From the Desk of Steven S. Bammel, PhD
October 15, 2024
RE: Let’s Write a New Narrative for Human Translation
ATTN: Translator Colleague
These are the lies.
First, they tell you a new technology will help you translate faster, that you can become more productive and make more money.
Second, they automate with that same technology. Before long, they’ve dumbed down your work.
Finally, they cut your rates or replace you.
Technology innovation in the translation industry has repeated this cycle relentlessly for as long as I can remember.
If you’ve been a translator for any time at all, you’ve noticed the same thing, haven’t you?
Do you think their game plan has changed now in the era of ChatGPT and other Generative AI models?
Of course not!
I bet you’re cynical about AI in translation, maybe even scared, angry, tired… You name it…
I don’t blame you for feeling that way.
I’m a translator just like you and I’ve had—and still have!—these emotions myself.
In fact, can I tell you something I’m especially tired of?
I’m especially tired of being told by people from the big LSPs, MT providers, and CAT/TMS tool makers how I’m supposed to just adapt to the new reality and get in line with their vision for the future… (Can anybody say “Pied Piper”?)
What’s their vision anyway? Might it be to eliminate competent linguists from as much of the translation process as possible?
Um, yes, of course.
Do you agree with me?
Let’s discuss machine-translation post-editing (MTPE) for a moment.
Are we of one mind about the mind-numbing grind and insulting mediocrity of MTPE workflows designed to suck the life out of the translation process and feed it back into lower prices?
How about that MT output they give you to work with that’s supposedly being trained to get better over time? Do they then tell you to revise that to “human translation equivalence” with a “full post-editing” process?
Might as well ask you to bend steel with your bare hands.
Am I wrong?
Or how about this?
They offer a low hourly rate, and tell you how many words you have to process within their pre-set budget. Boost your efficiency and finish early: they pocket the difference. Get stuck with awful MT that takes too long to fix: you take the hit.
I wanted to say that this work arrangement was designed by the devil himself.
But I decided that could offend some readers.
Aren’t you glad I didn’t say that?
So I’ll just point out that this insidious work arrangement is the most effective way I can think of to sucker punch translator morale, normalize the minimum common denominator, popularize machine speak into the culture, and drive the best linguistic talent out of the market.
Can we agree that MTPE wasn’t designed by or for translators?
For that matter, are you inspired when they tell you that their new AI tool will do all the “hard, mundane work” for you so you can focus on the advanced “human tasks”?
I didn’t think so.
How about all that lip-service to “human in the loop” workflows? What does that really mean to them anyway?
Even though they don’t generally engage in genuine dialogue with translators, we can still surmise their intentions from what they say and do.
BTW, do you like having your job title rewritten into terms that could describe just about anything? Now we’re “language professionals”, huh? So I guess accountants are just “numbers professionals” and actors are “impersonation professionals”?
We didn’t get into this profession to carry around a job title so amorphous it can be molded into a million different shapes.
As translators, it’s our job to be precise.
I’m a translator and until I stop translating, I’m not accepting another job title.
Am I preaching to the choir?
I absolutely do not believe that blind adoption of any of the newfangled AI features in the mainstream tools will get us anywhere close to future-proofing our careers as translators.
There’s no future in just following along, taking the work as it comes, doing it by the book, regardless of where it leads.
Then is there even a future AT ALL for translators in a world of AI?
I’m an AI optimist.
I’m not all gloom and doom. In fact, I’m the opposite.
What if I told you that the current Generative AI revolution that started with ChatGPT is a brand new opportunity? That we can turn the tables this time? That Generative AI changes everything for translators?
Even if you’re skeptical, do I at least have your attention?
I believe that with translator-centric tools and powerful AI workflows based on a revolutionary concept I call “expert-directed micro iteration,” we, as translators, can run circles around the big LSPs’ centralized and standardized workflows.
Of course, we have to work in the common tools of our trade.
And I certainly don’t mean that we can compete effectively against the big LSPs in the end-client market with a narrow translation service offering.
But we can do better; we can upskill in ways that the machines cannot replicate. We can often do so within our clients’ tools and workflows, plugging in our own tools and workflows to energize our position in the market.
Will the machines ever catch up? I have no idea; in the long run, I guess. But as John Maynard Keynes said, “In the long run we are all dead.”
In the meantime, we have a responsibility to ourselves, to our clients, to our families, to our profession (and probably to a few others I’ve forgotten in this list) to muster up the courage and march into the future with fingers on the trigger and our eyes wide open.
Are you up for a fight you can win?
Generative AI empowers translators.
Let’s consider four benefits for translators from Generative AI.
Generative AI Benefit #1 for Translators
Of course, Generative AI can enhance your translation productivity.
Whether you believe me at this point is up to you.
But I’ve seen the evidence… in the research, in my own work, and in the work of others. I don’t think there’s a valid argument to the contrary.
With the right techniques, Generative AI is too powerful not to give you translation superpowers. You can deliver better work than ever, faster.
If Generative AI can make you more productive, you’d have to find out how to use it, right?
But even if this were the only benefit of Generative AI, we’d still be trapped in the technology vortex I described above.
