I freelance translated full-time for about 10 years but transitioned to another career a little over a year ago. I’ve kept translating for my old company part-time just for some extra cash but recently they told me that if I want to keep working, I’ll have to take a 1/3 pay cut, saying AI tools were driving down prices.
For those of you still in it, how are you coping? Are the new tools allowing you to work faster so you can maintain income or even improve? Are you worried, optimistic, business as usual?
22 comments
It’s both reassuring and concerning to hear about the changes happening in the translation industry. On one hand, the introduction of AI tools has undeniably shifted the landscape, making tasks more efficient and allowing us to handle larger volumes. On the other hand, it’s disheartening to see the impact on our earnings, like the 1/3 pay cut you mentioned.
Personally, I’ve been working as a freelance translator for around 5 years now, and I’ve noticed a similar trend. While AI tools have indeed accelerated some aspects of translation, I’ve also found that they can’t replicate the nuanced cultural and contextual understanding that a human touch brings to the table.
In response to these changes, I’ve focused on specializing in areas where human interpretation is crucial, such as marketing copy, creative writing, and legal documents that require precision beyond what AI can offer. This has helped me maintain my income to some extent, but it’s a constant balancing act.
I’m a mix of worried and optimistic about the future. On one hand, AI has the potential to enhance our workflow and allow us to focus on more creative and intricate aspects of translation. On the other hand, there’s the looming concern of a continued downward pressure on prices due to the convenience of AI-driven tools.
I actually just finished up ‘reviewing’ a 40-page document that was translated by DeepL. It was utter gibberish. I more or less rewrote the whole thing.
The reasons why AI is still nowhere near useable, at least in Japanese, are:
– Japanese (fixed typo) is a high context language, it often leaves out the subject from sentences. AI gets lost with this and will mix up the subject from sentence to sentence.
– AI gets very very lost when dealing with bullet points, especially stuff like “Chapter 3 Article 4 Section 6” and trying to keep of indents and numbering and stuff.
– It’s still not very good at platitudes that need to be ‘reworked’ for English, eg. stuff like いつも大変お世話になっております。
– It will translate words differently almost at random, and you lose consistency, which is super important for technical documents (eg. client vs. customer, contractor vs. dispatcher, stuff like that).
The end result is that anything translated by AI, at least at this point, still absolutely needs to be thoroughly reviewed by a human, essentially adding a whole extra layer of work. If I had just translated the document myself from scratch I could have kept track of what terminology to use, formatting, etc. So I don’t feel my job is in danger yet.
DeepL does 70% of my work. I wouldn’t want to work without it again.
That said, anyone who relies on it without a professional human checking the results is a fool.
As an engineer working in japanese, I have to admit DeepL saves as lot of time / brain resources.
Copy/paste text and then check “manually” is less tiring that reading the whole thing. (N2 level since 2 years, targeting N1 some time soon FYI)
The effectiveness of AI depends on the type of translation, and how much the client cares about accuracy/natural English. Something like an infographic for a train station or a travel website, yeah sure, AI could probably handle that. It will not be perfect, but if the client doesn’t care then it gets the job done.
But longer, more nuanced translations…nah, AI can’t handle it. At least not yet. I’m doing some entertainment-based translation, and AI can’t really catch nuance, or create unique character styles.
I’m actually making more overall than last year. I recently did a comparison of this year and last and most months this year I’ve made more – significantly more, as much as 70% in some cases. The only difference is that the start of the new FY was much slower than last year.
I’ve only lost one direct client to machine translation – one that wasn’t high paying at all actually (despite being direct, they were my lowest paying company by far). I’ve had very large projects from other companies (2 jobs that were just over 1 million total) that gave me a bit of a savings buffer to quit my other jobs and try out full time freelancing.
So overall this year has been going very well for me. I don’t think AI is going to have a major impact on my work just yet, but I’m not naive and am starting to think about ways to strengthen my skill set and maybe transition into something else in the coming years if necessary. But I really enjoy this job and hope that doesn’t happen.
>but transitioned to another career a little over a year ago.
Tell us more
“AI” has been used in translation for well over a decade now. The work of tens of thousands of translators have been feeding machine translation engines for years, long before artists and content creators started complaining about Chat-GPT and the likes. On the plus side however, as machine translation is getting better, the end result is often superior to that of bad translators, making proofreaders and editors job somewhat easier.
An acquaintance of mine, a native English speaker whose Japanese is also nearly fluent due to his background, works as a regular office employee while also translating Japanese into English within the biochemistry industry at his company. This level of profession cannot be replaced by AI easily. However, AI translation technology is advancing fast, so individuals who translate random articles from Japanese to English (or vice versa) may indeed be at risk of job displacement.
I’ve come across numerous Japanese articles that were already translated by DeepL or Google Translate. I’m not sure who used to handle this task before AI translation became prevalent, but those individuals unfortunately lost their jobs.
Deepl is a turd. Chatgpt much better.
I hate when I receive “pre-translated” stuff by AI because it takes longer to check than it does to just do it myself.
