My Future as a Translator

Several years ago, before most of us were talking about AI, I made a decision.

In my first year as a translator, I had the opportunity to work for an online dictionary website. Over the next several months, I used my linguistic skills to help edit Spanish <> English dictionary entries. I was especially excited about the project because I had learned about the website while getting my master’s degree in Translation Studies, and it had quickly become one of the most helpful tools in my everyday translation work. It was an interesting project that paid well, and working on it made me feel like I was making a positive contribution to my new industry.

Several years later, the same company contacted me again to work on a new dictionary project. I eagerly accepted based on the great experience I’d had in the past, but during the initial trial contract it became clear to me that the work had a new purpose: to improve machine translation output. I probably could have worked on this project for months, if not years, and made quite a bit of money. Yet in the end, I chose not to continue working on the project. I explained to the project manager that while I appreciated the opportunity, I felt it was unethical to work on a tool designed to replace translators. It was the first big ethical decision I had to make in my career, but looking back, I feel confident that I made the right decision.

I have been extremely fortunate to have steady work throughout my 12-year freelance career. I now find myself in the midst of my first big lull in work as the industry is seemingly overrun by inferior AI translations, some of which are being generated by that same company. I was always able to find small translation, editing or copywriting projects to fill the gaps in the past, but those projects also seem to have largely disappeared.

As far as I can tell, the vast majority of opportunities available to professional translators at the moment involve refining subpar AI output, often in huge volumes for astonishingly low rates. Most of the advice from prominent voices in the translation industry seems to be to embrace AI by either using it ourselves or accepting post-editing or validation jobs.

Perhaps I am an outlier, but I have never accepted machine translation post-editing jobs. Years ago, a client for whom I regularly provided ES>EN translation and English editing services tried to save money by sending me documents they had clearly run through Google Translate, since my editing rate was lower than my translation rate. The machine translation was so bad that they ended up having to send me the original Spanish documents and pay the higher rate anyway because there was no way to fix the machine translation. Newer AI models may have improved considerably when it comes to translation, but they still make nonsensical errors, though they are often harder to find now that a higher percentage of their output seems reasonable.

I don’t know what the future holds for me as a translator, but I do know that I don’t want to spend my days trying to decipher and polish AI slop. I don’t like the fact that most AI models are trained on the stolen creative work of countless humans, from professional writers to anyone who has ever posted anything on the internet. I don’t like the fact that we are not being compensated for this theft of our resources (from creative work to water and energy consumption), even as it starts to take our jobs. I don’t like the fact that even the people who have created these AI models don’t fully understand how they work or what they are capable of. I don’t like the fact that by post-editing AI translations, I would essentially be accepting accountability for low-quality work done on the basis of theft.

So in this moment, I will try to keep going as a translator for as long as I can. I truly love translating and it has always brought me joy, even when I’m dealing with boring documents like financial reports, because I love using creativity and research to solve linguistic challenges. If that is no longer possible because humans will be relegated to doing excruciatingly tedious proofreading of AI texts, then I suppose I will be finding a new way to use my linguistic skills in the world. A way that respects humans and helps them connect with each other without relying on thievery and deception.

Why You Should Hire Me Instead of Using AI Translation

If the text you’re having translated matters to you, then hire a human translator like me who you can trust to do a good job using the skills that a high-quality translation requires. When clients hire me for a translation project, they aren’t just hiring some "thing" that takes text in Spanish and translates it to English. They are hiring:

• An eagle-eyed proofreader and fact-checker who identifies spelling, grammar and factual errors in the original text and informs the client so that they can be fixed.

• A skilled writer and editor who ensures that the translated text flows naturally and conveys the client’s message to the intended audience with clarity.

• A discerning researcher who knows how and where to find accurate information when lexical and cultural issues arise during the translation process.

• An experienced linguistic specialist who has developed a deep knowledge of both languages through years of study, and who continues to learn every day as languages and cultures evolve.

Reliable Fact-Checking & Editing: Skills Only Human Translators Can Provide

One of the many benefits of hiring an experienced translator (as opposed to using an unreliable AI translation) is that they can catch errors that have slipped into your text. On numerous occasions, I have been able to help clients fix mistakes in their source text, which is especially important when it is being used as a basis for translations into multiple languages. Sometimes these errors have been minor spelling or grammar mistakes, but in other instances they have been major inaccuracies that could affect their credibility as a business.

