The Localization Trends to Watch in 2024
Here at Milengo HQ, we’ve been asking ourselves what exciting developments will shape the localization industry over the next 12 months. And we’ve uncovered quite the mix – including all-powerful machines, the dream of flawless real-time translations, and just how unique human language truly is. Without further ado, here’s our pick on the most impactful localization trends for 2024!
Localization trend #1: LLMs will continue to evolve
Image credit: Michael Dziedzic on Unsplash
They occasionally hallucinate. They may insist on living in the past. But there’s no doubt that large language models (LLMs) are set to continue dominating the conversation in the localization industry in 2024.
The calm after the storm
There was no avoiding the hype when ChatGPT burst onto the scene back at the end of 2022. It looked like the end was near for traditional Google search, and the death knell had started to ring (somewhat prematurely) for entire professions such as content writers, SEO specialists, and translators.
But fast forward to one year later, and we find ourselves with a much clearer picture of what LLMs can – and can’t – do. Take Microsoft Bing for instance. The tech giant’s search engine hasn’t even come close to overtaking Google as the market leader, despite the fanfare surrounding its new AI-powered version. There is still no substitute for the expertise a human brings to the world of translation and creative writing.
2023 was the year of experimenting with LLMs and investigating what they could do. In 2024, the focus will be on integrating them with other software to enhance productivity (via “custom GPTs,” for instance) and exploring new applications to exploit the exciting potential the technology has to offer.
LLMs in translation software
LLM-driven features are increasingly finding their way into translation management systems (TMS), organically becoming parts of the translation workflow. Here are a couple of examples:
- The popular TMS memoQ incorporates texts from translation memories using its new Adaptive Generative Translation (AGT) feature, which is based on Microsoft Azure OpenAI. This generates translations that are more rooted in the context of the given text, reducing the margin for error.
- Cloud-based tool Lokalise offers the option to create “translation briefs”. To achieve this, it lets users specify the tone they want, as well as the text type and target audience. There is also the option to have ChatGPT automatically check translation quality and categorize errors using the DQF-MQM framework.
A new era for translation agencies
Translation management systems powered by advanced AI algorithms, such as Phrase and Lokalise, are getting closer to making the concept of on-demand translations a reality for companies.
However, rather than rendering translation agencies obsolete, this development will instead transform their role. They will go from merely producing translations to increasingly serving as long-term translation partners that:
- Provide consultation on implementing the best translation technology
- Manage complex localization projects in dozens of languages with ease
- Offer access to teams of highly experienced, native-speaking translators
- Perform quality assurance checks before publishing critical translations. After all, it’s rare that a multinational company has the wealth of linguistic and cultural expertise it needs for every single foreign market it wants to expand into
Localization trend #2: The translation profession will be reinvented
Image credit: Ross Findon on Unsplash
Translation is dead – long live translation
The first time someone predicted the demise of translation as a profession was in the 1950s.
It was 1954 when Georgetown University and IBM conducted an experiment which saw more than 60 sentences translated from Russian into English using punch cards. Elated at their success, the scientists involved predicted that the “challenge” of machine translation could be “solved” within three to five years. As a consequence, the world would no longer need translators.
But as we all know, that’s not quite how things turned out. Despite all the technological upheavals, the profession has demonstrated remarkable resilience over the decades.
There is no question that the arrival of LLMs has brought more cons with it than pros for translators. The proportion of translation jobs consisting of post-editing machine-generated content will continue to rise. Meanwhile, the perceived value of translation as a service is diminishing as getting a free translation these days can be as easy as clicking a button. At the same time, swathes of foreign language institutes and translation departments at universities are closing their doors as they don’t see a future for themselves. To make matters worse, the economic recession and increased price pressure are pushing many translators to leave the industry altogether – just as the crisis is reaching its peak.
Translators are the key to forging meaningful human connections
But it’s not all doom and gloom – there is also light at the end of the tunnel. Some companies have become disillusioned with the quality of machine translation and are rediscovering the value of human translations. LLMs will lack the ability to comprehend a text, the cultural sensitivity, and the linguistic creativity any good translation needs for the foreseeable future. This makes the expertise translators bring to the table vital.
Perhaps these technological limitations will make people see that human translators aren’t just walking dictionaries, and instead view them as specialists in fostering connections who use their communication skills and empathy to create valuable points of contact between cultures.
Heading into 2024, translators who would prefer to not spend most of their time post-editing in the future will have to adapt their profile to the new market situation. We have a few ideas about what they might need to do:
Channel their inner MacGyver:
Translators should prioritize sharpening their problem-solving skills because this is something machines often fail at. Translating a tricky pun, adapting a text to perfectly reflect the target culture, or developing a terminology strategy are all areas where humans can still add real value.
