Pages in topic:   < [1 2 3] >
Rates for ChatGPT/Bard Quality Assurance
Thread poster: Jo Macdonald
Jo Macdonald
Jo Macdonald  Identity Verified
Spain
Local time: 10:27
Italian to English
+ ...
TOPIC STARTER
Err..no Mar 31, 2023

Thanks for your comment Lieven but cyber sniping gets us nowhere.
Same standards as you I imagine.


 
Lieven Malaise
Lieven Malaise
Belgium
Local time: 10:27
Member (2020)
French to Dutch
+ ...
Sorry. Mar 31, 2023

Jo Macdonald wrote:

Thanks for your comment Lieven but cyber sniping gets us nowhere.
Same standards as you I imagine.



I didn't mean to offend you. I am just trying to understand why you should do fact-checking if you have a source text available. It just doesn't make sense to me that a convincing looking AI translation could trick you into not correcting errors after you have read the source.


 
Giovanni Guarnieri MITI, MIL
Giovanni Guarnieri MITI, MIL  Identity Verified
United Kingdom
Local time: 09:27
Member (2004)
English to Italian
Don't get it... Mar 31, 2023

if ChatGPT is fed a source text, why would it make things up? I did feed ChatGPT some translations to check the quality some time ago, and it returned a faithful translation, but not as good as DeepL. Do you think that ChatGPT would make things up when translating from a source text?

 
Mr. Satan (X)
Mr. Satan (X)
English to Indonesian
Prompt Apr 4, 2023

Giovanni Guarnieri MITI, MIL wrote:
if ChatGPT is fed a source text, why would it make things up? I did feed ChatGPT some translations to check the quality some time ago, and it returned a faithful translation, but not as good as DeepL. Do you think that ChatGPT would make things up when translating from a source text?


The user may enter a prompt that causes ChatGPT to alter the source texts. After all, a tool is only as good as the person utilizing it. I’ve also heard concerns about ChatGPT tailoring its answers based on each individual’s profile (political view, religious belief, brand loyalty, average income, medical records, geographical location, etc). It is unfortunate that ChatGPT is a proprietary software. Had the entire project was open-sourced, we would’ve been able to study it to fully comprehend its nature, the most strategic deployment scenarios for such technology, and the limits of its capabilities.


 
Giovanni Guarnieri MITI, MIL
Giovanni Guarnieri MITI, MIL  Identity Verified
United Kingdom
Local time: 09:27
Member (2004)
English to Italian
Not sure why... Apr 4, 2023

Mr. Satan wrote:

The user may enter a prompt that causes ChatGPT to alter the source texts.


Why would the user do that in a translation?


 
Mr. Satan (X)
Mr. Satan (X)
English to Indonesian
The User Apr 4, 2023

Giovanni Guarnieri MITI, MIL wrote:
Why would the user do that in a translation?


I’ve conversed with enough programmers and sysadmins to understand their woes.



 
Giovanni Guarnieri MITI, MIL
Giovanni Guarnieri MITI, MIL  Identity Verified
United Kingdom
Local time: 09:27
Member (2004)
English to Italian
Funny... Apr 4, 2023

Mr. Satan wrote:

Giovanni Guarnieri MITI, MIL wrote:
Why would the user do that in a translation?


I’ve conversed with enough programmers and sysadmins to understand their woes.





Mr. Satan (X)
 
Mr. Satan (X)
Mr. Satan (X)
English to Indonesian
New Age of MTPE Apr 20, 2023

So it appears that these types of jobs have started to populate in my language pair. Most of the outsourcers seem to treat them the same as MTPE assignments. That slightly concerns me. As I've stated before, the output of ChatGPT and similar software depends a lot on the users’ input. And I don’t know how good the prompter was in formulating their commands. It’s arguably prudent to ask for a preview of the texts before agreeing to anything. You may be in for a nasty surprise.

... See more
So it appears that these types of jobs have started to populate in my language pair. Most of the outsourcers seem to treat them the same as MTPE assignments. That slightly concerns me. As I've stated before, the output of ChatGPT and similar software depends a lot on the users’ input. And I don’t know how good the prompter was in formulating their commands. It’s arguably prudent to ask for a preview of the texts before agreeing to anything. You may be in for a nasty surprise.

It’s like back when I was still accepting MTPE requests. I was unlucky enough to receive translated documents that were blatantly churned out from an untrained MT engine. The results were so awful that you’d have to tear them down and redo the whole thing from scratch. Now, this does not mean I can justify the techno-phobia that some echoed throughout this website. Had LLM AI like ChatGPT or NMT like DeepL never existed, bad agencies/clients would still try to find ways to exploit the linguists. The problem here is not the technology, but the meatbags who abuse these tools and not giving a single damn to even employ them properly.

