Why do translation companies speak of AI if they mostly use NMT?

Let's give a brief, concise and direct answer: for marketing and market expectations. The term 'AI' sells. After the generative AI boom in the media, translation companies do not want to appear left behind in the technological wave. Although at one time 'neural machine translation' (NMT) sounded cutting-edge, today that label is overshadowed by the allure of the acronym 'AI'.

As a result, many companies continue to use their usual NMT engines, but now promote them as 'AI-powered translation' solutions. For example, DeepL describes its platform as 'the ultimate AI-powered translation solution for businesses'. On their website they state: Our AI technology provides accurate translations...

In practice, DeepL's core technology remains a neural translator highly trained with bilingual data (and recently expanded with a LLM). They have not stopped using neural networks nor have they discovered a new form of alchemy; they have simply begun marketing it under the broader umbrella of 'AI'.

Why this rebranding? There are several reasons:

Taking advantage of the trend

Customers from all sectors are excited about artificial intelligence. If a translation company speaks their language (quite literally) and promises them 'AI translation', it sounds more innovative than saying 'machine translation'. The term 'machine translation' may conjure images of overly literal translations or comical errors of the past, whereas 'AI' suggests intelligence, learning and modernity.

Informing the client

Paradoxically, many users do not know that modern machine translation was already AI. After ChatGPT, some executives or clients have started asking companies, 'Do you use AI to translate?'. If the company were to respond, 'No, we use NMT', the client may not understand that it is actually the same thing. To simplify things, they say 'Yes, we use AI’. In that sense, many people ask if AI can translate without realising that this capability is neither new nor exclusive to recent LLMs, since NMT was already doing it. So companies clarify themselves and say, ‘Yes, we offer AI in translations (we had NMT, which is AI)'

Maintaining relevance

After ChatGPT, there is a belief that a company not mentioning AI is falling behind. Even if they continue using the same tools, emphasising 'artificial intelligence' in their value proposition allows them to position themselves as part of the current digital transformation.

In summary, translation companies continue to predominantly use NMT, but they rebrand it as AI to capture attention. It's more a change of discourse than of underlying technology. Obviously, this does not mean that they are being misleading about its capabilities (a good NMT engine is indeed a powerful AI). However, it can cause confusion if clients assume that using ‘AI’ guarantees 'immediate human-like quality' in translations or implies that systems such as GPT-4 are being used when they are not. For this reason, it’s worth understanding what the ‘AI’ label actually represents in each situation.