Act 3: Reframing the Scene

What does the future hold?

Chances are, every person you ask this question will offer a different answer. And that’s as it should be (let’s leave the probable-words-in-a-sequence prediction to something else). The future will likely be composed of a multiplicity of parallel realities, much like a multi-colored glass vitrage.

Any discussion about the future comes with its share of responsibility. A responsibility toward ourselves, the teams we work in, and the communities we are involved in. A responsibility for those who come next — the next generation of talent that will be entering our industry.

When reading this closing section, ask yourselves: What are YOU doing to affect those around you positively?

Balázs Kis

Co-Founder and Chief Evangelist at memoQ

Balázs Kis is a Co-Founder and Chief Evangelist of memoQ. He has decades of experience in translation and translation technology. Balázs is a member of the Language Industry expert group (LIND) of the Directorate-General for Translation (DGT) of the European Commission, and is also, and is also involved in the AI Localization Think Tank.

There is a lot of talk and writing about the risks and shortcomings of AI, and how it is becoming a major player in all work related to language. I write some of these pieces myself (although I am far from being the most prominent voice in this space).

On the other hand, I have not seen a lot of discussions on how to prevent or stop all the bad things usually associated with AI from happening, at least outside the circles of compliance and research. However, before we can have this conversation in a wider circle, we need to have an idea of what a desired outcome looks like.

To facilitate this, I will present two best-case (or close-to-best-case) scenarios, the first from the perspective of people and organizations, and the second with technology at the center.

People: New premium, new professions, a stronger position

Back in 2013, the use of machine translation was already at the level that, according to some estimations, less than one-millionth of all translations were produced by humans. But the portion of the content that was actually translated by humans or with humans having the last word was more valuable than the seas of machine-translated text. This is connected to the purpose of the content.

Content becomes valuable when human life, livelihood, or the viability of a business depends on it, or if it represents such cultural or aesthetic value that it is inconceivable to use the cheapest form of automation on it. This value remains there regardless of the advances in AI or other automation techniques.

Today, it isn’t only about translation: More and more online content is generated by AI. In a way, AI models are the “great equalizers.” They produce uniform content in a uniform style. And, because AI-generated content is not consistently labeled, it gets back to the training data of foundational models, reinforcing the bland and uniform nature of the output.

Conversely, this also means that there will be a need for high-value content that is created or curated by humans, and the hope — and this is the best-case scenario — is that society recognizes that this is needed and begins to appreciate and pay for the real value of such content. In other words, human-curated content will no longer be a commodity because the contrast between cheap and premium becomes evident. Imagine lighthouses scattered over a grey ocean of AI-generated mediocrity.

However, AI is not the only factor that imposes fundamental changes on the language industry. Quiet and not-so-quiet changes have been afoot in the way client-side organizations look at the role of localization in their content processes. Localization is no longer an afterthought, but it also means that it isn’t a monolithic process that can simply be attached at the end of the content management process. The steps and motions of localization are now embedded in various stages of content creation. As a result, it is more difficult, if not impossible, to outsource “translation,” and it has become harder to separate localization from the rest of content-related work.

The outcome we want — and the one I hope for — is that the language industry actually becomes something more than just “localization,” that the profession becomes all things Language. I expect that the professionals and organizations in this community will have the opportunity to be the language specialists, content creators, and language managers they deserve to be.

Let me share another recent revelation: Businesses outside the language industry struggle to incorporate AI in their processes. (That isn’t the revelation: wait for it.) Most of the time, it is done for the sake of the AI, rather than for the business or its customers, simply out of fear of missing out. But there is little reliable information and education available for these organizations. Guidance on where to start is also scarce.

This is where the language community comes into the picture (this is the revelation!): Because modern AI has everything to do with language, our profession is the single most qualified group to educate everyone else on the benefits and the responsible use of AI and AI-based technology. If you ask me, I can’t wait for us to get there.

Tech: More reliable and more responsibly used AI

Language-based AI is the single most unpredictable form of technology humanity has ever invented. It lacks the precision we are used to from computers, and it also does not have the kind of motivation and structure that humans use to manage their tasks.

This is not a philosophical problem. If AI models stay at today’s levels, businesses will struggle to assess the value they derive from AI and will find it difficult to have a sense of a return on their investment. The only businesses getting reliable and predictable value from AI today are the providers and the consultants implementing systems for a subscription fee.

My hope — and expectation — is that this will change, and AI systems will pick up some structure and predictability. I have a feeling that this will happen outside of the realm of training data because curating and annotating data at the level that provides this reliability probably comes at a prohibitive cost and, although the large providers don’t share financial information, I have strong doubts they make a profit on their models. It is smaller models by smaller players that are where we can curate data more carefully; they might be (part of) the way ahead.

Today, there is a lot of noise around large language models and the technologies derived from them. Many people, organizations, media outlets, opinion leaders — advocates and critics alike — view AI as a force that, together with social media, implements power to control the population.

The dust must settle. If we’re talking about a best-case scenario, we say that the dust will settle. AI must become a tool that businesses and individuals use as an extension of their innate capabilities to manage their lives and business more efficiently.

Read the full 132-page Global Ambitions: (R)Evolution in Motion publication featuring vital perspectives from 31 industry leaders on the ongoing AI-spurred (r)evolution.

Get your digital copy today