Fortunately, this is not the only benefit.
Generative AI Benefit #2 for Translators
Because Generative AI is so accessible AND because linguists are the ONLY competent operators at the intersection of technology and language in this new paradigm, we have an opportunity to expand our role in the translation process for the first time in decades.
How would you like to add more value in your work, not less?
In fact, this opportunity comes to us just as three major paradigms of our industry are being crushed by Generative AI.
- Conventional neural machine translation is dying.
- It may take CAT tools years to fully adapt to Generative AI.
- And translation memory—a pillar of our industry for at least the last 25 years—has become a depreciating asset.
No wonder many of the big LSPs, MT providers, and CAT/TMS makers are seeing their stock prices crater!
Generative AI Benefit #3 for Translators
Generative AI gives translators our voice back.
No longer are we slaves to poorly trained machine translation in soul-crushing workflows.
With the Generative AI Iterative Translation workflow, we can make the AI do our bidding. We can command the AI to translate in our voice, in our style, with our words…
Translating in your own words is what you always wanted to do anyway, right?
And we can deliver quality at the top of our skillset, doing so more efficiently than ever.
Generative AI Benefit #4 for Translators
Finally, these three benefits lead to ONE AWE-INSPIRING CONCLUSION:
As translators learn to deliver better work faster than ever, we can unlock vast volumes of content for human-directed translation that were not previously cost-viable at quality.
And because the “big guys” are locked into their outdated paradigms and workflows, translators able to deliver at these super- human levels can look forward to a golden age of translation!
.
.
Uh-oh… Did I lose you?
Did I just hear you say, “Hold on, Steven… A golden age? Huh?… You’re full of sh…hype! Are you telling me that Generative AI can let me go back to doing everything I love about being a translator?”
In short, yes, I am…
But first…
Let’s deflate the hype.
I’d like to set some realistic expectations on how Generative AI can impact your work.
How would I know?
I’m a practicing translator (unlike a lot of the prognosticators out there talking down to translators).
I do 20-30 hours of translation a week from Korean to English. And I’ve been applying Generative AI to my work on hundreds of projects, large and small.
We’re hearing all over the place that AI is going to 10x office productivity. So if I was averaging around 4,000 words a day before Generative AI, I should be aiming for 40,000 words a day soon, right?
Hardly.
Of course, every job is different. But on average, I estimate that I’ve boosted my productivity by…
Wait for it…
3000%!!!
300%!!
30%!
That’s right. 30%.
This isn’t hype, but it’s not a number to scoff at either.
A 30% productivity increase could easily represent a $20-30K boost in annual earnings! (And for a full-time, all-engines-running translator, it’s likely even more.)
To repeat myself to avoid any possibility of confusion: I‘m not promising you that you’ll 10X your productivity with Generative AI.
At least, not in the immediate future, and not if your goal is to deliver your best work.
Which brings up…
The quality issue remains a challenge.
One of my favorite commentators on the impact of Generative AI is Gary Marcus.
Gary talks about the impact of Generative AI on quality. Here’s something he mentioned recently: Studies are showing downward pressure by Generative AI on coding quality.
Translation isn’t coding. But I think it’s safe to say that the same general dynamic is at work in our industry.
Again, I have the advantage of translating daily at my desk, seeing this stuff play out, and dealing with the aftermath.
More often than ever before, I’m being asked to revise “human translations” that turn out to be anything but that.
Has this happened to you? Have you seen the quality erosion?
If you’ve been working as hard as I’ve been working, I’m sure you’re witnessing the same thing.
But wait…
Didn’t I go on and on above about how Generative AI can make you a better translator, both in terms of efficiency AND QUALITY?
Yes, I did.
But read carefully: Just because Generative AI CAN make someone a better translator, doesn’t mean that it WILL.
Generative AI is an enabler of laziness in translation.
And this is exacerbated by ignorance in our profession about how to even master this awesome technology.
Unfortunately, without the right sense of professionalism and work methodology, the weaknesses of legacy neural machine translation can apply to Generative AI output as well.
Lazy work yields bad translations, regardless of the paradigm.
It’s easier with Generative AI to pass off bad work as good work.
This makes it more difficult than ever for top-notch translators to differentiate themselves in the market and earn what they deserve.
The need for good work hasn’t changed. But the means of getting there and promoting it have.
The economics of the industry have changed.
I learned in technology innovation class during graduate school that occupations like ours are “supplier dominated”. This means we are at the mercy of those who develop and own the technology.
Together with automation, technology innovation has commoditized our work. And this in turn has contributed to translation (aka “localization”) industry consolidation.
The result is less bargaining power for translators.
And for small LSPs.
In many ways, translators and small LSPs are in the same boat.
Therefore, we should avoid categorizing all LSPs as part of a monolithic, negative group.
Further, getting direct translation-only clients isn’t the answer for most of us either, as the above trends have made this harder as well.
I’ll hint at another way out later, but for now, the only way forward is to prosper within the dominant agency-subcontracting model.
I’ll also mention here that per-word rates are going to drop, and possibly drop fast.
Be ready for it.
