DeepL is the Beginning of the End for the Translation Industry

DeepL is the Beginning of the End for the Translation Industry

deepl making translators obsolete

 

Translation is like a roll of toilet paper – most people don’t really think about it until they’re sitting there without it, wondering how awkward it’s going to be to yell to their hardly sympathetic girlfriend for assistance.

Fortunately for the toilet paper, it isn’t in any danger of losing its job any time soon.

Translation? Well…..

I spend a large amount of time contemplating translation, its history, its role in the modern world, and perhaps above all – its future. 1 <—Click me I’m special

I used to work in the translation industry. I worked with countless tireless wordsmiths all over the planet on a daily basis. Through LATG I’ve met countless more, heard their stories, and listened to their concerns and discussed the future.

There is one thing that annoys translators beyond everything else. Beyond deadlines, beyond obnoxious clients who think they know better, and even beyond an empty coffee pot.

No, above and beyond all, translators hate it when you suggest that their jobs might be at risk.

I know full well what sort of shit I’m about to stick my head into here, as I’m sure the comments will soon show, but the truth is that the industry is almost definitely on the cusp of trouble. 


The Tower of Babel

We’ve been coming up with origin stories for humanity since always. Talking snakes, interplanetary gods with airplane-shaped spaceships, the thing with the elephants on the turtle, etc.

But one of the zanier origin myths from the Old Testament – Genesis 11:1-9 to be more precise – explains how after the Great Flood, having clearly learned their lesson and committed to years of rapid, incestuous procreation, humanity bounces back and kind of comes together into one big happy family, 2 speaking one unified language, and generally being not shitty to one another. They go about building a pretty sweet city that features a super tall tower thing – the Tower of Babel.3

In swoops God.

He takes a look around and freaks out about how nicely everyone is getting along, and how successful they all are in their unity.

In one of the more dickish moves God pulls in the Old Testament he snaps his fingers and suddenly they all speak different languages and can’t work together anymore. Seems a little contradictory considering that he just drowned everyone for being naughty like 10 minutes ago4 but the point is that he was weirded out by this and thought they were up to something. Their tower was really big and scary, and being in his angsty period, and being the biggest kid on the playground, God said “fuck it” and dealt with the problem the normal way.

Since these ancient Babel people didn’t have Rosetta Stone or Google Translate, and apparently no interest in learning languages, they all go their own ways and do their own thing. They were very sad.

This results in the poor people of Babel being scattered to the four winds, going off to build their own less unified, not-as-cool cities. Super inconvenient.

This is where our linguistic diversity derives itself from – if you’re not into walruses, that is. Or science.5


Okay…what’s your point?

Translation brings the world together – Babel is a story about how it was torn apart. Also, I just like that story.

Translation is incredible. If you step back and think about it for a little while there is certainly something spectacularly wonderful about taking the thoughts, feelings, customs and more from one group of people, scrambling it all around into a bunch word salad and other sounds, and then putting it all back differently on paper like an artist to serve to a totally different group of people, who – assuming you’re any good at your job – will now understand not just the words, but the meaning behind them. Communication will be achieved!

In this way, translators and interpreters are linchpins in communication between peoples around the world. Whether the purposes are for business, pleasure, emergency services, politics, legal stuff, or actually literally anything else, translations bring people together, foster understanding, and play an unseen role in fueling economic growth, among other things.

I suppose in a way translators are humanity’s way of sticking it to God.

It’s also hard. As someone working in the industry I’m more than familiar with the difficulties that translators face – and painfully aware of how many incompetent translators there are out there.

Without going on too much of a tangent, it really riles me up when a random person, with no training, who maybe speaks two languages just wakes up one day and decides “hmm, how can I make a little bit more cash…” then goes about starting their own freelance profile on some cheap site that churns out sub-par drivel that for some reason businesses buy because they can’t tell the difference. We’ve all seen the sites that try to sell us clothes and shit with descriptions that were clearly written by someone with the language command, and the writing skills of a drunken capybara.

*heavy breathing*

In any case, I’ll stop gushing about how awesome translation and its masterminds are, how cool and meaningful their work is, and how much I really like them.

Because whether they or I like it or not, they’re still screwed.