[Edited at 2023-04-20 06:44 GMT]
Collapse


Jo Macdonald
 
Jo Macdonald
Jo Macdonald  Identity Verified
Spain
Local time: 10:27
Italian to English
+ ...
TOPIC STARTER
Totally agree Apr 20, 2023

The problem isn't a tool, which might be more or less suited to a job, it's the people selling these things to clients as "acceptable" in need of almost no work, for a cheapo rate.

Mr. Satan wrote:

The problem here is not the technology, but the meatbags who abuse these tools and not giving a single damn to even employ them properly.

[Edited at 2023-04-20 06:44 GMT]


Mr. Satan (X)
Michele Fauble
 
Mr. Satan (X)
Mr. Satan (X)
English to Indonesian
What if… Apr 25, 2023

…the original author also used ChatGPT to write the source texts?

[mind blown]

Lieven Malaise wrote:
I am just trying to understand why you should do fact-checking if you have a source text available. It just doesn't make sense to me that a convincing looking AI translation could trick you into not correcting errors after you have read the source.


[Edited at 2023-04-25 04:24 GMT]


 
Lieven Malaise
Lieven Malaise
Belgium
Local time: 10:27
Member (2020)
French to Dutch
+ ...
His/her problem ... Apr 25, 2023

Mr. Satan wrote:

…the original author also used ChatGPT to write the source texts?

[mind blown]


... not mine.


 
Mr. Satan (X)
Mr. Satan (X)
English to Indonesian
Problem Apr 25, 2023

Lieven Malaise wrote:
His/her problem[,] not mine.


But you post-edit their piece of work after it was translated by ChatGPT. Then it becomes YOUR problem.


 
Lieven Malaise
Lieven Malaise
Belgium
Local time: 10:27
Member (2020)
French to Dutch
+ ...
Whut ? Apr 25, 2023

Mr. Satan wrote:

But you post-edit their piece of work after it was translated by ChatGPT. Then it becomes YOUR problem.


I don't know what you mean. If a client provides me with a translation that I am supposed to (post-)edit, be it through ChatGPT or DeepL or whatever, he also provides me with the source. So from my point of view the source is the only document that counts. If it's full of factual mistakes, then that is my client's responsibility, not mine. My only job is to make sure the source and the translation match.

[Edited at 2023-04-25 09:27 GMT]


 
Kay Denney
Kay Denney  Identity Verified
France
Local time: 10:27
French to English
. Apr 25, 2023

Lieven Malaise wrote:
I am just trying to understand why you should do fact-checking if you have a source text available.


I don't know about the texts you translate, but for me, fact-checking is part of the service I offer.
Very often, I will find errors in the source text. These might just be spellling mistakes (often foreign names) but sometimes terms are used wrongly or some other fact might be represented inaccurately.

Just the other day I had to translate a brochure about places of interest in a town in Burgundy. There were spelling mistakes, and a Duke of Burgundy was wrongly attributed another title. Now I check the spelling of all names I'm not 100% familiar with as a matter of course. It also happens that I studied that place and time in history in great detail at school, so I spotted the factual errors too. I checked them, after all I was at school 40 years ago and new information may have surfaced since then. Since the Internet was pretty well unanimously in agreement with Miss Pearson my dear history teacher, I corrected the text in my translation and left a note with links to flag the details to be corrected in the source text.

Marketing materials need to be flawless. The translator examines the source text in detail, it's perfectly normal to spot errors, and I think it's also perfectly normal to point them out to the client, so they don't get egg on their face over a howler.


Mr. Satan (X)
Jocelyne Cuenin
 
Mr. Satan (X)
Mr. Satan (X)
English to Indonesian
Localization Apr 25, 2023

Lieven Malaise wrote:
I don't know what you mean. If a client provides me with a translation that I am supposed to (post-)edit, be it through ChatGPT or DeepL or whatever, he also provides me with the source. So from my point of view the source is the only document that counts. If it's full of factual mistakes, then that is my client's responsibility, not mine. My only job is to make sure the source and the translation match.


Guess this is where we differ. I’m a localizer, so it’s my duty to make sure the translated texts are culturally appropriate, which may require fact-checking.

[Edited at 2023-04-25 11:33 GMT]


 
Pages in topic:   < [1 2 3] >


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

Rates for ChatGPT/Bard Quality Assurance







CafeTran Espresso
You've never met a CAT tool this clever!

Translate faster & easier, using a sophisticated CAT tool built by a translator / developer. Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools. Download and start using CafeTran Espresso -- for free

Buy now! »
Anycount & Translation Office 3000
Translation Office 3000

Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.

More info »