Poll: Can machine-translating English training data produce culturally sound AI? இழை இடுபவர்: ProZ.com Staff
|
|---|
This forum topic is for the discussion of the poll question "Can machine-translating English training data produce culturally sound AI?".
This poll was originally submitted by Ana Moirano. View the poll results »
| | | |
BBC has at least one TV soap opera ("Eastenders") that has been aired for several decades, I wonder if Proz is trying to follow in the same footsteps?
| | | | Zea_Mays 
இத்தாலி
Local time: 15:53
ஆங்கிலம் - ஜெர்மன்
+ ...
| | When in doubt, ask AI? Hmmm... | Oct 19 |
Having read and re-read the question - and then read it again just to be sure I'd not understood it - I asked Gemini what it means.
It replied:
Error: 503{
"error": {
"code": 503,
"message": "This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.",
"status": "UNAVAILABLE"
}
}
It seems Gemini hasn't any more time to waste on this than I have myself.
| | |
|
|
|
Frankly, this is one of the reasons why I visit ProZ.com less and less.
When even the questions posted on a platform for language professionals are poorly worded and ambiguous, it does little to inspire confidence in the quality of the forum.
| | | |
As the question relates to AI, let’s find out what the AI overview has to say on the subject:
"No, machine-translating English training data generally cannot produce culturally sound AI. While machine translation (MT) preserves basic semantic meaning, it fails to capture deep cultural nuances, local contexts, and unique societal values, often resulting in "cultural translationese" or Western-centric biases.
🎭 Limitations of Machine-Translated Training Data
... See more As the question relates to AI, let’s find out what the AI overview has to say on the subject:
"No, machine-translating English training data generally cannot produce culturally sound AI. While machine translation (MT) preserves basic semantic meaning, it fails to capture deep cultural nuances, local contexts, and unique societal values, often resulting in "cultural translationese" or Western-centric biases.
🎭 Limitations of Machine-Translated Training Data
• Loss of Idioms and Cultural Context: Machine translation engines frequently struggle with localized idioms, humor, and metaphors. This can lead to literal, unnatural translations that miss the intended emotional or social weight.
• Amplification of Western/Anglocentric Bias: Simply translating English data means the AI still inherits Western norms, legal concepts, historical perspectives, and societal benchmarks. It fails to reflect the organic worldview of the target language's culture.
• Grammatical and Social Mismatch: Languages with complex honorifics, gender dynamics, or relationship-based pronouns (like Japanese, Korean, or Arabic) cannot be adequately accurately mapped from English without native, culturally grounded contexts.
• Superficial Alignment: The AI may learn to speak the target language grammatically, but its underlying logic, ethical values, and behavioral expectations will remain fundamentally American or English-centric." ▲ Collapse
| | | | Thayenga
ஜெர்மணி
Local time: 15:53
உறுப்பினர் (2009)
ஆங்கிலம் - ஜெர்மன்
+ ...
| I have only one question. | Oct 19 |
When will these silly AI polls ever meet their very untimely fate? A fate that seems to be long overdue.
Maybe I should develop some kind of AI that will automatically delete all AI related polls.
[Edited at 2026-10-19 16:03 GMT]
| | | | Peter Motte
பெல்ஜியம்
Local time: 15:53
உறுப்பினர் (2009)
டச் - ஃபிளமிஷ்
+ ...
Thayenga wrote:
When will these silly AI polls ever meet their very untimely fate? A fate that seems to be long overdue.
Maybe I should develop some kind of AI that will automatically delete all AI related polls.
[Edited at 2026-10-19 16:03 GMT]
Financially AI is in a very bad shape. Once they bust, those AI things will mostly disappear, and what's left will be expensive.
The first one has fallen: Character .ai Others will follow. It's not a question IF, but WHEN. I hope soon enough
| | |
|
|
|
expressisverbis wrote:
When even the questions posted on a platform for language professionals are poorly worded and ambiguous, it does little to inspire confidence in the quality of the forum.
You're too polite. They can't even ask simple questions right!
ProZ has become such an embarrassment. I'm now actually thinking of changing my username to an anonymous one (after 20+ years) as I'm not exactly proud of being here. But that's just me.
[Maybe there should be a forum solely for interaction between AI bots (gemini, chatgpt, perplexity, grok, claude etc.) without human intervention, but it shouldn't be here, on a site for HUMAN language professionals.]
Happy Tuesday!
| | | |
Even if it's gibberish, at last we have a question with a definitive Yes/No answer.
Which is:
October 2026 - No
October 2028* - Yes
(*If not sooner)
Next.
| | | |
Maria Teresa Borges de Almeida wrote:
BBC has at least one TV soap opera ("Eastenders") that has been aired for several decades, I wonder if Proz is trying to follow in the same footsteps?
In the US, the Star Trek franchise has just celebrated its 60th anniversary.
| | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Poll: Can machine-translating English training data produce culturally sound AI? | Trados Studio 2022 Freelance | The leading translation software used by over 270,000 translators.
Designed with your feedback in mind, Trados Studio 2022 delivers an unrivalled, powerful desktop
and cloud solution, empowering you to work in the most efficient and cost-effective way.
More info » |
| | Wordfast Pro | Translation Memory Software for Any Platform
Exclusive discount for ProZ.com users!
Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