For example, a few years ago I was tasked with translating some facts about countries to be used in an educational game for children. The company, based in Latin America, had prepared the facts in Spanish, which I would translate into English. My work would then be used as the basis for translations into many other languages. As I worked, I came across several obvious issues, such as the inclusion of several facts about the “country” of Hawaii. I was able to inform the client that Hawaii is not a country, but a state, so that they could remove the incorrect “facts”. Perhaps the error would have gone unnoticed in the game, but it’s also possible that a parent playing the country trivia with their child might have seen it and wondered why they should be paying for an educational game that had such glaring errors in it. Thankfully, this risk to the company’s reputation was prevented because of my attention to detail.

The Dangers of Relying on AI Transcription in Translation Projects

I recently worked on a translation project that was a great example of the pitfalls of AI. I was hired by an agency to translate transcripts of interviews that had been conducted in Spanish. When I initially skimmed the documents they had received from their client, it seemed like a straightforward task. However, once I got started, it became clear that the transcription had been generated by AI, which the agency had not been told.

How did I know it was transcribed by AI? The transcripts were a mess, riddled with the types of errors a human would not make. For example:

• When the interviewers introduced themselves, their common female names (which were referenced elsewhere in the document) were changed to somewhat similar male names (think Jane > James or Anna > Andrew). Default male bias, perhaps?

• At one point a university with an unusual name was referenced. The AI simply invented a word based on the pronunciation (which produced no search results on Google), but I was able to correct it with an educated guess based on context.

• Every time uno or dos was said, the transcription cited the numbers “1” or “2”. However, in most instances the speakers were using terms like cada uno or los dos.

• Worst of all, text was repeatedly attributed to the wrong speaker. I didn’t have access to the original audio recording, but the issues were obvious. The interviewer would ask a question and the interviewee would start to answer, but then the transcript would switch randomly between speakers, as if the interviewer had the power to read the interviewee’s mind about their personal experiences.

In the end, this translation took much longer than it should have because the source text was generated by AI. Since it was for an agency that I have a great relationship with and the transcription issues were not their fault, I took the time to find and correct as many errors as possible, as well as sending them a list of the issues I had discovered so that they could pass it along to their client. I doubt I’ll ever hear what comes of it, but my hope is that the client might rethink their use of AI given the serious inaccuracies that were introduced into their transcripts.

The Secret Sauce of Translation

As machine translation post-editing has become increasingly popular with clients wanting to save time and money, some seem to have difficulty understanding why I decline such projects, despite my best attempts to explain all of the work that goes into a translation.  

So instead, let’s imagine that I’m a chef at your favorite restaurant, where you always order pasta with my homemade tomato sauce. Making the sauce requires a variety of time-honed skills and knowledge: developing a recipe, sourcing the ingredients and knowing how to adapt when their flavor varies, knowing how to use and cook the ingredients, monitoring cooking times and kitchen safety, and so on. 

You’ve always been happy with the service I provide and you love my sauce, but one day you decide that you should be able to get the sauce cheaper and faster. You walk in with a jar of the cheapest tomato sauce at the grocery store and ask me to improve it so that it tastes just like my homemade sauce. Surely it won’t take nearly as much time as cooking it from scratch, and using the jarred sauce instead of starting with fresh ingredients should decrease the cost as well. So you have a proposition: you’ll give me the jarred sauce to start with, and in exchange, you’ll pay me a fraction of the price once I’ve successfully turned it into my homemade sauce. You say it will be a great deal for me because the time saved will translate into increased output, and I’ll be able to get rich selling my delicious sauce in bulk. 

The whole idea is ridiculous, isn’t it? You obviously can’t just hand a chef a jar of store-bought sauce and expect them to magically transform it into their special recipe. The cheap store-bought sauce surely contains preservatives, plus any number of mysterious ingredients and additives that impact its flavor. It may contain lots of rosemary, while the chef’s recipe uses none. It would require considerable time and skill for the chef to refine the store-bought sauce so that it met their standards, but even then, it could never be the same as theirs. So many unknowns were introduced into the process by starting with the jarred sauce that they have no way of knowing if they were able to correct for them. For that matter, they cannot even guarantee that the sauce is safe for everyone who consumes it, since they don’t know all of the ingredients or how they were prepared. 

That is what it feels like when clients provide a translation generated by AI and ask for it to be “refined” so that it achieves the same high quality as a professional translation done by a human. It is excruciatingly frustrating work that generally takes much more time and effort than translating the text on your own. In exchange, you are offered considerably less money and the knowledge that you are furthering the demise of your industry, sacrificing quality and human ingenuity in favor of making a handful of companies richer. No thanks!