In-house translation – outside the box

Category: LangTech

  • Does shadow translation present a bigger threat to language services than MT/GenAI itself?

    Does shadow translation present a bigger threat to language services than MT/GenAI itself?

    As MT and GenAI offer seemingly free and instant translation, “shadow translation” is a reality for in-house language services. There is an inevitability in corporate life, that prohibiting people from taking shortcuts, will not stop them from doing so. Similarly, even if there is a path, there will always be a desire path. Well-meaning rules, e.g. banning free MT tools will always be broken, unless sites or tools are blocked.

    Such practices are typical examples of shadow translation activities. In organisations with a team of in-house linguists, shadow translation activities often form due to the bottlenecks in translation via language services, i.e. the need for translation outside the official, governed translation process.

    Shadow translation serves as an unofficial, parallel or informal translation conduit. They are produced outside of authorised translation workflows. Shadow translations also evade capture in translation management (e.g. job lists, translation memories, or terminology systems). However, they remain in use in the wild by staff members, from trainees through to board members.

    What kinds of translations are vulnerable?

    But what kind of translations do shadow translation cover? Frequently it may be translations lazily labelled as “for internal use” i.e. for quickly getting into the meaning of a text, for writing something up etc. But, it might be used by staff, if someone doesn’t feel confident enough to draft texts (e.g. e-mails, summaries or speaking notes) by themself. In some situations, staff members may even delegate translation to the non-professional translator bilingual colleagues. This is particularly visible in job descriptions for PAs with an expectation of multilingual written communication.

    In the MT/GenAI era, there is the complicit use of MT/LLM-based translation, often without subsequent review (i.e. the translation, due possibly to its deceptive fluency, is passed on “as is”) or not disclosed as being a translation. However, it may include cases of externally translated content that sits outside of internal translation storage systems.

    This can also result in siloing in departments and divisions, meaning that there are multiple diverging translations in circulation of the same texts (for example in the case of standard letters, text blocks etc.) This is one reason why I translate each iteration of the standard texts that form administrative decisions for the entire organisation. I ensure that there is a single source of truth, and then only need to focus on the ruling (Spruch) of the individual administrative decision in hand. In process management, where processes contain sample texts about decisions and mails, I also make sure only to include a sample ruling. Other parts of such sample texts refer to the current template to use for the rest.

    Is this an issue for translators?

    Shadow translation bears numerous risks. From the perspective of a language services unit or a translator in one, the sensitivity is different to that experienced by the users. Users are likely to have a slightly more cavalier approach, and see it as a “victimless crime”.

    Unofficial translations, without quality assurance processes, may diverge subtly from the legally binding or authentic wording of a translation. To provide an example, I recently saw a Copilot Agent use case. The purpose of the agent was “legal translation”. The agent saved parallel texts of a piece of EU legislation in two languages as “knowledge” into the agent. The prompt in the agent instructed it to translate a specific provision from one language into another. Without the prompt stating the Spanish text was an authentic translation of the English text, the agent decided to create a back-translation for its English version.

    This can quickly create issues (e.g. like distorted messages in a game of Telephone / Stille Post). It leads to parallel translations that are close but not identical. Potential outcomes can be incorrect wording invalidating decisions or misrepresenting obligations. This results in exposing an institution to being challenged about their decisions.

    But it’s “only” for internal use, right?

    The problem with claiming that something is “for internal use” is a downstream one later down the line. Months or years later an internal text may suddenly become an external one. When this happens, it is like a document crossing an institutional firewall. Translation in language services can frequently be a “black box” for non-translators. Translators on occasion lose themselves in their own processes or their perceived reality. In the worst case, they might actually neglect that they are translating for a specific audience (e.g. the “Clientland” analogy that Chris Durban frequently alludes to). Translators need to also consider how they label translations – and how it will be (continue to be) used after they deliver it to the customer in-house.

    Another issue from shadow translation is that it supports terminology drift – with key terms gradually changing their meaning across documents, or where different departments resort to different equivalents for the same source term, or where informal translations take on a reference quality. Variants can slowly leach into accepted use, and are difficult to reverse. In the human-machine translation age, the effects may be more pronounced, as the training of models has shown, as once something is in a model, it can be very difficult to get it back out of a model.

