| Pages in topic: < [1 2 3] | ChatGPT as a potential tool for translators Thread poster: James McVay
|
|---|
Lingua 5B 
Bosnia and Herzegovina
Local time: 17:12
Member (2009)
English to Croatian
+ ...
Ice Scream wrote: Lingua 5B wrote:
A diner is a small inexpensive restaurant in the US. This is news to someone???
Jo thought I mistyped it for dinner, I only referred to that part. But none of it was the point here. “English students” is one of the most commonly cited examples of semantic ambiguity, Internet is full of it, so I’m surprised GPT hadn’t managed to harvest it. I asked it a language/lingustic question on purpose, because someone in the thread said it was good/best with the language. Not.
| | | | Jo Macdonald
Spain
Local time: 17:12
Italian to English
+ ...
| Nah, didn't think you'd mistyped it | Feb 19, 2023 |
I often see clients using English with funky spelling in their Italian text for example.
You asked the chatbot if there was anything potentially ambiguous in that phrase and it didn't spot commonly confused words that could change the meaning of the phrase.
Lingua 5B wrote
Jo thought I mistyped it for dinner,
| | | |
Lingua 5B wrote:
Me: Is there any word or phrase that’s ambiguous in this sentence: A group of English students were eating at a diner.
GPT: There is no ambiguous word or phrase in the sentence "A group of English students were eating at a diner" as it is written. The sentence is clear and straightforward in its meaning, and there are no words or phrases that have multiple possible interpretations or could be misunderstood.
-
Actually, “English students” can be either English majors or students coming from England.
This illustrates why good translators have nothing to fear from ChatGPT and its ilk.
The AI doesn't see any ambiguity in the sentence because it only parses it one way. When I asked it to translate the sentence into Greek, it rendered the phrase as "English-speaking students"
When I changed the "source" sentence to "A group of English and Philosophy students were eating at a diner.", the result was "A group of English speakers and Philosophy students ..."
Obviously ChatGPT is strongly biased towards the use of English to indicate a language, rather than a nationality. That makes sense, given that the number of English speakers in the world is far greater than the number of English people, so the material on which the AI bases its responses is likely to have a similar bias.
With "A group of Philosophy and English students were eating at a diner.", ChatGTP got it right (at last!), by indicating that the students were studying the English language, rather than just speaking it.
ChatGTP is only as good as the data it was trained on - and often not even that good. I have found it useful for tracking down abbreviations and other special terminology. Others may not.
But to see it - in its present form - as some kind of potential professional competitor, or even as a threat, is just silly.
| | | |
Lingua 5B wrote:
I asked it a language/lingustic question on purpose, because someone in the thread said it was good/best with the language. Not.
If by "someone in the thread" you mean me, that is not what I said. Please go back and reread my post and please do not misrepresent me again.
Thank you.
| | |
|
|
|
| It obviously doesn't have a clue about snooker :-) | Feb 20, 2023 |
Philip Lees wrote: Cilian O'Tuama wrote:
It answered 147, clearly wrong. This is interesting, because if we consider how ChatGPT works, it's quite easy to see how it gets this kind of question wrong. *Technically, it is possible to score more, but only under very rare and special circumstances. Most people would consider 147 to be the correct answer.
147 (or 155) correct, but with blacks.
I rephrased my Q:
If I pot all 15 reds, each followed by pink, and then sink yellow, green, brown, blue and pink, what do I score?
...you will score a total of 147 points.
Here's how the scoring works:
15 reds with 15 pinks = 120 points
Yellow = 2 points
Green = 3 points
Brown = 4 points
Blue = 5 points
Pink = 6 points
So, the total score for a break consisting of 15 reds followed by pinks, and then yellow, green, brown, blue, and pink is 120 + 2 + 3 + 4 + 5 + 6 = 140 points.
However, in a standard game of snooker, the pink ball is always followed by the black ball, not the colors in sequence. Therefore, the maximum score in a standard game is achieved by potting all 15 reds with 15 blacks (worth 120 points) and then the six colors in sequence, with the black ball being the last ball potted for 8 points, for a total of 147 points, which is also known as a maximum break.
Am eagerly awaiting other massive shortcomings 
Maybe it is better with languages than with facts and common sense.
| | | |
Cilian O'Tuama wrote:
Maybe it is better with languages than with facts and common sense.
Sigh. ChatGPT (or any other AI) does not have "common sense" and "facts" depend entirely on the material it's trained on. How many times does this need to be said?
ChatGPT is a "large language model". It takes a piece of language and then manipulates it according to complex algorithms and with reference to billions of words of data. It then generates a response that incorporates some degree of randomness. It doesn't "understand" anything; it doesn't have any kind of sense, no matter how common; and it doesn't feel anything.
Expecting ChatGPT to show common sense is like expecting a fruit machine* to show common sense when you pull its handle.
*I added the picture because I see that the default meaning of fruit machine, according to Wikipedia, is now a device for identifying gay men, and I would hate to cause any confusion or ambiguity.
| | | | toasty
Italy
Local time: 17:12
Member (2013)
Italian to English
| another issue with MT and ChatGPT | Feb 21, 2023 |
There's a lot of talk about the quality of ChatGPT's translation and revision abilities, but what about confidentiality? Sure, ok students might try using it to cheat on their homework, but what about concerns like copyright and confidentiality/GDPR?
I'm currently working on the translation of a book that has yet to be published. I clearly can't be uploading confidential, proprietary text willy-nilly to a platform that does who knows what exactly with the text that's entered.