Stand out from the crowd:
As a wise person once said, “those who translate like machines will get replaced by machines.” Translators who deliver quantity over quality will find themselves struggling in the future. That’s why forward-looking translators must aim to produce texts with their own identity, that are tailored to the client’s tone and style, and that dazzle readers with high language quality. In other words, the type of texts that are in high demand in marketing, PR, and journalism. After all, text generated using AI might still inadvertently confuse or upset people (as Kanye West recently found out). Humans offer a way out of this “Uncanny Valley” of translation.
Advocate for their profession:
The hype surrounding LLMs is taking on bizarre forms. For instance, the Albanian president wants to use ChatGPT to translate a significant body of EU laws into Albanian, aiming to save millions in translation and legal costs in the process. But here’s the twist – ChatGPT itself advises against using the tool to translate legal documents when asked! The reality is that translations are increasingly being wrongly perceived as common property that is available free of charge. This caricature must be put to bed once and for all – not with vague opinions, but with verifiable facts and concrete examples. Translators should be prepared to explain to clients how LLMs work, use real-life examples to demonstrate the risks of unchecked AI translations, and prove the business benefits of human translations.
Human in the lead instead of human in the loop:
Human translators can prove their worth by advising their clients and supporting them over a longer time. This may come in the form of:
- Helping clients to improve existing company communication
- Pointing out errors and problems in source texts
- Providing support in applying gender-sensitive language
- Helping develop language-specific style guides
- Contributing to the development of multilingual terminology glossaries
Silently tinkering with texts will no longer be enough in the future – translators will have to be more vocal. They need to proactively expand client collaboration by initiating improvements (human in the lead) and thinking strategically instead of just about the text in front of them. It’s much more difficult to replace a partner like this than someone who simply delivers translations.
Our tip to read:
How do AI and language interact with one another? The Chartered Institute of Linguists (CIOL) has published a comprehensive study on the impact of AI on the future of the translation industry. Click here to read the document.
Localization trend #3: AI will become the hottest new voice talent in town
Image credit: Hunter Harritt on Unsplash
At Milengo, it’s our vision to overcome language barriers and make information all around the world even more accessible. What machine translation has achieved for written texts will also be within reach for audio content in 2024 – in the form of AI voices that switch from one language to another in the way a polyglot would.
AI means you can now listen to your favorite podcast in your native language
Spotify’s announcement of its new “Voice Translation” tool sent shockwaves through the podcasting world. Developed using OpenAI’s Whisper speech recognition system, it uses a speech-to-text model which includes speech synthesis to translate English voices into Spanish, French, and German. The nuances and rhythm of the speaker’s voice are also imitated with startling accuracy. Various English-language podcasts, including those by Dax Shepard, Monica Padman, Lex Fridman, Bill Simmons, and Steven Bartlett, were used as a pilot project. If you’re curious about the end result, then head over to Spotify’s newly created Voice Translations Hub. Listen to the different language versions of these podcasts and see what you think!
The AI-generated versions are certainly easy to listen to, with fact-driven, informative podcasts being the most likely to play to the technology’s strengths. On the other hand, personality and chat-based formats, podcasts with strong branding, and podcasts that deal with sensitive issues such as mental illness will probably struggle.
AI voices for the business
The potential of these AI voices is huge and of course, also something companies would want to investigate. Take training videos for example – these could now easily be made available in several languages. However, as is so often the case, the devil may well be in the details. The AI might find it more difficult to recognize accents, dialects, and tones of voice that are less common. Plus, AI still has plenty of room for improvement in its general understanding of human speech, which might cause issues.
For example, if an expert on a soccer podcast heaps praise on a particular team’s “wide players” – the fullbacks and the wingers – but the AI understands “white players,” then suddenly the praise turns into racism in the blink of an eye. What’s more, the nuances of verbal communication, such as emphasis – something which is often brimming with subtext (even simple statements such as “Sure thing!” can be meant in earnest or be dripping with sarcasm) – are also susceptible to misinterpretation.
Human quality control measures will be essential to protect the reputation and integrity of podcasters and other organizations.
Whether it be LLM-generated content, voice translations, or high-end human translations – these localization trends and developments will offer companies a wealth of options to choose from when planning their translation strategy. However, with great power comes great responsibility. This trade-off between authenticity, quality, information accessibility, and cost efficiency must be made on a case-by-case basis and should ideally be left in the hands of localization professionals.