    A question of accountability

    A further issue for language services can be down to the emergence of accountability gaps. The assumption might be that the error occurred in language services. Within language services, translation memory systems can resolve the team members involved. In turn, it will also identify the customer, reviewer and who signed off on the translator. Shadow translation can’t do that, other than identifying the string of communications. Nevertheless, language services may be caught in the crossfire. It can have the ability to undermine the authority of the official translation function in the blame game. Inconsistencies that they never contributed too nevertheless land at their door! Invisible translation output can also render quality metrics meaningless.

    Is shadow translation an accidental or malicious phenomenon?

    Ultimately, structural pressures are usually behind the creating of a burgeoning level of shadow translation. These pressures are due to time pressure (everything required now, or better still yesterday!) Available resources in language services might be creating the bottleneck – there are not endless supplies of human translators in all language combinations. AI tools seep out of every pore in corporate life (especially due to the “fear of missing out”!) unwittingly casting translation as being readily available on tap. Another phenomenon has been that of “multilingualism = translator” conflating speaking more than one language to being able to translate. Even children raised as bilingual/multilingual from birth will still gravitate towards one language over another in specific circumstances.

    However, sometimes the issue is down having a lack of clear rules, in order to distinguish working, informational, or legally or externally relevant translations. These issues are scaling silently as digital tools develop and potentially even improve.

    So this isn’t just about a translators trying to protect their jobs?

    When translators try to explain their continuing role in the GenAI-dominated landscape, they are accused of being protectionist. Persona prompts apparently can make everyone an “expert translator specialising in xyz in the language combination A to B”. But the institutional impact is more about governance rather than an out-and-out language one.

    Shadow translation circumvents document control, risk management, process transparency and accountability – a point missed by trivialising it as just being a discussion about quality. In this regard – it ties in a lot with the shattering of the triple constraint for translation, that I refer to in my blog posts about the “Expert in the Lead” approach. Particularly, given the fact that human in the loop approaches in translation do not adequately consider risk management and accountability.

    The “internal use only” caveat also is a risky approach. Internal documents (e.g. briefing notes) actively shape decisions, which in turn shape actions, which themselves have external and legal consequences. And internal translations can end up becoming de facto reference documents.

    Where there is an added dimension due to AI and MT is due to the fact that such tools also play a role in making shadow translation faster, more tempting to use, and harder to detect. Prompting courses invariably mention using prompts to try to “make the text not look like it is AI-written”. This is why institutions need clear policies to avoid uncontrolled MT output, inconsistent style and terminology, as well as false confidence and over-reliance on “good enough” MT output.

    We need to talk about cognitive load and cognitive downstreaming

    I increasing mention the concern about cognitive downstreaming. Offloading (linguistic) effort to tools or other individuals downstreams review, validation and responsibility involved. If you use a linguist only for the review stage, they have to try to make “informed” decisions without the full information – I’ve started to receive “Copilot translations” (the quotations are intentional!) in English – without being provided a source text. I enquire where the original German text is. After all, if it is a translation, it must be a translation of a German text.

    Sometimes it even emerges that the text was in fact prompting output. However, even the detail in the prompt provides me with information that is missing from the English, e.g. about the legal provisions forming the underlying basis (and quickly highlighting that there has been an unnecessary back translation). In the worse cases, the output has read like LLM output, where the German original also doesn’t respect the house authoring and formatting style.

    What should institutional practices do to get a grip on shadow translation?

    There is a clear near for greater formal distinction in the classification of translations as an approach to shadow translation management. Labelling is an issue that is emerging with the automation of translation of web content. I am fully aware that bad human translations also exist! Take a look at many multilingual tourism websites for private accommodation if you want!

    One potential categorisation could be as follows:

    • authoritative translations (binding, publishable)
    • working translations (informational, internal)
    • machine-assisted drafts (explicitly non-authoritative)

    Within each category, the permitted uses, visibility requirements and review expectations also need to be specified.

    Let there be light! Visibility of unofficial translations

    The way that shadow translation has shown that banning a tool only means that someone will look for the next tool to circumvent the ban, is possibly a sign than a more enlightened approach is to not ban shadow translation (as a process or its output).

