Is GenAI the New Rosetta Stone?

Early in my career, long before I started working at Kaplan International, I spent a number of years as a professional translator. I had the advantage of growing up as a native English speaker who also spoke Catalan at home, then later lived in Barcelona for five years. This put me in the unusual position of being able to translate everything from culturally specific film subtitles to specialized world music catalogs from English into Catalan and Spanish - and vice versa.

For years, I suffered from a high-standards syndrome so acute it led to a kind of perfectionism paralysis. Whether I was reading garbled instructions for an appliance or seeing cringe-worthy subtitles when I watched films at the cinema, it drove me up the wall. Should I actually translate the garble and faithfully represent the cringe? Or should I dare to improve on the original?

Many years later, Google Translate showed me just how far computing was from being trustworthy or reliable in this area. It could do something very basic (single words and very simple sentences), but you could not entrust it with the translation of texts - not even to an amateur level.

Yet the advent of ChatGPT is changing this field along with so many others. Even literary translation, which requires uniquely human inputs (creativity, cultural and literary knowledge, superb command of linguistic nuance), can begin with an AI draft. Many of those I’ve seen in recent months have been, frankly, remarkably good. With a bit of human oversight, they can be successively improved in a positive feedback loop until a final draft is achieved in a fraction of the time it would take to do it manually.

AI-Assisted Translation: Inevitable

This realization was sobering for a lover of language. It will surely have far-reaching implications for the business model of the translating profession as the willingness of clients and publishers to pay for high-quality translations diminishes. Some of my own friends and even family members who work as translators aren’t yet convinced of this new reality, but it is plainly inevitable. And it is coming quickly.

That said, the nature of game-changing innovations like generative AI is both cataclysmic and very hard to predict. Just as the rise of pocket calculators didn’t mean the end of rigorous quantitative analysis, the ability of AI to offer sophisticated translation assistance could lead to a range of outcomes, whether positive, negative, or just different.

The first is that a revolution in fast, accurate AI-assisted translation, combined with ever-improving portable digital technology, could dramatically reduce the incentive for many individuals to learn foreign languages. If a Portuguese business executive can travel to a client meeting in Seoul, with all participants equipped with earpieces that provide clear and instant translations of everything they hear and say, what incentive do any of the parties involved have to hire a translator, learn one another’s languages, or even get by with varying levels of English as a lingua franca? This brave new world isn’t here yet, but it may not be far off.

At the same time, I can imagine a scenario in which the possibilities of technology breathe new life into treasured but minority or even endangered regional languages, from Provençal to Gaelic, Pirahã to Garhwali. With the juggernaut of Global English no longer bearing down with quite so much urgency, exploring local languages with valuable historical, literary, and cultural meanings could have much greater appeal. Mastering English, often at the expense of less dominant languages, will no longer be the first order of business for anyone hoping to expand their horizons with knowledge and people from other countries, whether by reading, traveling, socializing, or doing business.

AI In Language Education: Personalization & Marking

Beyond language, the implications for just about every kind of learning are, of course, immense. One example is personalization. Much has been written about this, but in essence, it’s about tailoring learning to the specific needs of individual students. Imagine an AI bot that is fully familiar with a student’s work, personal strengths, and weaknesses and has an “understanding” of their potential. It could help a student who was bad at structure but great at style to think about ways of organizing their writing, or it could help a student who had plenty of story ideas but struggled with style to develop their own voice.

Another example is marking. Marking is often a repetitive, thankless task for many instructors. AI marking has improved vastly in recent years, and a well-designed, properly supervised marking bot can free up teachers to spend more classroom time with students or support students personally while the bot does the donkey work. It’s a win-win for both teacher and student.

Those of us in language education are already exploring ways in which these AI-assisted translation can be fruitfully applied to enhance the student experience and gain efficiencies at the same time. Additionally, the next era of AI will be led by those organizations that apply the technologies in ways that make them resilient, dynamic, and competitive.

My personal sense of what the future holds is that rather than panic about AI-assisted translation or get carried away with optimism. We are very much in need of a classic SWOT analysis - examining strengths, weaknesses, opportunities, and threats. In my own much-loved field of translation, I’m trying hard not to be a fossil who risks being a relic of the past. Instead, I think the key for all of us will be to stay level-headed about living with uncertainty but also to embrace the possibilities for expanding knowledge that lies ahead.