Today, I have a little story for you. It’s got a very catchy title – you’re going to love it. Ladies and gentlemen, I present to you:
A Brief History of my Attitude Towards Machine Translation and AI Generally: 2015 to Present
Are you sitting comfortably? Then let’s begin.
2015: I enter translation industry as freelance legal translator for the language combination DE>EN. Am aware that machine translation is a thing, that it is already well-developed and that it is going to seriously shake up the industry at some point. But right now, I’m far more occupied with getting clients on board, getting to grips with CAT tools – not to mention thoroughly pedestrian things like learning how to issue proper invoices and creating a website.
2017-2018: I keep being told that machine translation and AI are going to take my job away, but with DeepL (my MT tool of choice) still nowhere near being able to properly translate legal texts, I’m still feeling reasonably secure. I know I should be diversifying to spread my business risk, but with so much work coming in I don’t really apply my mind to it.
Apparently, a huge, “very lucrative” market for post-editing services is going to open up for us translators. All our customers will know immediately what post-editing is and that translators will be offering it. We will all become madly productive and efficient and live happily ever after. I’m not sure if I like the idea of doing post-editing that much, but carry on regardless.
2019: Gosh, DeepL is very good these days! Yes, time to acknowledge that it can now safely be used for certain legal translation projects. Also time to fully face up to the fact that technology might not just change, but completely gobble up my job – in the not-too-distant future.
I add MT and post-editing to my range of services, do the TAUS certification in post-editing, practice my light-/full-post editing skills and generally gear up to offer my clients this new service when the opportunity arises. One does, after all, have to move with the times.
2020-2022: PANDEMIC MADNESS.
During this period, I am no longer thinking about how machine translation is developing or what AI might do to the translation industry. The most immediate threat to my business is the fact that the entire economy has been put into deep freeze due to lockdowns. Am now far more occupied with thoughts like: am I allowed to use the public transport/go to the supermarket? When can I go back to the UK again to see my family? Why can’t we have a conversation about herd immunity or whether Sweden might actually be doing the right thing without everyone collectively freaking out? And, most importantly: am I going to go completely insane in this garden-less, balcony-less flat during lockdown?
One of the ways in which I avoid going lockdown-loopy is to dive into a new area of interest, SEO. I start to optimise the Spezialis Translations website to put my new skills into action and see what happens. (Spoiler: several top rankings in the Google search results for important keywords.)
2023: Finally realise that AI is going to be more significant and more far-reaching than I ever could have imagined. Start experimenting with the new toys in town: ChatGPT and DeepL Write. Am now good at SEO and decide to set up a business to offer it as a service – partly because I love it, but also because I clearly need to diversify away from a translation business fast being eaten up by MT/AI. Am impressed with the new opportunities AI brings but am also faintly terrified at the ramifications. Pretty much like anyone else who isn’t living under a rock in the Mongolian steppe, then.
Now what?
So, here I am in 2023 with an existing legal translation business and a fledgling SEO business. Both branches are currently being completely turned on their heads by AI. And no one can say with any certainty where this is all going or what it will mean for us all long term. The best we mere mortals can do is try and comprehend the situation as best we can and hash out a way forward which appears sensible. Not easy: we are being called upon to make big decisions based on the shifting sands of a fast-developing situation. And yet standing still and doing nothing is not an option.
To be honest, I had been a bit slow on the uptake with ChatGPT. And it hasn’t been the best tool to help me with the work I’ve been doing in the last six months either: DeepL Write has been my constant companion. But to get left behind with AI will not do, it’s just too huge an issue. Very few things are so groundbreaking that they spawn whole new disciplines almost overnight. And yet that is what has happened – see Prompt Engineering. No, now is not the time to look the other way.
You can make a lot of money if you keep up with this and acquire the right skills. On the other hand: fail to keep up and stick rigidly to old ways of doing things and you’ll become obsolete, as AI will very quickly take over the work you previously lived off.
So many opportunities! The question is: which ones to take? What does it all mean for translators from a practical point of view?
Time out
In order to answer this question for myself, I decided to take some time out to think, reflect and get the bigger picture. I needed to “sit in my thoughts” for a while. Let them percolate, associate, dissociate, coalesce…until I’d made sense of things. Then I’d have some kind of framework for decision-making going forward rather than simply reacting to events.
After zoning out and communing with the contents of my mind, these are the conclusions I have come to as regards the translation industry:
1. There is going to be an ENORMOUS shakeout.
This sorting out is going to be brutal – and it has already been underway for some time. AI is already performing a sizeable chunk of work that used to be done by humans – and that chunk is only going to get larger as the technology improves. We are going to see a lot of translation businesses going under in the coming months and years.
I think there are many freelancers out there who still have not comprehended the gravity of the situation and that they are about to be mown down unceremoniously by the AI juggernaut. Or they’re in active denial. Whatever. There is still a lot of shock and a lot of upset and a lot of disappointment to come.