... See more There's a lot of talk about the quality of ChatGPT's translation and revision abilities, but what about confidentiality? Sure, ok students might try using it to cheat on their homework, but what about concerns like copyright and confidentiality/GDPR?
I'm currently working on the translation of a book that has yet to be published. I clearly can't be uploading confidential, proprietary text willy-nilly to a platform that does who knows what exactly with the text that's entered.
Moreover, how can I know that what it's spitting out hasn't essentially been plagiarized from some other source?
https://www.forbes.com/sites/lanceeliot/2023/01/27/generative-ai-chatgpt-can-disturbingly-gobble-up-your-private-and-confidential-data-forewarns-ai-ethics-and-ai-law/
Has anyone here updated their confidentiality agreements with their clients to cover these matters? ▲ Collapse
| | | | Tom in London
United Kingdom
Local time: 16:12
Member (2008)
Italian to English
Anton Konashenok wrote:
Last week, a friend of mine had the following conversation with ChatGPT. It was actually in Russian, translation into English is mine......(cont.)
That is absolutely hilarious! On this occasion, LOL actually does mean what I was doing!
Post some more!
[Edited at 2023-02-21 13:41 GMT]
| | |
|
|
|
James McVay
United States
Local time: 11:12
Russian to English
+ ...
TOPIC STARTER | Replying to Sinan | Feb 27, 2023 |
Sinan Alobaidi wrote:
I don't think ChatGPT is a better translator than other MTs. However, it has the ability to edit texts, an ability that MTs lack. ChatGPT is great at rephrasing and rewriting texts by applying a certain style and matching the required context using its AI intelligence. I think what it is doing in this experience is a process of two steps, it translated the text, then rewrites it using AI, making it more readable and more human-like, not machine-like.
It is in the end a language-based AI and a literary genius that has access to a huge database.
PS: Maybe this is limited to my experience with ArabicEnglish.
Thank you. That is exactly my point.
I've been experimenting with ChatGPT as a translation aid for a couple of weeks now, using short news articles as the source text. If I give it a prompt that simply says, “Translate the following text,” ChatGPT gives me something that is virtually identical to a Google Translate translation. If I then tell it to rewrite the translation as a news article in a style suitable for publication, it usually does a good job of it. Sometimes, however, I get back strange results. I did that yesterday with a Ukrainian news article, but when I asked for a rewrite, ChatGPT gave me an essay on some random topic completely unrelated to the original text. It did that several times in a row. When I finally got it to do a rewrite, the quality was poor — much worse than I had come to expect.
I'm being asked more frequently to do a post-edits of a machine translations. I've usually declined, because the quality of the translation has been so poor, not usually because of obvious translation errors, but most often because of extremely bad style. However, I'm convinced that demand for post-editing of machine translations will increase and probably will become the norm in the not-too-distant future. I have hope that an AI-based translation aid — probably not this iteration of ChatGPT — will be able to improve MT style and let translators concentrate on finding the inevitable errors.
| | | | jyuan_us
United States
Local time: 11:12
Member (2005)
English to Chinese
+ ...
| Study is needed | Feb 27, 2023 |
A study is needed to reject or support the null hypothesis that Chat GPT is better than GT.
| | | | Daryo
United Kingdom
Local time: 16:12
Serbian to English
+ ...
| That in fact illustrates something else | Dec 17, 2023 |
Philip Lees wrote: Lingua 5B wrote:
Me: Is there any word or phrase that’s ambiguous in this sentence: A group of English students were eating at a diner.
GPT: There is no ambiguous word or phrase in the sentence "A group of English students were eating at a diner" as it is written. The sentence is clear and straightforward in its meaning, and there are no words or phrases that have multiple possible interpretations or could be misunderstood.
-
Actually, “English students” can be either English majors or students coming from England. This illustrates why good translators have nothing to fear from ChatGPT and its ilk. The AI doesn't see any ambiguity in the sentence because it only parses it one way. When I asked it to translate the sentence into Greek, it rendered the phrase as "English-speaking students" When I changed the "source" sentence to "A group of English and Philosophy students were eating at a diner.", the result was "A group of English speakers and Philosophy students ..." Obviously ChatGPT is strongly biased towards the use of English to indicate a language, rather than a nationality. That makes sense, given that the number of English speakers in the world is far greater than the number of English people, so the material on which the AI bases its responses is likely to have a similar bias. With "A group of Philosophy and English students were eating at a diner.", ChatGTP got it right (at last!), by indicating that the students were studying the English language, rather than just speaking it. ChatGTP is only as good as the data it was trained on - and often not even that good. I have found it useful for tracking down abbreviations and other special terminology. Others may not. But to see it - in its present form - as some kind of potential professional competitor, or even as a threat, is just silly.
I wouldn't be so optimistic about how many people would spot that "A group of English students were eating at a diner" CAN BE very ambiguous, and how many would swallow whatever ChatGTP sends their way before giving it first to a "poison-tester".
As for: "This illustrates why good translators have nothing to fear from ChatGPT and its ilk."
There is still a very real danger - not so much from ChatGTP & Co but mainly from the unrealistic end users and their middleman who grossly overestimate the capabilities of ChatGTP & Co. In a way, just a repeat of the "Machine Translation" saga.
[Edited at 2023-12-17 03:08 GMT]
| | | | | Pages in topic: < [1 2 3] | There is no moderator assigned specifically to this forum. To report site rules violations or get help, please contact site staff » ChatGPT as a potential tool for translators | Protemos translation business management system | Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!
The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.
More info » |
| | 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 » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
Translation news
