Humanity lost in traslation?

[XerXesXu]

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.

It's the errors in translation in the first three paras that threw the AI Detector. It's possible that without those low frequency errors they would have been flagged. I don't speak Italian, so I can't advise you on interpretation.

 

[devtek]

UPDATE:
I tried editing the text, cutting out parts that might have sounded AI-like (explanations that were too long, metaphors that were too predictable, etc.). Then I translated it myself, without worrying about making mistakes, and submitted it to the excellent editor Kenjisato, whom I thank for the speed of his review, as well as for the quality.
As a result, I submitted the revised text to several AI detectors, and the result was "a totally human text."
What drove me crazy was that the Italian text was always rated as 100% human, while the faithful English translation was contaminated by AI.
Thanks to Kenjisato and the other editors, this problem can be overcome and it can be a useful system for all non-native English speakers.
I'll finish the story now and try to submit it, hoping they haven't redflagged me...

 

[anthrodisiac]

UPDATE:
I tried editing the text, cutting out parts that might have sounded AI-like (explanations that were too long, metaphors that were too predictable, etc.). Then I translated it myself, without worrying about making mistakes, and submitted it to the excellent editor Kenjisato, whom I thank for the speed of his review, as well as for the quality.
As a result, I submitted the revised text to several AI detectors, and the result was "a totally human text."
What drove me crazy was that the Italian text was always rated as 100% human, while the faithful English translation was contaminated by AI.
Thanks to Kenjisato and the other editors, this problem can be overcome and it can be a useful system for all non-native English speakers.
I'll finish the story now and try to submit it, hoping they haven't redflagged me...
View attachment 2610443View attachment 2610444

Good luck! I hope you're able to get it through.

 

[devtek]

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.

A possible compromise solution might be to use Google Translate like old dictionaries. Look up one word at a time and then form the sentence without external or online help.

 

[Britva415]

The problem with any solution which proposes that the site "accept" non-AI stories in other languages before translation into English is the one we already have. Un-translated stories are already being rejected at what anecdotally (we don't have specific data) appears to be a higher rate than native-English-written stories.

The reason is: The site operators can't use their human judgement on the original text because they don't know the language. So there is no way for them to un-do a false-positive from an automated AI checker.

Making the site operators responsible for the translation wouldn't change that at all, when they already can't determine the provenance of the original.

 

[EmilyMiller]

The problem with any solution which proposes that the site "accept" non-AI stories in other languages before translation into English is the one we already have. Un-translated stories are already being rejected at what anecdotally (we don't have specific data) appears to be a higher rate than native-English-written stories.

The reason is: The site operators can't use their human judgement on the original text because they don't know the language. So there is no way for them to un-do a false-positive from an automated AI checker.

Making the site operators responsible for the translation wouldn't change that at all, when they already can't determine the provenance of the original.

Yup

 

[devtek]

The problem with any solution which proposes that the site "accept" non-AI stories in other languages before translation into English is the one we already have. Un-translated stories are already being rejected at what anecdotally (we don't have specific data) appears to be a higher rate than native-English-written stories.

The reason is: The site operators can't use their human judgement on the original text because they don't know the language. So there is no way for them to un-do a false-positive from an automated AI checker.

Making the site operators responsible for the translation wouldn't change that at all, when they already can't determine the provenance of the original.

In my case, however, there were no false positives. I had my Italian text checked by several AI detectors, and they all confirmed it was 100% human. I also posted the screenshot.

 

[Britva415]

In my case, however, there were no false positives.

Did you submit it to Lit? Did Lit approve and publish it?

Or was it your English translation which got rejected?

I'm just saying: Lit can have a false positive no matter what your own checking shows. They don't use the same tools as you to test for AI. And they won't take your screenshot as evidence.

 

[devtek]

Did you submit it to Lit? Did Lit approve and publish it?

Or was it your English translation which got rejected?

I'm just saying: Lit can have a false positive no matter what your own checking shows. They don't use the same tools as you to test for AI. And they won't take your screenshot as evidence.

We're not in court, I don't think "evidence" is needed. I explained what happened to me at the beginning of the thread, and I take the blame.
The screenshot was just to highlight the problem and ask for clarification, which the other contributors later provided, and I thank them.

 

[Britva415]

We're not in court, I don't think "evidence" is needed. I explained what happened to me at the beginning of the thread, and I take the blame.
The screenshot was just to highlight the problem and ask for clarification, which the other contributors later provided, and I thank them.

I was just surprised you didn't consider it a false positive, that's all.

 

[devtek]

I was just surprised you didn't consider it a false positive, that's all.

There were no false positives.
The story translated with Grammarly was rightfully sent back because the algorithm detected that the translation had been done with an invasive, prohibited tool. Many authors, non-native English speakers, translated their stories with Google Translate and even wrote it in their stories, as I pointed out in #25, and this led me to make the mistake. Obviously, I'm not making excuses; I should have paid more attention to the instructions.
I submitted the story in Italian anyway, and it's currently pending.

In #27, I recounted what happened next (thanks, Kenjisato), but only in the hope of being helpful to other non-native English-speaking authors.
Translate yourself, without fear of errors, and get help from an editor.
This should hopefully eliminate any misunderstandings about AI.