I think it was around 2019 when I realised that my time as a freelance translator has coincided with a kind of “end of times” for the job. It took another year or two to fully accept that prospect and take the steps necessary to move away from the industry towards something more “future fit”. There is a measure of grief attached to letting go of a business you built up, worked hard on, gave yourself to, and loved. It takes time to work through.
2. The market for post-editing for the artists formerly known as translators isn’t going to be anywhere near as big as we thought it was going to be.
Even if clients are aware that there is such a thing as post-editing, then it is by no means certain that those tasks will be given to translators. To get that work, freelancers are going to have to go on the offensive – actively informing their clients that they offer this service, what tool they use, how they maintain confidentiality, what savings are made etc.
Having super-duper modern translation software with the latest MT plugin is no longer a selling point if the client no longer needs translation services because they are generating texts in the desired target language using ChatGPT or translating texts themselves with MT.
3. At some point, the pendulum will swing back (slightly) in favour of translators/post-editors, but it will never be the same.
Disclaimer: I don’t have any data to back this up, so the following is all based on assumption, conjecture and personal observation. Sorry about that. But as far as I can see from the ground, the reduction in the amount of translation work available has reduced dramatically, even in the last 12 months. This could be to do with long-tail pandemic and lockdown effects, but I think it more likely that AI is the driver. Clients have already switched to DeepL Translate or ChatGPT to perform tasks which would previously have been outsourced to translators. As I mentioned above, translators can try and “claw-back” some work by going on a post-editing awareness spree.
However, if you are determined to stick around in the industry, it might be worth standing back and waiting out the current wave of AI enthusiasm a bit. (Hopefully you’ve been smart and built up some financial reserves to get you through lean times like this. If you haven’t…good luck with that job search.)
Just like everyone else, people and companies who previously ordered translations are engaged in a process of experimenting and understanding what AI can do for them and what it can’t. New boundaries and processes are being established. And that takes time.
That process will inevitably involve some level of incident as people invest too much trust and expectation in tools that either have not reached maturity, or which haven’t developed in the way expected, or which go rogue. In other words, plenty of users will learn the hard way that blindly relying on MT/AI output in all cases is a bad plan. Facts need checking, machine-translated texts need careful reading and editing. That takes time, effort and critical thinking. And that’s where language providers can swoop back in to save the day and ply their services.
However: the amount of work which will become available as part of this “backwash” will be very small compared to what we knew before. AI is already good, and it will get much, much better. Only highly complex, legal or technical texts will need a thorough review after machine translation. Therefore, only few, highly specialised translators will still be in work after this revolution. The rest had better start reskilling and looking for other things to do.
When I wrote this article about the advice I would give to young freelancers now entering the market, I said I would not advise them to forget about freelancé translation and do something else. I’ve rowed back on that opinion already. That wasn’t my inner cynic speaking – it was my inner realist. Unless these rookies are already super-specialised in a certain field or offering some kind of offbeat language combination where there is plenty of demand but very little supply/MT capability for the foreseeable future, entering the AI-driven translation industry as a freelancer at this point would be like diving head first into a swimming pool with only 5cm of water left at the bottom. Splat.
4. Some areas of translation might even see a rise in their status or a renaissance.
So far, I have been pretty darned negative about the future of the translation industry. There is no point in “sitting in your own thoughts” unless you are prepared to peer into the dark corners in order to draw your conclusions.
But I’m not all doom and gloom. While I think that the times when I could earn my living entirely off legal translations DE>EN are well and truly over, some translators might find themselves being left high and dry by the AI revolution. A bit like the animals that managed to get on Noah’s ark before the biblical floods arrived.
To risk stating the absolutely obvious: AI may be able to deliver (more or less accurate) knowledge, formulate texts which communicate that knowledge, provide a clean solution for a coding problem, or translate a contract. What it will never, ever do is…be human.
There will never be a substitute for true personal interaction. Humans will always want to hear from other, real humans about their real, human experiences. That’s why Reddit is so popular and why there have been “blackouts” over there as contributors seek to protect their fundamental asset – authentic, human-produced content – from AI-generated tat. In the AI world, there will still be more than enough space for well-crafted and moving stories written by real people which teach, inspire and move other real people.
For this reason, I think that literary translators are going to be reasonably safe. At this point, I do not believe that a machine will ever be able to comprehend a literary text, its historical/political/social context, or the relevant linguistic/dialect nuances in the text to the extent necessary to create a high-quality literary translation. Nor do machines have the emotional intelligence required to reflect upon and choose the words in the target language which will move the reader in the way the author intended.
But I could be wrong. Things change so fast.
—–
Other articles on Spezialis Translations blog:
A little anthology of iffy translations
Looking back on 8 years of freelance legal translation
UK vs. US English: the nitty gritty legal translation edition
Further reading:
“AI is killing the old web and the new web struggles to be born” by James Vincent on The Verge
—–
Photo: MEFTAHYs-PROTOTYPE on Envato Elements