You're either the captain of this vessel or you're overboard!

Introduction

It's undeniable that we've entered a new revolutionary era with the advent of current Neural Machine Translation and AI technologies. Positioned at the leading edge, we have already begun to witness this transformation, and this paper is our way of sharing those experiences with you.

The performance of machine translation, progressing through its different phases, delighted us and caressed our sense of human dignity until it arrived at the neural stage. For this reason, the remarkable success of Neural Machine Translation (NMT) was initially overshadowed by the shortcomings of earlier machine translation attempts. 

As it became more specialised and was fed with increasingly large amounts of data, NMT reached an incredible level. However, acceptance of NMT by public would have required time. Artificial intelligence (Al) shortened this process. Yet, with the emergence of Generative Al and its related applications, it swiftly became extensively popular worldwide, changing public perceptions and speeding up the introduction process.

For example, let’s compare it with Trados. Ever since Trados was first launched in the market in 1984, along with other CAT Tools and Platforms, it has continued to be an essential source of expertise for translation agencies and translators. Unlike CAT Tools technologies, the arrival of ChatGPT4 marks a different era, having quickly become a significant part of our lives and gaining rapid popularity. And therefore, numerous translation firms or start-up initiatives have seized upon this development, making transformative moves with related applications.

Present Circumstances

Today, NMT has become a distinct speciality; the performance of different MT engines can vary based on expertise and language pair, and selecting the most accurate NMT engine for the desired outcome is a technical decision awaiting us.

The situation with Al is slightly different. Utmost care must be taken in giving verbal commands to ensure AI translations do not skip information or produce incorrect translations. As you may have noticed, when dealing with AI-generated translations, accuracy is a priority we must attend to. When we have shown our translator colleagues examples of both NMT and AI-generated translations using different commands, we found there was not much difference, but there was a tendency to prefer NMT. On the other hand, we see that new algorithms related to translation are being added to artificial intelligence, making it progressively smarter.

Conclusion

There is already a general consensus that the way of working in the translation industry will shift towards NMT or AI post-editing, and the recent developments unequivocally confirm this. The direct companies we work with and other translation agencies are now cautiously requesting NMT post-editing, but human translation accuracy and natural language. This indicates we may be working on more NMT post-editing jobs, but rates might slightly decrease depending on NMT output quality. 

Just like this year, factors such as geopolitical tensions, inflation, and intensive international competition could lead to challenging years where companies might find some relief in their translation budgets. Work may not come to an abrupt halt but could continue in a steady flow.

Ultimately, we find ourselves in the very scenario this saying describes: You're either the captain of this vessel or you're overboard!

Author: Volkan Güvenç 
General Coordinator 
at Alafranga Language Solutions