| Vom Thema belegte Seiten: [1 2] > | Off topic: ChatGPT and multilingualism Initiator des Themas: Maria Teresa Borges de Almeida
|
|---|
Another interesting article published on ‘The Economist’ (April 1st 2023), this time about ChatGPT and its “multilingualism”. Enjoy!
P.S. I am not sure if this is the right place to post this thread, so please feel free to move it to a more appropriate forum…
“Speaking in many tongues
ChatGPT may make things up, but it does so fluently in more than 50 languages
The hype that followed ChatGPT’s public launch last year was, even by ... See more Another interesting article published on ‘The Economist’ (April 1st 2023), this time about ChatGPT and its “multilingualism”. Enjoy!
P.S. I am not sure if this is the right place to post this thread, so please feel free to move it to a more appropriate forum…
“Speaking in many tongues
ChatGPT may make things up, but it does so fluently in more than 50 languages
The hype that followed ChatGPT’s public launch last year was, even by the standards of tech innovations, extreme. OpenAI’s natural-language system creates recipes, writes computer code and parodies literary styles. Its latest iteration can even describe photographs. It has been hailed as a technological breakthrough on a par with the printing press. But it has not taken long for huge flaws to emerge, too. It sometimes “hallucinates” non-facts that it pronounces with perfect confidence, insisting on those falsehoods when queried. It also fails basic logic tests.
In other words, ChatGPT is not a general artificial intelligence, an independent thinking machine. It is, in the jargon, a large language model. That means it is very good at predicting what kinds of words tend to follow which others, after being trained on a huge body of text—its developer, OpenAI, does not say exactly from where—and spotting patterns.
Amid the hype, it is easy to forget a minor miracle. ChatGPT has aced a problem that long served as a far-off dream for engineers: generating human-like language. Unlike earlier versions of the system, it can go on doing so for paragraphs on end without descending into incoherence. And this achievement’s dimensions are even greater than they seem at first glance. ChatGPT is not only able to generate remarkably realistic English. It is also able to instantly blurt out text in more than 50 languages—the precise number is apparently unknown to the system itself.
Asked (in Spanish) how many languages it can speak, ChatGPT replies, vaguely, “more than 50”, explaining that its ability to produce text will depend on how much training data is available for any given language. Then, asked a question in an unannounced switch to Portuguese, it offers up a sketch of your columnist’s biography in that language. Most of it was correct, but it had him studying the wrong subject at the wrong university. The language itself was impeccable.
Portuguese is one of the world’s biggest languages. Trying out a smaller language, your columnist probed ChatGPT in Danish, spoken by only about 5.5m people. Danes do much of their online writing in English, so the training data for Danish must be orders of magnitude scarcer than what is available for English, Spanish or Portuguese. ChatGPT’s answers were factually askew but expressed in almost perfect Danish. (A tiny gender-agreement error was the only mistake caught in any of the languages tested.)
Indeed, ChatGPT is too modest about its own abilities. On request, it furnishes a list of 51 languages it can work in, including Esperanto, Kannada and Zulu. It declines to say that it can “speak” these languages, but rather “generates text” in them. This is too humble an answer. Addressed in Catalan—a language not on the list—it replies in that language with a cheerful “Yes, I do speak Catalan—what can I help you with?” A few follow-up questions do not trip it up in the slightest, including a query about whether it is merely translating answers first generated in another language into Catalan. This, ChatGPT denies: “I don’t translate from any other language; I look in my database for the best words and phrases to answer your questions.”
Who knows if this is true? ChatGPT not only makes things up, but incorrectly answers questions about the very conversation it is having. (It has no “memory”, but rather feeds the last few thousand words of each conversation back into itself as a new prompt. If you have been speaking English for a while it will “forget” that you asked a question in Danish earlier and say that the question was asked in English.) ChatGPT is untrustworthy not just about the world, but even about itself.
This should not overshadow the achievement of a model that can effortlessly mimic so many languages, including those with limited training data. Speakers of smaller languages have worried for years about language technologies passing them by. Their justifiable concern had two causes: the lesser incentive for companies to develop products in Icelandic or Maltese, and the relative lack of data to train them.
Somehow the developers of ChatGPT seem to have overcome such problems. It is too early to say what good the technology will do, but this alone gives one reason to be optimistic. As machine-learning techniques improve, they may not require the vast resources, in programming time or data, traditionally thought necessary to make sure smaller languages are not overlooked online.” ▲ Collapse
| | | | Philip Lees 
Griechenland
Local time: 05:02
Griechisch > Englisch
| Greek data point | Apr 12, 2023 |
On a whim yesterday (i.e. before reading Maria Teresa's post), I asked ChatGPT, in English, to check a short Greek email text I'd written for errors. Its answer was in Greek, which I wasn't expecting. Very good Greek, too.