Moore’s Law

You know how tech geeks are always talking about how computing capacity doubles every 18 months? You always think to yourself “that’s a weirdly specific number of months. Is this a real thing? Probably just some geek talking about geek stuff. Shut up geek…”

Turns out this is actually a thing – or at least it was, we’ll get to that. It’s called Moore’s law after Gordon Moore, the really smart, really rich businessman-scientist responsible for co-founding Intel. 6

Moore’s law remains tangible. You can look at your phone and compare it to the computational monstrosities we used 60 years ago. The entirety of the computing power needed to send people to the Moon is inferior to the computing power of your desktop or the mobile device you’re reading this on right now. Tech advancement is exponential, not linear, and it’s not going to slow down any time soon. 

Moore’s law has its issues though. At its most basic it is the observation of how the number of transistors in integrated circuits seems to double in a given period of time. This doesn’t mean that the computer you buy next year will be twice as good as the one you bought last week, though. Furthermore, Moore’s isn’t moving as quickly as it once was. The rate of computing capacity advancement is 100% reliant on the technical challenges in making stuff smaller and about a hundred other things that make it difficult.

Moore himself has said that there’s a good chance that it will cease to apply in the relatively near-ish future.

Regardless of whether Moore’s law may or may not become applicable to machine translation in the future, it’s a good way of underscoring how technology has improved the way it has over the past 75 years, and how it will undoubtedly continue to do so barring some global catastrophe.

The machine translation and artificial intelligence (AI) topic has certainly been beaten to death, including in at least two or three posts on this site, but my sentiments have, upon professional evidence, begun to take a turn for the negative.


Get ready for your robot overlords

Here I’m going to talk about the two things that those of us in the translation industry should start to worry about. We’ll start with the thing you should maybe worry a little bit about, and then follow it with the crazy shit you should maybe worry a lot about.

In direct contradiction to the words I’ve said in the past – perhaps in large part due to denial and wishful thinking – I do think the translation industry is in trouble. Not today, not tomorrow, and probably not next year, but it is nevertheless in trouble and things are going to get rough for translators sooner than we’d all probably like.

Google Translate has been the subject of ridicule from countless angles and people over its 12 year lifespan. We love to hate it thanks to its historically god-awful translations and the hysterically inaccurate results you can produce by inputting even the most basic of sentences. Don’t get me wrong, GT has its uses, and it can translate to and from about a hundred languages, so I guess that’s cool, but I wouldn’t stick my English medical records in it and hand them to my German doctor any time soon.

If your mission is to discover the Mongolian term for “fish cakes” it’s perfect, just don’t ask it to translate the recipe and expect gourmet dining. 7

While still not great, it has improved tremendously since its inception, and you can now get a better translation with it than ever before.

One of Google Translate’s biggest problems is that it’s a secondary concern for Internet titan Google. For a company with its hands in all things cool – including devastatingly powerful AI, self driving cars, fancy mobile tech, knowing what you had for lunch, and promoting lunar exploration – The Goog is kinda busy. GT is the least important thing Google executives have to think about on a daily basis, and it shows. You’d think a company with that much money, influence, and innovation could do a better job, but no.

Dr. Gereon Frahling would probably agree with me here. An ex-Google employee himself, Frahling decided it would be way cooler to go off and actually make translation software that doesn’t completely suck.

So he did. Sort of.

And this is one of the things translators should be a little nervous about.


The somewhat disconcerting thing – going off the DeepL end

In 2007 and 2008 Dr. Frahling and Leonard Fink came up with a cool new company called Linguee. Based out of Cologne, Germany, Linguee was unique in that rather than trying to transpose words and grammar from one language to another in the same fashion as its forebears, it instead set about collecting millions of translated sentences from all over the world and across many languages and tossed them together into a massive database of fully formed sentences.

It functioned very similarly to a super-cool dictionary, but rather than just getting the word you’re after, it gives you an entire sentence, providing a nice dose of grammar and some proper context along the way.

In and of itself Linguee isn’t really a translator so much as a dictionary of sentences, but that was just the beginning.

The coolness came to a head in August 2017 with the launching of Linguee’s translation AI. Named DeepL, this scary translation beast dips into Linguee’s now utterly massive database of sentences, word pairings, idioms, and other natural snippets of language, and instantly parses them to provide you with an exceptionally accurate translation.

It uses something called convolutional neural networks to provide something a little better than the average. It has some seriously badass semantic capabilities and actually adjusts the translation as you type and offers a deeper nuanced reply than GT has ever been able to manage. DeepL is a small quantum leap for translation technology, and you probably haven’t heard about it until now.

