Lilt CEO Spence Green’s observation on SlatorPod that localization has become one of the key use cases to show return on investment in AI is something most people in the industry can agree with. And when some big names like Reddit not only make AI-enabled localization part of their growth strategy, but also make good on their intent, as they announced in September 2024, the skeptics ought to notice, too.
The technology is certainly there for all who wish to open the door to automation of multiple localization tasks.
Transformed content workflows and customization of target outputs via trained/fine-tuned large language models (LLMs) are just two examples of what can now be done. But there is a lot more.
What retrieval-augmented generation (RAG) and AI orchestration can help accomplish, and the ability to process ultra-mega-massive volumes of content using AI, are also great ways to boost that precious ROI… even automating quality assurance to different degrees based on the type of content.
Still, quality assurance remains one of those subjects for experimentation and debate. For that and much more, it is humans who need to get their heads around retooling processes, roles, and the back office so that no new problems are unnecessarily created as AI is implemented.
We asked readers if AI (LLMs included) has changed their day-to-day work in the past 24 months. Over a third of respondents (37.5%) said that has not happened at all (!). A little less than a quarter (23.6%) are on the opposite end of the spectrum saying it has done so radically. The rest, two equal groups (19.4% each respectively), said it has somewhat changed it or just a bit.
Benched or Still in The Game?
“Bouleversement” is a French word that perfectly describes the sort of situation some in the language industry are facing as it evolves toward increased AI automation: beyond upset or disruption, more like an upheaval.
Back in March 2024, readers could already feel upheaval and pressure increasing as they answered polls on bankruptcies, the relevance of language service provider offerings, and individual use of AI. Two quarters later, the upheaval appears to gain momentum.
The market will continue to change. More translators will become AI training and QA experts, and more project managers will become a combination of AI process designers and sentries. Regardless of role, most in the industry are likely feeling the pressure, with generalist language service providers (LSPs) and linguists facing tough choices.
We asked readers how the talent situation is in this environment in their corner of the language industry, and half (50.0%) responded that there is an oversupply, and the talent situation is super competitive.
Interestingly, a third of respondents actually face a talent shortage (30.0%). For the rest (20.0%), the talent situation is balanced.
Practical and Tactical AI Adoption
At SlatorCon Silicon Valley in September 2024, the growing level of confidence in generative AI and large language models (LLMs) in localization at the enterprise became quite clear.
Slator research featured 20 practical applications of large language models (LLMs) in translation and localization. And Hameed Afssari, Uber’s Head of Globalization, commented later in a conference panel that Uber has “tested or implemented 12 out of the 20 already.”
It is not uncommon now to find an equal measure of scientific research on LLMs and language and language technology companies leveraging them for features and functionality on localization products and services, and Uber is but one example.
2024 Slator Pro Guide: Translation AI
The 2024 Slator Pro Guide presents 20 new and impactful ways that LLMs can be used to enhance translation workflows.
Small, medium, and large companies, as well as freelancers, and just about everyone with a job in the industry, have been exposed to LLMs in one way or another.
We asked readers how they would describe their company’s level of LLM adoption, and close to a quarter (24.5%) is in the awareness phase. Two groups of respondents (20.4% each) are either still in the discovery stage or actively using LLMs.
The rest of the respondents said their level of adoption is inactive (16.3%), operational (10.2%) or systemic (8.2%).
Winding Down or Ramping Up?
Something else that was discussed in the rooms and halls at SlatorCon Silicon Valley in September 2024 was the featurization of AI translation in many products (where translation did not exist before).
There are several cases. Adobe offering AI in beta mode dubbing and lip-syncing API and Asana revealing “AI Teammates” as a tool in the management platform to write, translate, update, and even answer questions, are two of them. Both are on the localization buyer side and what this means for the host of localization vendors that can be impacted by such internal featurization at large tech companies and elsewhere is speculative at the time this article is written.
Slator Pro Guide: Audiovisual Translation
The Slator Pro Guide: Audiovisual Translation is a concise guide to audiovisual translation, including dubbing, subtitling, access services, AI dubbing, AI captions, and more.
Whatever the level of speculation and uncertainty, with ChatGPT now an old mate at almost three years old, surely language access businesses have a different outlook for Q4 2024 than they did a year ago.
We asked readers what their business outlook is for the rest of 2024, and over a third (34.6%) said it is positive. For a little over a quarter (26.9%) of respondents, things are looking slightly negative. For one in every six respondents (17.3%), the horizon is neutral. For the rest, the outlook is either very negative (13.5%) or very positive (7.7%).