    I’d suggest destigmatisation by recommending to improve their visibility. However, there would be some requirements, namely:

    • enforced use of labelling (e.g. also involving the categorisations mentioned above).
    • requiring storage in a shared environment, with the source text available in the same location.
    • compulsory disclosure about the use of MT/GenAI (this can also help reduce cognitive downstreaming).

    Favouring a visibility approach, over prohibiting shadow translation, can also help ensure a steady work for professional translators. For example, quality scoring of shadow translations can form a ranking for permit the prioritising items for translation to ensure greater availability of authoritative translations.

    Visibility on the other hand, can also mean improved centralisation and availability of language data (e.g. from translation memories or terminology databases). Two such approaches are what RWS calls “generative translation” or Kaleidoscope calls “TAG (terminology augmented generation). I’m hoping to be able to convince my employer to allow more effective terminology sharing than in the old-fashioned “bilingual glossary”. That product is no longer cutting edge, and is high maintenance. Each update consumes a lot of time for a product in an old-fashioned format.

    The role of translators in AI governance

    Back in 2022, I joined our Innovation Lab, and chose from the outset to take a “voice of reason” approach, which would question use cases, AI governance issues and the like. I tried to be open to innovation, while also ensuring that the recognition of the continuing importance of human translation . They recognised and understood my valid concerns. I emphasised that I was risk aware and possibly somewhat risk averse. Labelling myself the “Spaßbremse” (spoiler of the fun) prevented anyone else from using it in an accusatory tone.

    I contributed towards my institution’s AI governance – by highlighting the need for review, and highlighting issues of accountability, reputational loss, and credibility. More importantly, it guaranteed my seat at the table. Langtech solution providers currently struggle to be heard. BigTechs demonstrate their (lack of) knowledge of translation and seem hell-bent on ignoring incumbent niche players. In this regard, my approach was about being heard rather than just to be an end-user.

    I’ve also used it to cultivate my role as a “risk sensor”. Deceptive fluency of MT output frequently fails to identify inconsistencies in source texts, ambiguities in policy, legal uncertainty and procedural gaps. Trying to bring the issue of “shadow translation” onto the table is also a case of highlighting that it bypasses governance, rather than my being protectionist or territorial.

    So what about a bite-sized takeaway?

    Shadow translation is a hidden reality of any organisation with a multilingual environment. Failure to govern shadow translation will lead it to self-govern. Self-governance is inconsistent, invisible and allows the propagation and permeation of risks.

  • Who’s in/on the lead in early 2025?

    Who’s in/on the lead in early 2025?

    In December 2023, I wrote about the state of human-machine translation as we headed towards 2024. The technological march of machine translation had dominated 2023. From personal musings over the last twenty years, comparing my professional situation with 12 months before, for the first time in 2022 and 2023 my outlook was a more pessimistic one in consecutive years.

    My view was one of human translators being pushed towards fighting for scraps at dumping MTPE rates. More were considering moving away from translation or increasing their other activities than moving towards focussing solely on translation. In house, I had started to receive editing and revision requests that “didn’t seem quite right”. They seemed more fluent than their authors’ previous drafted texts, but also weren’t quite factually correct. In other cases, the inconsistency of terminology shone through. The sinking feeling was that my own descent towards MTPE drudgery had begun. The profession shared my pessimistic outlook. The fragility of (self-)employment relationships, needs for efficiency and cost-cutting amid difficult financial times were also apparent.

    2025: a turbulent start

    When I started sketching out this article in mid-December 2024, I didn’t know how 2025 would begin in terms of technological announcements. DeepSeek was not on the radar – by the end of January it was everywhere. Possibly from a translation professional’s perspective the most interesting aspect was OpenAI’s complaint in late January that new upstart DeepSeek was using “its” data. That’s right, the same data that OpenAI itself had unashamedly scraped to train itself. Excuse me Mr. Altman while I locate the sub-atomic-sized Stradivarius.

    In recent weeks, I’ve read a number of people saying that this could be positive for easing OpenAI’s (perceived?) monopoly. For many, ChatGPT has become a metonym for AI. Others think it could herald a torrent of new solutions – some fear one that might finally be able to translate (impacting their endangered volume of translation work and pushing them further towards MTPE’s clutches). And that was before the latest development of Elon Musk expressing his wish to buy OpenAI.

