Poll: How do you feel about language companies integrating AI localization capabilities?

Rabie El Magdouli wrote:

CAT Tools, MT and AI… The only ones who benefit from this are the translation agencies that continue to squeeze the poor translator.

Johan Beyens wrote:

I think the translation industry is missing the mark by focusing too much on AI-driven translation, especially in MTPE workflows. It feels like we’re underutilizing both human expertise and AI potential. AI shines as a supporting tool, not a replacement. Tools like ChatGPT and Perplexity are very useful for tasks like analyzing source texts, extracting terminology, and spotting inconsistencies. These things make translation smoother and let translators focus on what they’re great at—actually translating.

But using AI to generate full translations often creates more work for humans, not less. Instead of using humans to fix AI’s mistakes, why not let AI handle the prep and QA (obviously under the supervision of a human expert), and leave the essence of the work to people?

Philip Lees wrote:

Here's a personal data point to fuel the discussion, though. Last week, on a whim, I asked ChatGPT to review a medical research abstract, fixing any language errors and making sure that the final text was no longer than 250 words. I also asked ChatGPT to comment on any deficiencies it found in the original text.

The result was scarily good. Its comments targeted the same things mine would have, and it found a couple of inconsistencies in the use of abbreviations that I had overlooked.

On the other hand - and this is important - it failed to correct one sentence where what the authors had written was not what they meant, as I was able to judge from my knowledge of the subject area and the type of text.

So the important thing here is that it shows the difference between a person (me) who understands the text, and a computer program that doesn't.

On the other hand, these things are starting to get so good that my opinion of them is changing. This was the free version of ChatGPT, and it's much better than it was when it started, a couple of years ago.

I'm now thinking that while its performance is still not good enough to satisfy me, a professional translator and editor, it is probably good enough to satisfy a lot of people. I do feel something about that, and that something is deep unease.

Ana Vozone wrote:

I have added a few ***** to illustrate the expectations/madness....

"We are reaching out to invite you to participate in a new project that we are launching in 10 days and running until April. Our primary objective is to assess the available supply for this project *******within a two-hour timeframe*******.

Task Description:
The task involves ******evaluating AI-generated responses****** based on a given context and query.

Annotators are required to:
* Language: Portuguese (European)
* *******Identify potential hallucinations (e.g., irrelevant, inaccurate, or fabricated content) at the sentence level********.
* Assess the overall relevance of the response.
* Answer predefined evaluation questions and, in some cases, provide brief comments to explain annotations.
* Average Time per Task: Approximately 5 minutes per task.

Action Required:
Please provide the necessary details in the following spreadsheet to confirm your participation.
Your CV

Confirm that you have done such project before

Price per task


If this project aligns with your expertise and interests, we’d be delighted to join us.

(In response, I told them my hourly rate saying that, if they accept it, I would provide the remaining information. Never heard from them again, of course.)

Daryo wrote:



Or translated in plain-speak ... you are given 5 minutes to "polish / finalise / weed out hallucinations in" what some unsuspecting client of some AI company will think was entirely the "brain-child/original work" of some AI software. The kind of people who are in awe of this new gadget and convinced that AI that could replace everyone tomorrow morning ...

I had few "offers" of that kind myself.