The idea of humans being reduced to acting as the “moral crumple zone” (see my last blog) is, clearly, frightening – and a highly undesirable outcome for translators and post-editors, and in the end for clients, too.

So how do we make sure that we are not just the scapegoats, the ones that take the blame when things go awry? After all, even the creators cannot quite understand, or explain, how neuronal networks produce their results. And since we have no, or very little, control of the process, it does not seem rather odd that the responsibility for any errors lies with us.

Everyone knows how difficult it is to define what is and what isn’t a good translation. How much effort is needed to achieve the desired quality, how much time needs to be spent, and what is the price? Make compromises with regard to any of those three elements, and quality will inevitably be jeopardized.

In current discussions about human translators vs. AI (to be found on many translation companies’ websites and forums) much is made of particular “strongholds” of humans, i.e. those areas where human translators are considered to be indispensable and where AI simply cannot compete. Only humans, it is said, can bring special knowledge, special flair and knowledge of the world, which is required to produce a text that is flawless in all respects.

It is only natural of course that we humans should be proud of what we can do using our intelligence and intuition, and that we should defend our turf, our uniqueness, that which sets us apart from machines.

I have identified 8 categories where it is generally believed that humans are better than machines. In each case I have come to the conclusion that it may no longer be sensible to make a choice between Human OR AI. Instead, I would suggest it is best to engage in a collaborative approach between these two – to get the best of both worlds.

At Alpha we have been working on this during 2024 and will continue to make advances in this respect, as we hone our skills and gain not just experience but also confidence in our new best friend, AI.

1. Cultural nuances and sensitivities

• Machines are not great yet at understanding cultural subtleties, idiomatic expressions, humour and irony, context-specific meanings and jargon – how could they? Human translators will typically outperform them when it comes to adapting content to make it resonate with the target audience and respect cultural sensitivities.
• It is not only literary works, marketing materials and slogans that require adaptation (transcreation) rather than direct translation. It happens where the reader is being directly addressed as an individual, rather than just a generic user. The machine lacks that human touch. Little phrases like ‘Es tut uns leid’ or ‘bitte’ are often omitted by machines, which can make the text feel impersonal and negatively impact the user experience (UX). These small details are crucial for creating a connection with the audience. Sometimes you want a conditional clause, or a question in the target language, that may not be present in the source. Changes may include gender-neutral language, and choosing the most suitable form of address (formal/informal), breaking up long sentences into shorter ones, etc.
  • At Alpha, we are working on prompts and instructions defining style, tone of voice, register of the text; form of address, degree of friendliness, casualness, etc.

2. Ambiguity and Polysemy

• Words and phrases with multiple meanings require a thorough understanding of context to choose the correct interpretation. Human translators can analyze the broader context, they can use their commonsense, and make informed decisions, whereas machines may misinterpret context, leading to errors in meaning. For example, the word ‘bank’ could refer to a financial institution or the side of a river, depending on the context.
  • We provide our AI with as much context as possible regarding the subject-matter, the type of material, the audience, and the client’s intentions (if we have that information). We carefully check the “tricky” passages in-context.

3. Emotion and Tone, Register

• Capturing the emotional tone, style, and voice of a text is a complex task that requires human intuition and creativity. This is especially important in literature, poetry, speeches, or any content where the emotional impact is critical or where empathy is needed.
  • At Alpha, we identify texts that are best left to a creative human translator. Even in cases where human creativity is essential, AI can be a helpful tool for overcoming writer’s block or refining particularly challenging sentences. The key is knowing when and how to use AI effectively.

4. Subject-matter expertise

• Most people would say that specialized fields like legal, medical, or technical translations, require human translators. Only the can bring domain-specific expertise and the ability to interpret complex concepts accurately whereas machines lack the necessary know-how. While some human translators certainly are subject-matter experts, many are skilled generalists who rely on research and collaboration with clients to ensure accuracy. Just ask yourself: Why would an orthopaedic surgeon, a vet or an engineer double-up as a translator? Out of love, in their spare-time? Or a CFO or a marketing expert? (The most obvious reason is self-evident: such professionals have a far higher earning potential if they spend their time doing what they are best at).
  • At Alpha, we are keen to use AI to our best advantage, for example to extract key terminology, do the basic research. This then allows human translators to work on the all-important term base (when the client does not provide one). We ask for this to be approved by experts on the client side, then feed it into the engine..

5. Quality Assurance, accountability – and commonsense

• While AI tools can produce fast translations, they often require human review to ensure accuracy, logic, consistency, and appropriateness. Human translators can identify errors, inconsistencies, or awkward phrasing and mis-interpretations that machines might overlook.
  • For any texts that are important or intended for long-term use, we recommend a careful post-editing stage. We do not recommend ‘light editing’ for important texts, as it often fails to catch critical errors or inconsistencies. However, it can be useful for straightforward materials, such as updates to user instructions or simple Q&A websites, especially when the engine has been trained on approved materials.

6. Handling sub-optimal source text

• When the source is faulty (“funny” grammar, careless writing with poor choice of words, omissions, lack of logic, or factual errors, etc.) it can be difficult for the machine to make sense of a text. Sometimes it is difficult even for a human translator, particularly if it is a one-off job, with no context or reference material, or when there is no briefing or support from the client. Or if for reasons of “diplomacy” (personal sensitivities) it is difficult or impossible to ask for clarification.
  • At Alpha, we find that a combination of AI and translators works well in this situation. What often helps is a dialogue between “(wo)man and machine”. For example, it is possible to interrogate AI, saying: “This passage seems rather unclear, can you detect what the author is trying to say?” You can then have a discussion that will often provide surprising results, helping to deduce the author’s intended meaning – and a translation.

7. Adaptation/Customization for Target Audience

• Human translators can tailor translations to specific audiences, considering factors like age, education level, or regional preferences. This level of personalization is difficult for machines to achieve without extensive human input.
  • At Alpha, in such cases too, we work collaboratively with AI, using prompts and dialogs, to achieve best results, empowering AI to assist our translators for best results.

8. Creative and Artistic Translation

• In areas like literary or fantasy-game texts, film subtitles, human creativity is essential to preserve the artistic intent and ensure the translation feels natural and engaging in the target language.
  • At Alpha, quite frankly, we are often surprised how well AI copes with these kinds of materials that were (and still are) considered “off-limits” for machines. Of course, not everything can be used “as is”, but remember that humans too, often need to go over their texts, rework them, read them out aloud, etc. Even literary greats like Thomas Mann spent days refining their drafts. Similarly, AI can provide a starting point, but human creativity is essential for the final polish.

Humans who can shake a final version out of their sleeve are rare. Why should we expect a machine to do this at the drop of a hat? So again, when style matters and creativity is an essential ingredient, humans and machines must learn to work together. They can produce 3, 4 or more variants and arrive at the best possible solution.

Humans are better in some respects, machines in others. The best outcome, to benefit from the strengths of each, in my opinion, is when the two make a team. As in any relationship, this takes time and effort to develop. But it’s well worth it.