Should translators and interpreters fear AI, or embrace it?

This week, let’s talk about whether, as translators and interpreters, we should fear AI or embrace it, or some of both? Or something else! I have a lot of random and conflicting thoughts on these questions, but here’s an attempt at something coherent!

In the past couple of months, I’ve participated in a few webinars on AI, including demos of Chat GPT and Gaby-T. Wherever you stand on the “fear or embrace” spectrum, I think it’s really important to understand how AI tools work and what they can and cannot do, so that you know when they might be helpful. I think this is also true of machine translation: whether you find MT really helpful or totally useless in your own work, you need to know what it is and how it works, so that you can offer an intelligent response when clients ask you about it, and so that you can make an informed decision about whether you want to use it.

When it comes to the new generation of AI tools, I think we need to divide the question into two components:
-How useful, if at all, can these tools be to us?
-How might they affect the language professions and the language industry in years to come?

Those are two different questions: sort of like, what can these tools do for me, and what might these tools do to me? At this point, my sense is that:
-AI tools may end up being very helpful for translators and interpreters, and particularly for interpreters
-AI tools may also radically change how the language industry operates, particularly for translators who want to work with agencies

I’ve honestly been really impressed with the demos of AI for interpreting prep: being able to tell an AI tool, “Take this document, extract the keywords, translate them into French, and make a two-column list of the French and English keywords,” is a huge time-saver, if (biiiigggg “if”) you know enough about the subject matter to check the AI tool’s work. Having an AI tool that could reliably display place names, dates, people’s names, book titles, etc. during a speech (which is a feature that’s being incorporated into various AI tools) would be hugely helpful for conference interpreting. And I personally find it so much easier to write e-mails in French in Google Workspace now that I can quickly draft an e-mail on my US keyboard, without any accent marks, then go back and accept or reject Google’s suggested changes to the French text. I am personally really enthusiastic and excited about these kinds of AI innovations.

I also think that automated interpreting is such a huge challenge that it’s not worth stressing out about it for the moment.

I’m less excited about how I think AI may affect certain segments of the translation market (emphasis here on “think,” “may,” and “certain segments”). This is because:

• There’s already a pretty big market for “good enough” translations: translations that are neither great nor terrible. And if a client’s main concerns are price and speed, with quality in third (or distant third) place, AI-assisted translation plays right into that market. For an end client that really doesn’t care about translation, and is only doing it because they have to, it’s appealing to get the translation done faster and cheaper (more on this below).
• Lots of agencies seem to be investing heavily in AI-leveraged translations. Companies such as Lionbridge, RWS, and TransPerfect (and many others) have significant website real estate devoted to AI-powered options for their translation clients.
• Lots of large end clients have more and more content, and tighter and tighter budgets; to be realistic, lots of large end clients like pharmaceutical and technology companies are legally obligated to produce huge amounts of translated content (software end-user license agreements, patient information sheets) that almost no end users or patients ever read. These companies may be exposing themselves to legal risks by procuring ‘good enough’ translations, but they may know that and do it anyway, for cost reasons.
• I think that AI may make significant inroads into the middle of the translation market, which is where most translators are working. You just don’t see that many translators saying, “I leverage every possible piece of technology and I translate 5,000 good-quality words a day,” and you don’t see that many translators who are really, really successful in the high-end direct client market. It seems to me like most mid-market agencies are looking at AI-powered translations in at least some way, because the market demands it.

What would I recommend in terms of business strategy, given the above?

• First, a few caveats. I may be way off base here. I don’t use AI-powered tools myself, I don’t generally work for mid-market agencies, and most of my active marketing energy these days goes into conference interpreting, for clients who use no AI or MT tools at all. My translation direct clients are much more likely to ask, “You don’t use any machine translation tools, do you??” Also, even super-smart, super tech-savvy people get the technology trajectory wrong, for all kinds of reasons. Google co-founder Sergei Brin famously said in 2012 that fully autonomous, self-driving cars would be “a reality for ordinary people in less than five years,” and 12 years later, that whole concept seems to have lost a lot of mojo. The same thing could happen in our industry.
• That being said, I would encourage anyone reading this to think about whether or not you’re interested in working on AI-assisted translations. I honestly think that there’s enough high-end translation work, for clients who really care, to keep a lot of people busy for a long time. I actually think that the high end of the translation market is kind of underexploited by freelancers (a topic for another post!). And/but, the reality is that most translators aim for the path of least resistance: mid-market agencies, which, I think, may go the AI-assisted route in a big way.
• I would also encourage anyone reading this to think of diversifying into something that is not so prone to AI incursion. I think lots of translators these days are branching out into specialized writing: not churning out blog posts for $10, but things like writing specialized reports that require a writer with a certain type of certification, or ghost-writing articles and presentations for C-level business executives, or doing medical or legal writing for clients who see $150-$200 an hour as a pretty reasonable rate. For what it’s worth, one of my favorite websites on this topic is Make a Living Writing, which covers all kinds of topics from choosing a writing specialization, to writing technique, to writing business practices. Also consider translation specializations (like official documents, genealogical records, etc.) that will never be automated because the actual files are too hard for a computer to read.
• The same advice you’ve been hearing forever still applies: aiming for the high end of the market, where clients care much more about quality than price, is the way to go in any freelance knowledge profession, because there is no way that AI can compete with human creativity and linguistic resourcefulness. See French to English translator Ben Karl’s comparison of his translation and various AI translations as an example.

As a final observation, and in line with the final bullet point about freelance writing, one thing I see coming as a result of the AI boom is a growing income disparity among translators. Again, I see interpreting as a whole different case here, because it’s so much harder to automate; I have thoughts on the future of interpreting, but those are for another post. To me, it seems to have become more common to see freelance writers talking either about writing for content mills and making (literally) $8 an article, or about landing $15,000 worth of freelance writing work in one week, with not a lot of “mid-market” work left. In fact, I found it really interesting that in Make a Living Writing’s 2021 reader income survey, the percentage of their readers (10%) who reported earnings of 50-75K (which I would call a “mid-market” income) was only one percentage point higher than the 9% of readers who reported that they earned more than 100K. I’m wondering (and again, this may be completely incorrect) whether translation is going in much the same direction.