DeepL is still far from perfect, and you’ll still discover a number of grammatical fails, but it gets better all the time as the Linguee database expands. It’s not quite the same as legitimate artificial intelligence, it’s just a really good algorithm with its own crowd sourced database – not to mention a small, dedicated team of pretty smart people.

DeepL is, as of 2020, available in ten languages including English, Spanish, German, Polish, French, Japanese, Simplified Chinese (presumably Mandarin), Portuguese (European and Brazilian), Russian and Dutch. I highly suggest you check it out the next time you need to translate something.

It’s not currently a substitute for a professional translation from a human with thoughts and life experiences – or professional training in translation – but if you need to translate a sentence here and there for normal everyday purposes, or you need to get a solid gist of something you’re reading, it’s a pretty nifty tool. Go take a look!

Still, it’s good enough to give professional translation agencies the chills, and its sudden appearance on the machine translation scene has given us a gloomy glimpse into the future of the industry. It’s like those scenes in movies where the evil robot is lying dismantled on the floor of the laboratory in the dark, and suddenly one metallic, red eye flashes open and then they quickly cut to the next scene.

And just like the movies, this is just the beginning. The real problem we’re really worried about is coming, and it’s coming fast…


The scary thing – embrace the singularity!

This is the part where I worry a lot.

The super short explanation of the singularity is that it’s the point at which artificial intelligences attain cognitive functions on par with those of human beings and the ability to improve themselves autonomously. It’s considerably more complicated than that and I’m already going to have over 4,000 words in this post, and as much as I’d love to talk about it all day, you probably aren’t here for a lesson in robotics and super-human intelligence anyway.

It’s a contentious subject with scientists falling all over the spectrum in regards to when it will happen, and what will happen when it does, but most agree that it will happen.

The majority of scientists seem to fall somewhere in the middle between “this could be the best thing ever” and “this could be worst thing ever.” Some prominent names fall on the ends.

Drs. Ray Kurzweil and Michio Kaku, who are positively giddy about the possibilities of human-level AI and like to say fun things about the transcendence of the human race, immortality, world peace, galactic expansion, etc etc etc.

Equally brilliant dudes like Stephen Hawking and Elon Musk, on the other hand, are freaking out. They’re pretty confident that human-like AI with the ability to learn for itself and improve itself would be an extraordinarily slippery slope that could end humanity in the blink of an eye. Musk in particular is tweaking so bad that he went out and made his own brain-computer interfacing company to help merge humans with technology before the technology merges humans with being not alive. 8

I’m an optimist and a little bit of a transhumanist, and science and technology are my bread and butter, so as much as I like Hawking and Musk, I’m going to go with the other guys because these dudes are depressing and everything is shitty.

Anyhow, most experts believe that we’re going to experience this within the next few decades with some suggesting low estimates like the 2030s or 40s, with others suggesting that it will be more like the 70s.

But once again, I digress. 9

More to the point, the reason that artificial intelligences haven’t been very good at translating so far is because they currently lack the ability to truly learn anything about humans or our myriad languages that vary tremendously.

Thanks God, real helpful bro…

They are still effectively mindless and haven’t achieved the level of autonomy needed to actually understand stuff. They basically just detect the information they’re told to, gather it together, then spit it at you on command. They can’t think for themselves or interpret the meaning of those ones and zeroes.

Yet.

They will be able to one day, and it’s not a matter of if, it’s a matter of when. And when computers reach a point at which their data collection skills are advanced enough to actually pick up on the subtleties of human interactions, the translation industry will be tragically catapulted into irrelevance.

DeepL’s algorithms foreshadow this emerging capability. While still just a fancy algorithm that reads sentences in a database, it is massively outperforming its competition and should be giving us all pause. A year ago I thought this sort of thing was still way ahead of us, but since August I’ve come to realize that we’re not so far from human-like translation as we might have thought.


We’re all losing our jobs to automation

Despite the desperate claims of translators everywhere who insist that robots will never achieve true understanding and cognition, it’s just not a future-proof job industry. It’s probably going to happen. 

This is where I like to make the blacksmith argument. We don’t need an awful lot of guys with disproportionately huge right arms sitting around smoky shops pounding horseshoes and making knives all day. Our horseshoe demand is pretty low because most of us don’t ride horses anymore, and those who do have them made in factories. Same thing with knives. They’re laser cut and built from higher quality composite metals. Any blacksmiths reading this are about to go red in the face – but these things tend to be higher quality. It’s just more efficient, if somewhat less romantic and cool. We don’t have a lot of coopers anymore – because we don’t use an awful lot of barrels. 

