Hardly a day goes by without a new story about artificial intelligence (AI) and its effect on our lives. As language professionals, we are already feeling the impact of this rapidly evolving technology. On 7 March 2024, Southern SIG hosted an online meeting where we debated how best to persuade clients of the added value of a language professional. Using some examples of mistakes made by AI systems, the 29 participants enjoyed an open and lively discussion, sharing a variety of experiences and offering diverse perspectives.
We all acknowledged that some of the typical mistakes of the early large language models have, in fact, already been rectified. For example, the word-for-word style of translation has largely been eliminated by the newer deep-learning neural network models, which translate at the sentence level. Furthermore, as these newer models are exposed to more data, they continue learning and improving. This means that even more of the early advantages of translators over machines – such as an in-depth understanding of both source and target language, understanding context and nuance, and capturing the intended meaning and tone of a text (as opposed to literal translation) – have become or will become eroded. While discussing each of these points, time and again we found ourselves asking the same question: ‘Can AI do that?’ Rather than a convincing ‘no’, the answer was often ‘not yet’. And while there were different viewpoints on exactly when AI might be able to close the gap, there was little doubt that it would.
However, it was not all doom and gloom. Some members pointed out that certain texts require more out-of-the-box thinking or specialist expertise, and clients recognize this and prefer to hire the services of a translator or editor. It was felt that for certain fields, such as creative works, this is unlikely to change. Members also reported that many clients also realize that a machine translation is not an end product in itself and they use professionals to post-edit their texts. While post-editing seems to be one of those love-it-or-hate-it fields of work, it is an example of an area where the services of a translator or editor are still needed and valued, albeit only in some quarters. Those familiar with post-editing explained that the quality of machine-translated texts depends, not surprisingly, on a dedicated corpus and high-quality copy in the first place. There is no denying that there is pressure on rates for post-editing, and we discussed the importance of checking a text before deciding whether to take it on and, if choosing to proceed, also ensuring that the fee corresponds to the work involved. However, both of these stipulations already apply to translating and editing work in general. Those who enjoy post-editing were quick to point out that there are actually some advantages for the translator: leave the boring parts of the work to AI while you concentrate on the more interesting and challenging projects.
So, how do we market ourselves going forward? Fortunately, there were plenty of suggestions, many of which focussed on one theme: our human qualities. A translator or editor can collaborate with their client, ask questions, request clarification and check for meaning (some words just don’t translate easily into another language and can only be properly translated by having a discussion with the client). Language professionals use all their real-life experience and subject matter expertise to work with their client, not just translating or editing the text, but improving it. They ensure that the text is suitable for purpose and appropriate to the target audience. They can also differentiate between what is true and what is false and identify and correct errors in the source text. Unlike AI, a good translator or editor double checks and triple checks their work to minimize the chance of an error; although, of course, humans still make mistakes too. In short, language professionals offer flexibility, specialist expertise and partnership.
In addition, we might be able to address the area of cost in some cases. A common reason for using machine translation is the low cost. However, in certain situations, there are two strong arguments to counter this. Firstly, as one member pointed out using a marketing campaign as an example, the cost of a translator or editor is just a drop in the ocean compared to the total cost of the campaign, and it is far more cost-effective to avoid a preventable mistake at this stage of the process. Secondly, given that the improved AI models consume much more energy, if you factor in computing power to the cost of a machine translation, a human translator may not be quite so expensive after all.
Several different fields of work, language combinations and areas of specialization were represented at the meeting, each with their own set of challenges and opportunities. While some people are optimistic about the future, others are less so. However, we all agreed that the role of the language professional is changing and that we will need to adapt to the evolving AI landscape.
Note
With thanks to Taylor Steed for sharing her comprehensive meeting notes with me.
Blog post by: Linda Comyns Website: www.lmcenglishcommunications.nl/ |