Also, translation is just part of my work, so I haven’t received anything like a pay cut. I think most people that work in global positions realize that machine translation, while sometimes useful, is definitely not quality or a desirable deliverable
Only positive impacts so far. Implementing it into my existing workflow has increased my productivity substantively. I’m riding the wave with a beer in one hand until it crashes.
Personally I think we are in a (brief) golden age right now – AI is good enough to be a major productivity boost but not good enough to take away jobs, Japanese companies have huge English-language disclosure requirements that they cannot handle in-house, and will pay premium rates to outsource their translations.
It’s definitely going to impact the market in the future, but make hay while the sun shines.
The old company is full of shit, and have no idea what they’re talking about. I do video productions and often hire translators, or I do them myself. There’s still a lot of manual labor. These apps are great assist, but Japanese is a little special case due to nuance. Writing Japanese emails takes so much mind games, I hate it.
I got a 7% pay raise last year, and then 4% this year to keep up with inflation.
AI isn’t particularly useful for technical subjects like finance (which is my field of expertise) and honestly, I’d rather do the whole thing from scratch rather than having to correct multiple errors every other sentence.
Still, whether AI gets to the point where it can automate the translation process entirely or whether upper management at these companies eventually convince themselves that it can, the writing is on the wall for this industry.
Exactly how long it’ll take is anyone’s guess, but I’m guessing we have maybe another decade or so. Who knows?
Once I get permanent residency and enough money to buy land and build a house, I’m out.
I haven’t actually decided whether I’ll just hang on until I get fired or whether I’ll quit of my own volition at some point, but I can’t imagine myself being corporate cattle for much longer, i.e., for another 5+ years.
Saying it’s taxing on one’s mental health is quite the understatement: it’s indentured servitude to people who hate you and covertly (sometimes even overtly) think you’re replaceable.
Rather than living in constant fear of being fired and not being able to pay the bills, I’d rather develop useful skills like growing my own food so I can keep expenses to a bare minimum and stop worrying about money.
With enough DIY skills I reckon I can keep my living expenses under 20,000 yen plus health insurance and social security, and below a certain income threshold you can get exemptions for those, too.
I translate instruction manuals for professional equipment and related tech documents, stuff like DeepL can’t replace me for the same reason why my company can’t outsource our translations – you need to fully understand said equipment/devices.
I need to use the devices while translating an instruction manual, I need to ask tons of questions to the Japanese writer and the engineers.
Who knows if I’ll still be employed in the future, but I think there are certain translation areas that will always need a human to at least proofread the content. Maybe not a restaurant menu, but stuff like instruction manuals, contracts, etc.
I’m not a professional translator, but I do unauthorized fan translations of manga, presumably because I’m a weirdo.
I don’t use machine translation at all, and I can spot bad machine-done manga translations a mile away. Google’s sentences sound terrible. DeepL makes nice-sounding sentences but just leaves entire clauses out when it gets confused. Both of them get completely the wrong idea about what’s going on from time-to-time.
Someone who knows Japanese can “babysit” the machine translators and get a decent outcome, I guess, but I think a lot of people will run stuff through a machine translator and then hand it directly off to English-only (or whatever the target language is) editors, who will make it sound “nice” based on assumptions or guesses about what the English output of the machine translator is, and that just causes too many translation errors.
Maybe it would work better with more “standard written Japanese” than “spoken Japanese written down and then broken up into small speech bubbles”, but I doubt it.
I’m sure it’ll improve eventually, but for now it all looks like ass.
My wife is a translator for a construction company who mainly deals with the US naval base in Sasebo, she loves openAI apps, makes her tasks so much easier. And since japan is so behind with systems and AI tools, she can do the work of 3 people. This bumped her to the head of her department.
A lot of people are pointing to DeepL and Google Translate, but those are not the translation services companies are using. There are background AI companies with very expensive translation services that are extremely competent – those are the services that translators will have to compete with in the near future.
I got asked to do a pre-translated check of some AI translated stuff. The sentences were all well formed, but like others I don’t think it translates terms consistently enough etc.
It was more work than doing the translation myself.
I’m not a professional, and I don’t think it’s ready for prime time. I think it has it’s uses (quick checks of documents when the budget doesn’t extend to a translator).
On the positive side it could increase the high value jobs for actual humans (delicate documents, full book translations) if people are educated on it’s shortcomings.
Unfortunately though, it is oversold and people look at the first couple of sentences and think that because it feels like quality that they don’t need to spend the money on a real translator. Then bad jobs, creating less pleasant work, like checking, and essentially retranslation and going back and forth endlessly, are born, and they are assigned lower values.
And for the true AI translation effect, how is it infecting you wives?
My experience tells me that it really depends on the content. For formal documents like legal, contract stuff, DeepL seems to be more consistent as the language used are more precise.
ChatGPT on the other hand shines more on less formal conversations or information. It is usually able to discern what the subject is when omitted.
But to say thay your pay needs to be 1/3 because AI – yeah whoever said that certainly has a turd brain.