Machine translation has been around for years, with acknowledged limitations. We’ve all seen the amusing back translations from Google Translate, which, although improved, are still somewhat imperfect. But AI translation? Like with everything AI, there’s been a lot of hype around its capabilities, so let’s unpack its pros and cons.
We’ll cover…
- Promises of AI
- Consequences of AI editing
- AI translation for marketing & SEO
- How to implement AI into a translation flow
How AI translation is changing the industry
The promise of AI translation is not dissimilar to the rollout of AI in any other industry – it promises to work faster and for cheaper than humans.
And there are many instances where AI translation is making significant progress. Integrated into Google Translate, smartphone users can now easily and accurately speak into the software and have it output something in a completely different language. We can all take a picture of a menu and see what it says without having to interpret complicated arm gestures from a waiter.
It’s also being rolled out into subtitling and dubbing processes, as seen on the likes of YouTube.
There’s no doubt about it, it’s making our increasingly globalised lives easier. So, what are its limitations?
Human-in-the-loop: AI translations and editing
The biggest downfall with AI translation is its susceptibility to error, which, if you don’t speak the second language, you can’t see. AI translations need to be carefully proofread and edited by translators, which can be more difficult than you might think.
Translators editing AI translations might struggle to create translations as good as they could come up with on their own, because they’ll be affected by the AI translation that they’ve read. This interference means that translators can quite often identify that a translation is wrong, but struggle to come up with an alternative, as their brain is occupied with the AI translation. Figuring out a better translation often requires taking a step back, making editing AI translations a time-consuming – and often frustrating – process.
AI translation should only really be used for situations of minimal consequence, like ordering from a menu, or for straightforward texts where mistakes are unlikely to happen because a literal translation will suffice (such as a plain language internal email).
How well does AI translation work for marketing and SEO?
Growing a global business is tough, because creating impactful marketing in multiple languages and multiple regions is tough. But AI translation promised to make this easy. Has it worked?
In short, no. There are several reasons why AI translation doesn’t work for global marketing strategies…
Marketing and SEO require cultural context
SEO is about creating helpful, relevant content for users in a target market, using key phrases that they input into search engines. Keywords are chosen based on knowledge of the target culture’s search behaviour, which is based on their cultural context. And what does AI translation have? A distinct lack of cultural context.
AI and machine translation will convert texts word-for-word, ignoring differences between cultures. The impact of this is often nonsensical translations that – at worst – can offend target customers, or – at best – completely miss the mark. Human insight is required to shape international and multilingual marketing campaigns so that their impact, rather than their exact content, is translated.
Side note – if you’re wondering how successful SEO translation is carried out, I wrote a blog all about it.
AI is inaccurate
Another priority for Google when ranking webpages is the quality of content. Non-sensical, inaccurate translations that are clumsy to read won’t get your content seen.
Consistent errors in translated content are also off-putting to customers, who lose faith in your brand.
Is it ever okay to use AI for translation?
We’ve discussed how using AI translation can be more time-consuming than it seems, require meticulous editing, and have damaging consequences for SEO and international reputation. So, is there a point in using it… ever?
The answer is yes – but only for certain types of content. Content that needs to be highly localised or accurate, such as marketing or legal texts, should be left to a human. But texts that are repetitive, or can be translated more literally, can be a good case for AI translation (with proofreading from a human).
The language you are translating in and out of should also be considered, as AI engines are not universally proficient. Spanish, for example, is much better understood than Arabic.
One final thing to think about is potential data security issues that can arise from using AI translation models. Information from uploaded documents will be used by the tool in future responses to other users, so it’s essential to ensure that no confidential information is shared. Companies such as Samsung actually banned the use of ChatGPT by staff members after confidential information was interpreted by the AI.
To conclude…
Whilst there’s no doubt that AI engines will continue to develop, there’s currently no foolproof system for using AI translation tools for all of your translating needs. AI translation outputs need to be tested and evaluated on a case-by-case basis, which is time-consuming in itself.
My recommendation for global brands looking to tackle a new market is to use AI translation for low-risk situations like internal documents, ensuring that they are still proofread and that documents are secure. Then invest in a marketing translation partner who can assist create culturally-relevant, fluent translations which suit the target audience, without the potential risks of AI translation.
Keen to find out how we could work together? Check out my services here.