As AI Adoption Accelerates, the twin issues to address for me are transparency and the looming cliff edge that is the loss of trained translators who AI relies upon to feed its memories and edit out its mistakes.
First up the transparency issue. As machine translation becomes increasingly common across the language services industry, the elephant in the corner of the room (or l’éléphante au coin de la pièce, as an AI engine near you might be tempted to say), is that many businesses are being kept in the dark about what they are actually paying for. And I wonder if that’s also the case in other industries? While artificial intelligence has undoubtedly transformed translation workflows, questions remain over whether clients are benefiting fairly from the efficiencies it creates.
Many translation agencies now offer Machine Translation Post-Editing (MTPE), a process in which AI-generated translations are reviewed and corrected by human linguists. The approach can be highly effective for large volumes of technical or informational content and often enables faster turnaround times. As we know.
However, I’ve started to notice a growing gap between the savings generated by machine translation and the discounts passed on to end clients.
There’s nothing wrong with using technology to improve efficiency; as founder The Aktuel Translation Group nearly 35 years ago, that has always been one of the three golden rules of translation (alongside accuracy and reasonable pricing). The issue is commercial integrity. End users deserve to know how the margins really stack up and translators need to be fairly compensated.
By way of example, cost savings to agencies by using machine translation that is then edited are around 40-60%, whereas my experience shows the savings passed onto clients hover around the 20% mark. That’s a big difference.
So, whilst MTPE is frequently marketed as a lower-cost alternative to traditional translation, the reduction in price offered to clients is often modest compared with the actual reduction in production costs. At the same time, professional translators report being paid significantly less for post-editing work despite retaining responsibility for the quality and accuracy of the final text.
So, the rise of AI also threatens the very market it claims to help (‘It doth mock the meat it feeds upon,’ to quote Will S). This is because these underpaid but highly experienced translators are now going off and do other things that allow them to pay the bills and go to the supermarket. When they’ve gone, all we’ll have is AI run by people who are underqualified.
To drill down to specifics on the pricing issue, large agencies for human translation charge around 15p a word for common languages and pay the translator around 7p a word. For machine translation they charge around 12p a word and pay a translator 3p a word to edit. That’s an enormous difference to their margin.
Simply put, in an AI scenario:
Agency margin rises by 41%
Translator pay falls by 57%
Client prices fall by only 20%
By not passing on full savings and underpaying professionals, confidence in the industry is sapped and we start to lose our most valuable resource.
And it doesn’t need to be like that. The Aktuel Translation Group is committed to providing 100% human translation as standard at a transparent rate of what clients would normally pay for machine translation, plus editing. It’s a price point designed to remain competitive with AI translation while ensuring higher standards and that experienced linguists are fairly compensated for their work.
This is because I still believe high-quality human translation should not be treated as a luxury service and it doesn’t need to be. My view is simple: AI is only worth the candle for a client if they receive a full and fair discount. If not, pure human translation can cost the same and has fewer risks involved of AI errors not being spotted.
AI technology has a long way to go, and human editing is still almost always required. For legal translation, AI will often treat keywords like ‘possession’ and ‘ownership’ as the same, when they clearly have very different legal connotations. For texts requiring a degree of idiomatic understanding, AI will often translate a phrase like ‘I don’t think he’s wrong’ as ‘I think he isn’t correct’; AI very rarely can reproduce authorial voice or add empathy or charm to a text that requires it … one could go on.
If I were running a business seeking translation services, I would encourage you to really look at whether you are really making savings and whether you can afford to take the risk of using AI. I would also ask them (and agencies) to examine their conscience and ensure that the people who make it all possible, the linguists with years of training and experience and genuine love of the language, are paid fairly.
This article was written by a bod, not a bot.
Original article can be found here.
Recent Comments