Maybe somebody else would like to try asking ChatGPT in Portuguese (for example) to check a text in a different language, and see which language it chooses for its reply.
| | | |
The article’s author did ask ChatGPT a question in Spanish, Portuguese and Danish and said that even though the answer contained some errors the language was impeccable…
| | | | Tom in London
Vereinigtes Königreich
Local time: 03:02
Mitglied (2008)
Italienisch > Englisch
| Tired already | Apr 12, 2023 |
I'm tired of ChatGPT already. What's the next terribly exciting thing they want us to think about?
[Edited at 2023-04-12 08:02 GMT]
| | |
|
|
|
Philip Lees
Griechenland
Local time: 05:02
Griechisch > Englisch
Maria Teresa Borges de Almeida wrote:
The article’s author did ask ChatGPT a question in Spanish, Portuguese and Danish and said that even though the answer contained some errors the language was impeccable…
I'm more interested in the cases where the answer is in a different language from the question, as in my example. What made it switch?
As for Tom's comment, I too am tired of the interminable discussions about whether AI will put translators out of business (it won't, at least not in the near future), but I'm fascinated by the technical details of what it gets right and what wrong, and how it adjusts its behaviour in response to a prompt or series of prompts.
This is perhaps because of my second career, as a computer programmer. I asked ChatGPT the other day to generate a visual basic script for use in Word that would convert small numbers written in full (one, two, three, etc.) to their numerical equivalents (yes, I know there are many potential pitfalls here, e.g. avoiding atrocities like fr8, or asi9).
It said, "Sure I can," and then proceeded to produce a script that wasn't even close to working. When I pointed this out, it admitted it and apologised. Then it produced something that looked as if it would work, but was a horrendously inefficient piece of coding.
So I don't think computer programmers need to worry just yet, either.
| | | |
I see now what you meant. I'll do it later on and I'll come back with the results...
On another note, we all know that plagiarism in universities is not a new topic, but I wonder if it will be more difficult to detect…
| | | | Tom in London
Vereinigtes Königreich
Local time: 03:02
Mitglied (2008)
Italienisch > Englisch
Maria Teresa Borges de Almeida wrote:
I see now what you meant. I'll do it later on and I'll come back with the results...
On another note, we all know that plagiarism in universities is not a new topic, but I wonder if it will be more difficult to detect…
As a former university tutor I can assure you that plagiarism is easy to detect, if you know the student and you have been following their progress and talking to them.
If you read an essay they've written that uses vocabulary they would not use, or attains a level of literacy you know they could never attain, their essay is probably plagiarised. Then there are other web tools that will help you to detect plagiarism.
From the student's viewpoint it's important they should know that plagiarism is the most severe offence you can commit (academically) and may well result in being expelled.
[Edited at 2023-04-12 09:57 GMT]
| | | | Kay Denney
Frankreich
Local time: 04:02
Französisch > Englisch
A friend recently posted on social media that she had detected the use of ChatGPT in essays by two different students. She didn't explain how she twigged because she didn't want the students to merely perfect the art of cheating. She was pretty chuffed.
| | |
|
|
|
When I taught translation at university, the only work I received was actual translations. How would you tell whether these were AI-assisted? And don’t say because they’d contain mistakes, because the students would make mistakes too.
Then again, maybe tomorrow’s translators (and students in general) need to be cheating with AI. Presumably it will be the new normal.
| | | | Lingua 5B
Bosnien und Herzegowina
Local time: 04:02
Mitglied (2009)
Englisch > Kroatisch
+ ...
| Robotic flow | Apr 12, 2023 |
Kay Denney wrote:
A friend recently posted on social media that she had detected the use of ChatGPT in essays by two different students. She didn't explain how she twigged because she didn't want the students to merely perfect the art of cheating. She was pretty chuffed.
Aside from a robotic flow and occasional logic errors (can be 1 error in 1k words, but enough for a teacher to spot), there are tools that can detect this. But it will be clear to an experienced teacher after the first two sentences.
| | | | expressisverbis
Portugal
Local time: 03:02
Mitglied (2015)
Englisch > Portugiesisch
+ ...
| Le Zhang
China
Local time: 11:02
Mitglied (2021)
Englisch > Chinesisch
+ ...
| seriously worried | Apr 13, 2023 |
To be honest, chatgpt is making me quite anxious about my professional future as a translator, so much so that I can barely focus on work in recent days. I'm thinking of changing courses, but have no idea what else to do, as the old Chinese saying goes, "夜来思量千条路，晨起依旧推磨卖豆腐after a thousand what if's over the night, the second day's still occupied by beancurd making."
| | |
|
|
|
Lieven Malaise
Belgien
Local time: 04:02
Mitglied (2020)
Französisch > Niederländisch
+ ...
| AI might be too good for cheating | Apr 13, 2023 |
Ice Scream wrote:
When I taught translation at university, the only work I received was actual translations. How would you tell whether these were AI-assisted?
