Last year, I met a young translator reinventing herself.
She studied Translation for five years at a local university. We met online by commenting and engaging on each other's LinkedIn posts. When we met, she was looking for another way to make money. Her agency cut down her rates and couldn't find her as many gigs as before.
Who to blame? AI!
AI isn't disappointing at translation. But it doesn't score A+
Her story came to mind when I started my own translation project, and I quickly understood why she was looking for another job.
Recently, I translated Street-Smart Coding to Spanish. As a native Spanish speaker, I could have done it easily myself.
But to make it faster, I used Copilot with a simple prompt. I specified a tone, voice, and style. Latin American conjugations and vocabulary instead of Spaniard ones, for example.
I was surprised by the results.
Copilot translated chapters with almost no fixes. Of course, there were places where the phrasing made it clear the text was machine-generated.
In English, we say "wear all hats" when someone has to do multiple tasks alone. Copilot translated word by word. The same expression in Spanish ("usar todos los sombreros") makes no sense at all.
Copilot struggled with coding terms like "parser combinators." In Spanish, we use a completely different term. Direct translation doesn't work either.
That's when I jumped in. But Copilot handled most of the job in just a couple of work sessions.
What if it isn't only translation, but coding too?
This made me rethink one of the lines I heard the other day:
"AI won't take your job. It will change your job description."
(Credits to Kevin Kelly)
AI may already be turning translators into proofreaders. Coders could be next. Maybe the world won't need as many coders, and coding may no longer mean typing symbols anymore. Who knows?
In any case, I predicted AI won't take our jobs by 2034. Let's see if I was right. In the meantime, I'd like to pick a DIY skill and double down on my creative and writing skills—just in case.
When AI generates code faster than any of us, you need more than syntax to stand out. That's why I wrote Street-Smart Coding, a roadmap to build hype-proof skills.
Top comments (21)
Translation has been a moving target for years... the use of AI to translate was an unforeseen advantage. But, as you say, there still must be a human taking care of fixing those parts that are not "automatic", I mean, which still have to do with human imagination.
There still exists the problem of AI hallucinations, or, more precisely, the resort for AI to invent a new reality instead of fixing the real problems (eliminating test cases, for instance).
The other day I was listening to an expert in automatic driving. He said that AI was in charge of driving the car, while a simpler, rule-based AI was in charge of verifying the main AIs driving. I imagined that the rule-based system was needed in case the driving AI decided to abandon the road in order to avoid traffic...
That's where our real edge lies
Or decided to bump into pedestrians for no reason :P
There is something sad about this shift... even though my career path led me to the IT industry, my initial goal at the university was to become a translator. It always seemed to me like a creative profession, and some magic is lost with knowing that AI is doing most of the job now. There were already some parts of the original book lost in translation, but a translator's job, as a kind of co-author of a local copy, was to keep the magic in. I'm afraid this is what we are losing when allowing AI to do most of the job.
The example with idioms is spot-on. I'd also add poetry – it's the hardest part to translate, even manually, as it requires translation of multiple layers: not only words, but often also sound, and – most importantly – the feelings that the text brings.
Great point! We can notice it in all the slop flooding the internet these days.
AI changes the job description before it removes the job.
Coders shouldn’t panic, but they should pay attention. AI is already changing the job. The safer path is learning how to review, debug, design systems, and use AI well instead of only writing code faster.
I also like to think AI isn't taking jobs, but changing job descriptions. (Credits to Kevin Kelly for that frame)
That’s a good way to put it. The job isn’t disappearing overnight, but the expectations are changing fast. Writing code is becoming only one part of the role. Understanding, reviewing, and owning the outcome matters more now.
the translator parallel is sharp. my read: the people who thrive move up a level, from doing the task to orchestrating + verifying the system that does it. that's literally what I build at Moonshift: agents build + deploy + market a SaaS overnight, and the human job becomes designing the harness, not writing the code. good provocation. first run's free if you want to feel the shift firsthand.
I think this part is where we should focus on. AI can do numerous things in a decent way, of course. But there's a certain limit regardless of fields. And only experts can make AI overcome the limit. That's why I consider AI as a booster for skilled people.
Translators would no longer have to tranlate all the phrases. But only they would be able to fix 'wear all hats'. Developers are the same. They would no longer have to write code from scratch, but only they would be able to optimize the system and fix bugs.
I like to imagine AI as a powerful calculator: it needs someone who knows Math, otherwise it's a toy
This is the way!
The real barrier to entry has never been writing code itself, but rather architectural design, hardware-software synergy, and deterministic local logical loops. AI is reshaping the toolchain, forcing us to transition from "code monkeys" to system architects who command the entire landscape.
Translation and code generation share a deeper structural property: both look like 'turn input X into output Y' from outside, but the value sits in the judgment about which Y is appropriate for this audience. AI is excellent at the surface transformation and indifferent to the judgment, and that's exactly the part clients quietly stopped paying for first.
AI right now is just the next high-level language -- C++ -> Python all over again. Every abstraction we've added traded understanding of the layer below for shipping speed, and it never killed the job, just moved the floor up.
The real shift isn't panic, it's generational: the devs who still know how the machine actually works will quietly age out, and the new ones will treat the whole stack as magic. Clarke called it -- sufficiently advanced tech is indistinguishable from magic. We're just pouring on another layer.
Coders panicking? We've been pretending to understand legacy code for years. AI is just another thing to Google.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.