Artificial Intelligence in Translation

Since the launch of ChatGPT (Chat Generative Pre-trained Transformer) by OpenAI in November 2022, there has been a real hype around this innovative tool as well as similar software that has been launched continuously since. A range of sectors have been identified as being particularly exposed to the consequences of the rise of artificial intelligence (AI), among them highly skilled professions such as translation and interpreting. Having experienced AI’s rapid development and impact on the translation industry in recent months myself, I have been wondering how AI may transform the future of professional translation. Here are my initial thoughts:

Client demand for such tools is high.

With human language at its heart, it is not surprising that translation is seen as a prime industry where the impact of AI will be felt. Large translation agencies have, for example, been launching new departments dedicated to training AI engines to help their clients develop tools that summarize texts, generate questions and answers about them as well as new texts based on certain parameters and prompts from users. But is machine learning, the basis of AI, something completely new in the translation sector? Over the years, translators have already been regularly using various machine-learning tools, such as translation memories and neural machine translation engines. Hence, the translation industry has been a “tech-heavy” sector for a while, and translators have indeed been forced – and been able – to adapt to new technologies. Whilst initially clients may have wondered why they should still employ translation agencies when they can get texts machine translated by engines available on the internet, they soon realised that the quality can be patchy, especially where English is not part of the respective language pair. Privacy is also an issue given that once texts have been translated by tools like Google Translate or DeepL, they are in the public domain. The experience from machine translation is that the machine will not replace the human translator in the foreseeable future since the quality of human translation remains unmatched. It is rather the human translator who uses the technology and improves the quality of its output.

❝Will this be any different with AI? Cultural sensitivities and linguistic nuances require human translators to be involved in the professional translation process. This hasn’t changed despite the creation of widely accessible cloud-based translation tools. ❠

The human translator remains the master of the translation process, even for light post-editing tasks that produce “rough and ready” translations for non-public purposes. I have been involved in some AI-driven projects, for instance testing the text-generative capacities of AI or its ability to summarise texts and engage in a discussion about them. So far, my experience has been mixed in that AI-generated texts seem to be either quite generic and non-creative with uninspiring storytelling or sounding logical but containing some serious errors. I recently reviewed an AI-generated text in the field of economics that arbitrarily mixed the macroeconomic instruments of monetary and fiscal policy. Let alone the dangers such errors pose to the uninformed reader, from being factually wrong to politically biased or even manipulative, such examples of both fictional and non-fictional texts prove that AI-generated texts still require human revision and, as far as different languages are involved, the “human touch” of translators.

So will the future model be one where human translators work alongside AI? Surely, the deployment of AI in translation will develop further at rapid speed, but it is uncertain what level of excellence AI will achieve, even with all the human input currently going into AI training. This also applies to large European languages such as English, where AI development efforts are currently particularly focused on. It seems to me that it may take quite some time before AI can be equivalent to the human brain as far as translation is concerned, if ever.

Nevertheless, the translation sector will inevitably have to adapt to routinely using AI and it will therefore be crucial for translators to adapt and be able to use this new technology on a daily basis, too. Rather than only translating from scratch, AI output evaluation and editing AI-generated texts may well become new key areas for translators to work in. I would describe my involvement in AI projects to date as that of being a “linguistic consultant”, both in terms of the source and target text. And this human oversight aspect does not even touch upon the difficult topics of ethics and regulatory oversight of AI, currently intensely discussed areas that go far beyond the role that AI is likely to play in translation in future.