Over the last few years, Machine Translation (MT) and Artificial Intelligence (AI) have both revolutionised the translation industry. From Google Translate 20 years ago to today’s sophisticated Neural Machine Translation (NMT) models like DeepL, AI and automatic tools have made translation faster and more accessible, as well as eliminating human error. However, just because there are no human errors doesn’t mean that there are no errors at all, just none of the ones that people tend to make.
Despite the benefits of AI tools, there are many reasons to be wary or, at the very least, extremely careful when it comes to their use. Now, I’m not here to bash AI. Just like many other professions, language professionals around the world are now attempting to harness the power of technology, but there are drawbacks and issues to take into consideration. So, let’s take a look at what it’s all about.
The Rise of Machine Translation and AI in Translation
AI tools use deep learning algorithms to process and translate text with unprecedented speed and efficiency. Unlike older machine translation methods, which use sets of linguistic rules, AI analyses vast quantities of multilingual data to predict the best translation for a text or document. This means that it is able to produce more natural-sounding translations that fit the context better.
These tools are rapidly becoming more and more widespread, and almost all businesses, large corporations and SMEs are using them, whether it’s to translate their websites or content or provide customer support. Healthcare, legal services, politics… there is almost nowhere that AI hasn’t now touched.
A Gamechanger for Good?
- Speed and Efficiency – AI and MT can process much larger volumes much quicker than a human translator. In fact, they can produce results almost instantly even if the document is hundreds of pages long. This is obviously a huge benefit when time is of the essence.
- Cost – Automated machine translation services are generally cheaper than human translators, making them much more affordable for small businesses with limited budgets.
- Improved Accuracy Over Time – As they process more and more data, in theory, AI should get increasingly more accurate (although we’ll see the caveat with this assumption in a second).
Or a Liability?
- Lack of Cultural Nuance – One thing that is very important to understand is that machine translation and AI tools themselves do not understand anything. They simulate understanding. It’s an illusion. AI uses probability and huge datasets to draw conclusions, not real-world experience, knowledge or understanding. As a result, machine translation can often struggle to process cultural context, sayings or humour. This can mean that translations, while accurate, can seem rather unnatural to native speakers.
- Limited Accuracy for Complex Texts – Machine translation is great at simple sentences where very little contextual knowledge is required. But for anything technical or text loaded with jargon, it can get into something of a pickle very quickly. This can make it a hindrance rather than a useful tool when translating technical, legal or literary texts. Indeed, a recent study has found that AI is not particularly good at accurately summarising the news.
- Dependence on Training Data – Garbage in, garbage out. AI and MT are only as good as the data used to train them. If there are gaps or errors in the data, the output will reflect this.
- Security and Confidentiality Risks – This is a big one that many companies overlook, including many translation agencies. The vast, vast majority of tools available are online tools rather than native software tools. As a result, every time a text is fed into the machine, someone else has a copy of it. This is where NDAs are critical. Even then, for some document types, it’s just too big of a risk to allow AI to access proprietary information.
Translation in the time of AI
So, are human translators a thing of the past? Absolutely not. Despite advances in AI technology, it is still a long way from being reliable enough to produce accurate, natural translations that are fit for purpose without human supervision. This is particularly the case in fields that require extensive cultural or subject-specific knowledge. Like many translators and translation agencies, I offer Machine Translation Post Editing as an option when customers need a translation urgently. Obviously, they are fully informed of all the pros and cons (which are not just limited to the main ones listed above). AI isn’t replacing translators, but it’s a very powerful tool which, used the right way, can help everyone.
Need to Translate Large Volumes Quickly?
AI-driven machine translation is transforming the industry, but it’s no silver bullet. Any company that believes that AI alone is going to solve its problems is in for a rude awakening. Without human supervision and critical thinking, AI is like a high-powered car with no steering wheel. It might get you there quickly… or get you nowhere fast. If you need to translate regular or large volumes of text, feel free to get in touch to see if human-supervised AI or MT might be the solution for you.
EN 
ES