was entirely thought up by me) could be flagged as AI?
That's what drives me crazy.
The original Italian is so clear that it can hardly be translated any other way. (I'm no expert in Italian, so would have to check scherzare in a dictionary for its nuances, but otherwise even I can translate that sentence at sight.) When 'friends' or 'his friends', in 'his house' or 'his home' or 'his own house' - any AI translator and any human translator would have to come to almost exactly the same wording, so I don't know how it could possibly be AI-like.

One tiny thing. You said 'joked with' just now, but 'wanted to joke and tease' is what you tested earlier. 'Joke' can't take a direct object, so 'joke and tease Alessandro' is technically incorrect - so one might hope AI wouldn't choose that, whereas you, skilled in English but not perfect, might.
 
The original Italian is so clear that it can hardly be translated any other way. (I'm no expert in Italian, so would have to check scherzare in a dictionary for its nuances, but otherwise even I can translate that sentence at sight.) When 'friends' or 'his friends', in 'his house' or 'his home' or 'his own house' - any AI translator and any human translator would have to come to almost exactly the same wording, so I don't know how it could possibly be AI-like.

One tiny thing. You said 'joked with' just now, but 'wanted to joke and tease' is what you tested earlier. 'Joke' can't take a direct object, so 'joke and tease Alessandro' is technically incorrect - so one might hope AI wouldn't choose that, whereas you, skilled in English but not perfect, might.
I tried asking ChatGPT why this phrase is considered AI. It replied that the joke is too good to be human. :oops:
 
Of course it is! No mere human could come up with good jokes, or use consistently correct grammar or punctuation. Worth reading? Must be an AI wot wrote it. They're after our jobs.
 
Of course it is! No mere human could come up with good jokes, or use consistently correct grammar or punctuation. Worth reading? Must be an AI wot wrote it. They're after our jobs.
It’s important realize (and I’m sure you do, this is more of a general comment) that the only thing that genAI is about, has ever been about, and will ever be about in the future, is making a handful of already obscenely rich men even richer. It’s not a tool to aid humanity, it’s not a staging post on the way to AGI. It’s a money-making Ponzi scheme and way too many people have bought what the snake oil salesmen are selling.
 
devtek, in #1 you posted your Italian text and an English translation. I'm not clear about this. Is the English translation exactly what came out of the translation program, or did you adjust it?

Because if it's the translation program, it's done a lot of 'cleaning up' and adjustment that it didn't need to do. Here are three examples of the Italian, the English you quoted above, and an equally good but more literal English translation:

dei migliori avvocati di Milano, nel campo del diritto d'impresa
one of Milan's top corporate lawyers
(one of Milan's best lawyers in the field of corporate law)

I suoi onorari erano
He was now earning
(His earnings were)

quando era studente universario
while studying at university
(when he was a university student)

You obviously know English well. You can see that the more literal translations (mine in brackets) would be easier to do for a program that simply did translation.

Then there are parts that just aren't in the English translation:

più giovane di lui di due anni
(two years younger than him)

tutti i colleghi la ammiravano e rispettavano
(all her colleagues admired and respected her)

If you adjusted the English text and chose to do this, that's fine. But if it's an automatic (AI) translator making these quite big changes, that might be a problem. These are bigger changes than the sentence I analysed in my earlier comment.
 
devtek, in #1 you posted your Italian text and an English translation. I'm not clear about this. Is the English translation exactly what came out of the translation program, or did you adjust it?

Because if it's the translation program, it's done a lot of 'cleaning up' and adjustment that it didn't need to do. Here are three examples of the Italian, the English you quoted above, and an equally good but more literal English translation:

dei migliori avvocati di Milano, nel campo del diritto d'impresa
one of Milan's top corporate lawyers
(one of Milan's best lawyers in the field of corporate law)

I suoi onorari erano
He was now earning
(His earnings were)

quando era studente universario
while studying at university
(when he was a university student)

You obviously know English well. You can see that the more literal translations (mine in brackets) would be easier to do for a program that simply did translation.

Then there are parts that just aren't in the English translation:

più giovane di lui di due anni
(two years younger than him)

tutti i colleghi la ammiravano e rispettavano
(all her colleagues admired and respected her)

If you adjusted the English text and chose to do this, that's fine. But if it's an automatic (AI) translator making these quite big changes, that might be a problem. These are bigger changes than the sentence I analysed in my earlier comment.
You're right. I posted a screenshot of a translation I edited, following the suggestions for lowering the AI index.
The point is that some of the "incriminating" sentences, highlighted in yellow, were entirely created by me, and the translation, as you correctly said before, is very faithful to the original.
 
Right, that's fine then. And yes, that final sentence in yellow - it's a direct translation and it's ridiculous that it could be seen as AI.
 
I don't know how much of a problem this actually is if you translate with basic tools. Google Translate doesn't claim any rights to what's translated.

If I copy somebody else's story and post it online with a big disclaimer saying "I don't claim any rights to this story", I'm still in breach of their copyright.

And if I use a LLM to generate content, depending on how much I lean on the LLM, there's a possibility that nobody owns the copyright on the resulting work.

And many stories by non-English authors clearly state that they were translated with Google Translate.

Recent ones? Google Translate has been around for a long time, but as I understand it the technology under the hood has changed.
 
Back
Top