While it makes me uncomfortable to equate translation to these now irrelevant or soon-to-be irrelevant industries, and I do find it rather awful and worrisome – since it also directly impacts my own job – I think it’s in for a similar fate.

And why wouldn’t it be? Instantaneous translation of a quality equal to or greater than that of an expert. Free, and readily available. Everyone who can will choose this route.

The machines aren’t going to have it easy, though

Not all translators will suffer this fate. As a commentator on Facebook was quick to point out – and he’s right, of course – languages are constantly evolving. This means that whatever AI we’re employing will need to keep pace with that change – likely made faster by the Internet or whatever futuristic mass communication system we’re using in 30 years. This will certainly prove a challenge for the AI, but if we’re capable of reaching the point at which computers reach human level intelligence and can improve themselves the way we’re fairly confident they will, I imagine the AI will be able to pull this off just fine. It would have no more difficult understanding the latest slang than you do, and it would probably learn about it first.

There are, however, many languages that don’t have dictionaries in the first place, let alone representation on the world’s largest translation systems. There will probably always be a need for translators who work with these languages for as long as they are still spoken. I’m talking about the industry as a whole though. Sure, there are still blacksmiths – just not very many.


It’s not just translation stuff, either

You know those cool universal translator things that they have in Star Trek or other sci-fi series? We’re not there yet, but we will be, and in the meantime it can feel at times like we’re wasting our time with languages. Why bother if we’re going to overcome them.

The technology is already emerging and it’s getting better every day. With voice recognition software like Siri and Alexa, with Skype and Google generating instant translations, this isn’t just some far off dream, it’s happening right now.

Learning languages will start to become less relevant to a lot of people, which, for me, is even sadder than the loss of the translation industry. It could be dragging down the interpreting industry, the language education industry, and countless other industries as well.

Now this issue is a little bit different, because we’re not just going to stop learning languages entirely, but it is worrisome and it gives me pause. This is another thing I think about a lot. How can I really reconcile the fact that I write this blog, talk about learning languages, am actively trying to study them, and am otherwise obsessed if I fundamentally believe that the world is headed in a very different direction.

There are plenty of benefits to learning a language outside the scope of professional needs. It’s good for your health, it’s cool, it helps you understand other cultures, etc, you know the rest. However, most people don’t learn languages for these reasons. They do it for business, or specifically for travel, or other things that they deem highly “practical.” They don’t do it for shits and giggles, and they’ll probably stop if they don’t have to. Unfortunate, but that’s how people are.



Conclusion

This whole mess has got me really uncomfortable and frustrated. I’m caught between two opposing ideologies that I really like: technology, science, robots, and super cool sci-fi stuff, and languages and their related industries.

In any case, if you’re a translator, interpreter, or you work in the industry, you’re probably thinking one of three things right now:

  1. You’re a moron, shut up, we’re fine, translation4eva!
  2. Oh, shit….
  3. Ain’t nobody got time for this, deadline is in 2 hours….wtf, where’s my coffee – oh ffs, is SDL updating Trados again?!?

Chill. You don’t have to dust off your CV just yet. You don’t have to switch majors to engineering. You don’t have to worry about hoarding your money like a madman for that early retirement you didn’t ask for. You’ll be fine for a while. We’re talking long term here. Not as long as we’d all probably like, but probably at least another couple of decades. C3P0 isn’t on the horizon, for now.

Don’t stop studying languages or innovating the industry. A day may come in which we transcend our modern conception of voice-to-ear-to-brain speech, instead merging our consciousnesses with a universal cloud, or communicating “telepathically” with Cyborg Elon Musk’s craziest new invention, but it isn’t this day, and that future event is even further away – if it ever happens at all. 10

As a species we are often tragically nearsighted, and these are things we need to consider for the long term. Tech such as voice recognition software, automatic, real-time translation AI, super accurate algorithms with huge databases like DeepL are just the beginning. It’s going to be a wild ride, but I think that we’ll come through it on top.

We always have before.

 

Like it? Don’t miss the next one!

* indicates required

Apex-editor of Languages Around the Globe, collector of linguists, regaler of history, accidental emmigrant, serial dork and English language mercenary and solutions fabricator. Potentially a necromancer. All typos are my own.

Tags: , ,