Few industries have been hit by AI as hard as translation. Rates are plummeting. Work is drying up. Translators are considering abandoning the field, or bankruptcy. These are their stories.
Thanks for sharing our stories! Just a quick update: I was indeed forced out of the translation industry, and I’m now working in marketing as a bilingual copywriter. My boss even makes me use AI to speed things up (regardless of quality issues). Worst of all, he trusts everything ChatGPT says, and says that if I double-check the AI's copy or do my own research, I’m just wasting precious time.
These people grossly underestimate the problem of hallucinations. You should try to educate them about it. Let him use the system with no error checking with his name and reputation on it 😆
What I'm hearing from this is that the AI buzzword gave corporations an excuse to turn cultural products — notably online content and video games — into indistinct, barely intelligible mush. Really does seem like we're heading into the post literate age.
I get really angry when I hear about this concept of translators being replaced because my job requires me to work with translations all the time and I specifically request human translation rather than machine translation (even with human supervision and redaction) because the latter sucks. If it happens that we need to use a machine, sometimes we have to look for a native speaker to go through the original text for us because we have zero clue what is going on. Give it a typo or any human error or slang and it’s dunzo. Respect to all translators and I hope decision makers will soon realize that they need your work rather than half-assed machine produce.
I majored in Translation, have a Master's in Translation Technologies and have been working in Localization/Translation management for over 10 years.
When I joined my current company (a tech company where I work as Localisation manager), I was able to cut costs without being asked to - by using freelance translators instead of agencies (which add a big markup) and implementing translation memories. I pay an above-the-market rate to my translators, who I have good relationships with, some of them for over 10 years.
Over the last 3 months, my managers keep pushing me to cut costs. As I read this piece, I have an Excel sheet open to see how to cut 30% of translation costs - and of course, "AI" needs to be involved somehow. Not because it's proven to be good at what it does (it isn't), but because it's a task the AI salespeople love showcasing, partially because it creates okay drafts sometimes, partially because only localization professionals can tell when a translation is bad.
I manage +30 languages. I speak 4 of these fluently, and other 6-7 I can "vaguely" QA. That's enough to tell that LLMs CANNOT be trusted with proper uncontrolled translation, of pretty much any content. They rarely commit grammar or spelling errors (they're LANGUAGE models after all), but they're inconsistent with themselves, they don't really follow prompts, they can't tell how many characters a sentence is: and of course, only a translator or localization person can tell.
I am in my mid 30s and, even if my job is "safe" for at least a year or two (they need someone to Copy and Paste the GPT output after all), I am already thinking of reskilling to a blue collar job. I will probably need to abandon the country that I love and where I moved to, go back to living with my family, take shit jobs, lose my "middle class" status, but I will be damned if I spend years of my life reskilling to be replaced by AI again or ever have to sit across a privately educated manager talking about KPIs, OKRs, career development.
I'm a Japanese/English interpreter and translator. Unfortunately, for my language pair at least, I think ChatGPT and other LLMs have been a huge step forward in automated translation. Like night and day compared with deepL or google translate. It's actually dealt quite well with human errors, slang, and jargon, in my experience. Not as good as a seasoned pro, but with a non-specialized text, it may well deliver a better output than someone just starting out.
Do I think companies should hire AI translators instead of humans? Of course not. But to discount LLMs as no different than previous technology totally contradicts what I have seen, personally.
My heart is with all these folks, translation is an art form and thus a canary in the coal mine situation. The observations of “flattening,” “loss of nuance and characterization” should be alarming for all of us.
The testimonies of translators in the OP are heartbreaking to read. At the moment, though, there might be a few industries where the flow may be going the other way. One is financial services, at least in certain niches such as the asset management and hedge fund space.
My wife and a colleague formed a boutique to service this area with translation about 12 years ago, after she’d had a long career on the operational side of that business in Japan. All her contractors also must have operational experience with some type of assets, e.g. bonds, equities, swaps, funds of funds, etc. She’s seeing an uptick in orders after a short earlier decline, as clients are getting frustrated with MT work product.
Maybe other areas that require a lot of domain expertise can provide some glimmer of hope that human translation will still be valued by customers.