DeepL makes translation and interpretation errors and is stylistically far from flawless, but makes virtually no spelling and grammar mistakes, let alone typos. In Dutch (not the easiest language to write) that should be a huge red flag for any teacher dealing with inexperienced students.
Any wicked student would of course add a few 'mistakes' to cover up for his godless behaviour.
| | | | Denis Fesik
Local time: 06:02
Englisch > Russisch
+ ...
| Is this text machine-translated? | Apr 13, 2023 |
Le Zhang wrote:
after a thousand what if's over the night, the second day's still occupied by beancurd making
And if it's MT, which engine was it? I'm asking because I ran the Chinese text through Q-translate, and the translations I got from Google and DeepL were definitely clearer although I did understand the main point without them (according to DeepL, the daytime routine consisted of pushing the mill and selling tofu while Google just wrote about selling tofu – can't tell how accurate those translations were because I can't read Chinese).
I can tell why I'm not worried just yet. Whenever I get to read ChatGPT's output (in videos), I get bored within seconds. Some people make a living by producing texts of similar fashion and are probably worried by how effortlessly a machine can replace them. I once had to write two papers on Russian history to get a passing grade in university, so I went to the library, got some books, and wrote the papers without using a computer. The teacher didn't even read them. I guess I could have used ChatGPT to have the same work done in seconds, but the point is that the output would be as bland, disinteresting, and idea-less as anything I've ever seen this AI thing produce. As for interesting and insightful texts, I have yet to see one translated by a machine in a way that wouldn't make me want to rewrite it from scratch. I keep checking, and AI keeps failing me every time; it can only do basic stuff for which no one in their sound mind will want to hire a translator in the first place.
As I already mentioned in other threads, this may have to do with my language pair: we just don't have huge corpora of really robust, professional translations on which AI could be trained. Any good translator may create such a corpus, but it will be relatively small. We have old documents that are well-written but don't exist in English, and we have modern translations from English that are written horribly (it's the overall picture as I see it). There are people claiming that Soviet-time translators (and members of all other trades) were real pros unlike today's slackers, but whenever you get to read the official English titles of old GOST standards, for example, you begin to doubt the universal truth of such assertions. A lot of the developments in 20th-century science and engineering in the USSR were relatively isolated from what was going on behind the Iron Curtain, and the result is that we often talk about the same things in very different ways, so you have to be mindful of those differences when translating. I noticed, btw, that certain English texts that had to do with nuclear power engineering and space technology were drawing on something that was originally written in Russian – there's no way I'll overlook a language pattern that's characteristic to Runglish
| | | | Lingua 5B
Bosnien und Herzegowina
Local time: 04:02
Mitglied (2009)
Englisch > Kroatisch
+ ...
| A few mistakes | Apr 13, 2023 |
They can add a few mistakes but can't change the entire flow and subtext. Impossible. We are talking about essays here, not translations.
I will paste three passages below. Can you tell me whether some of them have been produced by GPT?
Passage I
I settle on the couch with my coffee and book in hand. As I turn pages, the sound of rain outside creates a background melody that adds to the peaceful ambiance. It's a moment of pure bliss as I lose myself in th... See more They can add a few mistakes but can't change the entire flow and subtext. Impossible. We are talking about essays here, not translations.
I will paste three passages below. Can you tell me whether some of them have been produced by GPT?
Passage I
I settle on the couch with my coffee and book in hand. As I turn pages, the sound of rain outside creates a background melody that adds to the peaceful ambiance. It's a moment of pure bliss as I lose myself in the pages of book, forgetting about the outside world. I glance up from my book and see the raindrops sliding down the window, and I feel grateful for this moment of stillness.
Passage II
Children and adults alike need to experience how rewarding it is to work at the edge of their abilities. Resilience is the product of agency: knowing that what you do can make a difference. Many of us remember what playing team sports, singing in the school choir, or playing in the marching band meant to us, especially if we had coaches or directors who believed in us, pushed us to excel, and taught us we could be better than we taught was possible.
Passage III
When we arrived there, the house was empty. We tried calling them but nobody was answering the phone. So I took a stroll to the neighborhood to check if somebody could help us. There was an old lady down the road who was very friendly but said she didn't have contact with Mary and Steve and wouldn't know where they might be at that point. So we went to the village to have lunch and make plans for the rest of the day. We enjoyed some tasty local food in this lovely picturesque village.
[Edited at 2023-04-13 10:02 GMT] ▲ Collapse
| | | | | Vom Thema belegte Seiten: [1 2] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » ChatGPT and multilingualism | 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! » |
| | 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 » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
