Have you used or considered using AI for your content or translation needs? If so, you won’t be alone – ever since ChatGPT hit the scene, companies all across the world have been experimenting with its cost-cutting and efficiency-boosting potential. What’s more, at my LSP, we have seen first-hand how clients are eager to leverage this technology to cut costs and speed up turnaround times.
But it is important to understand that AI is not free of risk! And although its user-friendly, chat-based interface makes it seem a cinch, it would be foolish to believe that using AI requires no special training or expertise. Without the right skillset and knowledge, the unconstrained use of AI can have many negative unintended consequences.
Below are just some of the risks associated with using an AI solution without proper human oversight and review:
No way to safeguard against mistranslations
The biggest risk of using AI to translate your content is that the meaning of your original text may be changed. AI-based translation tools are trained to always offer a translation even when the meaning of the original may not be clear or unambiguous. They are also prone to hallucinations, which means that they invent texts based on word-order predictions rather than other inputs such as context or subject knowledge.
It is for this reason that translation agencies will always include a human review step as part of all AI or MT-based services. This allows them to pick up on errors or mistakes that might otherwise be missed. So don’t be fooled by sleek and confident-seeming AI interfaces – the content they churn out may not be reliable, and without a deep-seated knowledge of all the languages involved, it will be impossible to tell for sure what issues may lurk within.
This, of course, is the key risk for direct clients – it is impossible to assess quality across multiple languages without the muscle of a professional, multi-lingual team.
The worst case scenario here – and the biggest risk – is that you put out problematic content without realising it. This could then begin sowing the seeds of doubt and generating mistrust among your readers in certain markets. And to make things even worse, without the proper linguistic expertise, it may take several weeks, months or even years before you are able to detect the source of the problem. By this point, lasting damage may already be done.
Nuance and attention to detail go out the window
Translation is more than just swapping words in one language for words in another. The full localisation process involves carefully combing over every aspect of the text in all its minute detail, carefully selecting words and phrases in order to give each translation the best possible chances of success.
While AI may be impressive in its ability to rapidly generate results, there are many aspects of communication that it is just not capable of considering – from subtle references to the latest cultural phenomena and playful alliteration to local turns of phrase and differences of dialect. The more colourful or creative a piece of text, the more difficult it is for the machine to retain all of the elements that make it unique and engaging in the first place.
What this means is that the text you have worked so hard to craft loses all of its appeal in translation.
Even if it may seem fine when read as a back translation, in the target language it will likely sound flat, uninspiring and perhaps even outright incomprehensible.
And if the purpose of your text is to reel in and convince readers, then putting a dull shadow of your original out into cyber space is about as good as pointless. At best, it will have little effect whatsoever – at worst, it will confound your readers and turn them away from your brand.
You may accidentally disclose sensitive or confidential information
Using AI or MT to quickly translate high volumes of text may seem like an efficient way to save both time and money – but do you have a good overview of all the all personal data that may be contained in your files? As humans, we naturally and readily pick up on any and all references to private individuals, including details that can be used to indirectly work out a person’s identity. AI programmes, on the other hand, may let stray pieces of sensitive data through the net. The consequences of this can be wide ranging – from embarrassing disclosures to fines from the data protection authority or even lengthy and costly claims for damages.
What’s more, data protection law also requires organisations using AI in the EU to document how they address the data-protection risks. This bureaucratic process requires safeguards to be developed and implemented, and it will inevitably eat away at some of the efficiency gains from using AI.
You may contribute to the spread of harmful misinformation
Translation is not a binary process. It is not a case of right or wrong, or swapping one word for another. Every sentence is laden with many shades of nuance – each noun and adjective has been chosen from a range of different options, reflecting the conscious intent and unconscious beliefs of the writer as well as other, much subtler factors and influences. It would therefore be foolish to believe it is an easily automatable process that can be carried out by machines without risk of corruption or change.
Right now, our feeds and information sources are being flooded with dis- and misinformation. Some of this content is produced maliciously by malign actors with their own agendas. But other parts are being produced unwittingly, by automated content generation programmes that tend to hallucinate or invent their own reality when they are unsure. Some of this content comes from machine and AI translations.
The question we need to ask ourselves is what happens as the material we read online becomes increasingly untrustworthy? How will we, as humans, adapt to this problem and find solutions we can trust? At this juncture in time, it is important not to contribute further to the problem, and to introduce processes that will enhance the reliability of the content we create. This means involving expert, human linguists in all translation projects.
Make the most of automation – The safe way!
The point of this article is not to scare you off ever using MT or AI again. On the contrary, we believe that AI is a fantastic tool that will inevitably revolutionise not only the translation industry, but the entire world. The risk is not automation in and of itself – the risk is using these tools without the proper supervision or knowledge.
COMUNICA, has been embracing MT and automated solutions from their introduction, using them effectively to boost our efficiency and offer more competitive rates to our clients. We are able to do this successfully not just because the systems we use are cutting edge, but because we can combine them with decades of human knowledge and cultural expertise. This allows us to ensure that all of the automated translation solutions we offer are accurate, nuanced and secure.
So don’t take risks by going solo – make the most of automation the safe way. Make the most of automation with us. Contact us today for a no-obligation quote and more information.