Thanks for sharing our stories! Just a quick update: I was indeed forced out of the translation industry, and I’m now working in marketing as a bilingual copywriter. My boss even makes me use AI to speed things up (regardless of quality issues). Worst of all, he trusts everything ChatGPT says, and says that if I double-check the AI's copy or do my own research, I’m just wasting precious time.
These people grossly underestimate the problem of hallucinations. You should try to educate them about it. Let him use the system with no error checking with his name and reputation on it 😆
What I'm hearing from this is that the AI buzzword gave corporations an excuse to turn cultural products — notably online content and video games — into indistinct, barely intelligible mush. Really does seem like we're heading into the post literate age.
I get really angry when I hear about this concept of translators being replaced because my job requires me to work with translations all the time and I specifically request human translation rather than machine translation (even with human supervision and redaction) because the latter sucks. If it happens that we need to use a machine, sometimes we have to look for a native speaker to go through the original text for us because we have zero clue what is going on. Give it a typo or any human error or slang and it’s dunzo. Respect to all translators and I hope decision makers will soon realize that they need your work rather than half-assed machine produce.
I majored in Translation, have a Master's in Translation Technologies and have been working in Localization/Translation management for over 10 years.
When I joined my current company (a tech company where I work as Localisation manager), I was able to cut costs without being asked to - by using freelance translators instead of agencies (which add a big markup) and implementing translation memories. I pay an above-the-market rate to my translators, who I have good relationships with, some of them for over 10 years.
Over the last 3 months, my managers keep pushing me to cut costs. As I read this piece, I have an Excel sheet open to see how to cut 30% of translation costs - and of course, "AI" needs to be involved somehow. Not because it's proven to be good at what it does (it isn't), but because it's a task the AI salespeople love showcasing, partially because it creates okay drafts sometimes, partially because only localization professionals can tell when a translation is bad.
I manage +30 languages. I speak 4 of these fluently, and other 6-7 I can "vaguely" QA. That's enough to tell that LLMs CANNOT be trusted with proper uncontrolled translation, of pretty much any content. They rarely commit grammar or spelling errors (they're LANGUAGE models after all), but they're inconsistent with themselves, they don't really follow prompts, they can't tell how many characters a sentence is: and of course, only a translator or localization person can tell.
I am in my mid 30s and, even if my job is "safe" for at least a year or two (they need someone to Copy and Paste the GPT output after all), I am already thinking of reskilling to a blue collar job. I will probably need to abandon the country that I love and where I moved to, go back to living with my family, take shit jobs, lose my "middle class" status, but I will be damned if I spend years of my life reskilling to be replaced by AI again or ever have to sit across a privately educated manager talking about KPIs, OKRs, career development.
You obviously have an impressive ability to figure stuff out and manage things at a high level. Don’t let them sell you short!
I'm a Japanese/English interpreter and translator. Unfortunately, for my language pair at least, I think ChatGPT and other LLMs have been a huge step forward in automated translation. Like night and day compared with deepL or google translate. It's actually dealt quite well with human errors, slang, and jargon, in my experience. Not as good as a seasoned pro, but with a non-specialized text, it may well deliver a better output than someone just starting out.
Do I think companies should hire AI translators instead of humans? Of course not. But to discount LLMs as no different than previous technology totally contradicts what I have seen, personally.
My heart is with all these folks, translation is an art form and thus a canary in the coal mine situation. The observations of “flattening,” “loss of nuance and characterization” should be alarming for all of us.
The testimonies of translators in the OP are heartbreaking to read. At the moment, though, there might be a few industries where the flow may be going the other way. One is financial services, at least in certain niches such as the asset management and hedge fund space.
My wife and a colleague formed a boutique to service this area with translation about 12 years ago, after she’d had a long career on the operational side of that business in Japan. All her contractors also must have operational experience with some type of assets, e.g. bonds, equities, swaps, funds of funds, etc. She’s seeing an uptick in orders after a short earlier decline, as clients are getting frustrated with MT work product.
Maybe other areas that require a lot of domain expertise can provide some glimmer of hope that human translation will still be valued by customers.
Thanks for sharing your story.