Ignoring the inevitable by going on about how doctors don’t drop their rates just because they have fancy technology is not helpful.
The way to fight back is to increase productivity dramatically.
Defend your bottom line and resist hourly rates at all costs (we’re not attorneys!) by focusing on upskilling and volume growth.
It’s not going to be easy.
Even with the new advantages we enjoy through Generative AI, we still have our work cut out for us. We have mountains to climb and battles to fight.
And I have to be honest with you. Nobody has completely figured this all out yet. Everything is new and the landscape is changing at a ridiculously fast pace.
I’m on this journey too, interacting with other like-minded translators, applying new techniques, learning new things every day, and documenting as much of it as I can.
They say it takes 10,000 hours to become an expert in a field.
It’s only been about 15,000 hours since ChatGPT dropped.
Alas, I have not committed every single waking minute since late 2022 to learning about Generative AI.
So I’m not an expert.
But I am a seeker, and I’m on this journey with you. I have been working and investing hard; some would say I’ve been obsessed.
And together, our community has documented a lot of hard-hitting insights that can power up your workflow.
If you could tap into the current value we’ve unearthed, and be around for what our community discovers going forward, would you want to be part of that?
Here are the elements of what we’ve built so far.
Element #1 – Software
My team and I have created a tool called CotranslatorAI, which is designed to empower translators to reach our potential in Generative AI.
Is the tool finished? Is it perfect? No, of course not. We’re still adding features (with a major update scheduled for Q4 of 2024).
But CotranslatorAI is unique, innovative, and powerful. It is already delivering value to hundreds of users.
And it’s available in the Starter edition as a free download.
Would you use a tool that helps you deliver more value? Check it out at CotranslatorAI.com.
Element #2 – Training
I’ve now published many hours of in-depth training for translators about exploiting Generative AI.
The core of this content is the Micro Iterative AI Frameworks. This training provides you with frameworks to help you bridge the gap between knowing about AI and effectively integrating AI into your daily work as a translator.
I’ve also just released a beginner-level course, AI Gameplan for Translators.
This is not the shallow “Hey, here’s another cool thing AI can do” type of training. Of course, it’s got that too. But this is focused, intense, full-concentration content that will take you deeper than anything else out there.
Also, once you’ve registered, you continue to have access to the same course.
And we keep improving the materials.
This means that as we learn new techniques and as the technology evolves, you can look forward to improving content without further purchases.
You can even ask questions and get support as you study.
Are you interested in learning how to apply Generative AI directly to your workflow? Find out more at CotranslatorAI.com.
Element #3 – Community
We’ve created AI4LanguageProfessionals, an online community for translators (and yes, others in the community, so “language professionals” is a broad term that applies) to discuss and stay connected to everything about AI in translation. The discussions there can get pretty deep, and it’s a great place for translators to be right now.
You’d like to be among the first to find out about new ideas, new use cases, new prompts, and new models, wouldn’t you?
Sign up for AI4LanguageProfessionals at CotranslatorAI.com.
We have a transformative vision for translators.
So what are your choices in the face of everything that’s going on in our profession?
Here are some scenarios.
First, you can just keep doing what you’re doing.
The big LSPs, the MT providers, the CAT tool makers, and other industry players will keep chipping away. Your earnings will continue to fall, and you’ll have to make do.
Second, in this process, you might lose heart and quit the business.
Some are already doing that.
Third, you can upskill and overcome.
Remember what I said earlier about using expert-directed micro iteration to one-up anything the big players are doing with their algorithms?
That’s your ace in the hole, your secret weapon, and the key advantage you can rely on when the odds seem stacked against you.
Remember, while the industry “illuminati” may have their machinations, your individual ingenuity and adaptability are powerful assets that can help you navigate challenges and seize opportunities.
And what comes next?
Here’s one potential vision of the future to consider:
Generative AI provides adjacent market access. Thanks to the direct line we have into the technology, we can seize opportunities for bundled language services.
As a result, translators can begin transforming themselves into “multi-lingual content operations specialists.”
Here’s what this will achieve:
- Provides a comprehensive, affordable, non-metered AI- powered service suite to SMEs without in-house resources
- Leverages the skills translators have developed over the years, overcoming past commoditization trends
- Dramatically grows a market for language services that individuals are suited to fill Have I clarified what all this means yet? Am I sure this is how it’ll all play out?
Of course not. Even Bill Gates recently said he was wrong about how he thought computers would be used in the future.
But one thing we are sure of: We’re not going to succeed by sitting on our hands. There is no action plan in that.
If we do this right, I believe we can get out front and transform our profession into one that adds value to global communication in ways that we have yet to fathom.
Let’s join forces!
We’re working to build a movement that will guide our tribe of translators into the future.
If my message resonates with you,
If you’d like to figure this Generative AI thing out in your work,
If you’re keen to add new capabilities to your skillset, and
If you want to do so while also delivering better work than ever,
Then I invite you to join me on this journey.
Are you ready to begin?
I hope to see you on the other side!
Steven
P.S. No AI was used in the writing of this message; I had too much soul to bare to rely on a machine!