    The schism between the translation industry and the translation profession

    The trend of recent years of a divergence in approaches between the translation industry and the translation profession continues. It had been a pandemic edition of the Translating Europe Forum (TEF) that first pushed the Human in the Loop (HITL) agenda. At first sight, its deceptive allure took me in. Over time, I became aware of the weasliness of the term “Human in the Loop” for translation. HITL is misused: it fails to define the expertise level of the human, and does not advocate the human retaining control/leadership. The industry seems to be revising its estimation somewhat with new term “Human at the Core” which is closer to my “Expert in the Lead” approach than “Machine in the Loop”, but is still coined by the industry. My “Expert in the Lead” concept is also about coming down on the side of the profession over the industry.

    Fresh hope from the industry?

    A piece from late in 2024 by Arle Lommel for CSA did give me some hope that the industry is also coming round to the fact that HITL will not sustain human translators in the human-machine translation era. One remark in that piece captures why HITL gets it wrong, and how that “janitorial role” of HITL will not be fulfilling.

    “[…] “human in the loop” models – a sort of window dressing for post-editing – … often relegate expert linguists to an essentially janitorial role, sweeping up “bad MT” (quality checking and correction) and cleaning up AI messes. Instead, CSA Research has shifted to describing augmented translation as “human at the core” because, at the end of the day, empowered linguists will be making the decisions, aided by technology.”

    The Language Sector Slowdown: A Multifaceted Outlook, Arle Lommel for CSA Research

    Looking back to my assessment in December 2023, I opened with the following paragraph:

    The debate about the future of (human) translation and changing role of translators is the biggest topic in translator circles. 2023 has been the year of the (unstoppable?) march of machine translation. Within a year of bursting onto the scene as an unknown, OpenAI’s chatbot, ChatGPT, can apparently also translate. Human translators increasingly face tighter, more competitive markets. Many are not even consulted about their replacement by MT solutions, but maybe grudgingly offered MTPE work. And there are talks of tightened budgets and gloomy outlooks of recession. So are the days of out-and-out translators numbered?

    Michael Bailey, transl8r.eu blogpost – December 2023 – Who’s in/on the lead as we head into 2024?

    As I prepared to write this post, I asked fellow professionals over LinkedIn how they viewed the situation. A modest little poll on LinkedIn among my network of fellow translators returned a slightly blurry snapshot. I asked pretty much the same question as I have been asking myself over two decades. From over 80 responses, less than one quarter of responses viewed their situation more optimistically. In contrast, 45% view their situation more pessimistically, and the remaining third see it as unchanged from the previous year. From those responses, a number of in-house translators and specialists in less common language pairs seemed more optimistic. Of the positively inclined, many were offering premium services with a narrow specialist focus. A few reported that new areas of specialism had emerged that compensated for the slowdown in business in other areas.

    Busy-ness and Business

    Some responses mentioned improved levels of “busy-ness”, but qualified the improvement being due to time-consuming customer acquisition drives. For others, new services and specialist areas had arrested the slump, but hadn’t banished doubts about the long-term future. In a few cases, new revenue streams opened up from (re)activating new language pairs, although a number I connected with did not realistically view adding further language combinations as a potential solution. Others viewed that the situation was no worse than a year ago, but had also not improved. For some of these this kind of struggle was a “new normal” – the glass was neither half-full nor half-empty.

    Of those viewing the situation more pessimistically, several commented about an acceleration in the shift towards MTPE from “pure” translation work. Many freelancers lamented that their “valuable and valid” contribution was unable to outweigh their customers seeking “value for money”. By value for money – they mentioned diminishing rates (whether by line, character or page) or more MTPE work. A couple also said that work from major agency clients drying up had impacted them. In other cases the agencies had shifted towards an MTPE-based model instead of “classical translation”. Some others mentioned that reorganisations and mergers had meant that major customers had already reviewed the situation. A couple of respondents mentioned that smaller companies had been absorbed into larger groups with in-house language services.

    Payment Practices

    One contact also said that their pessimism was fuelled by longer payment times, although still within the agreed timeframe – a potential sign of agencies also suffering from cashflow issues. Amid ongoing cost of living issues (price inflation outstripping wage/salary increases, or downward pressure on rates), the financial squeeze becomes more apparent.

    By delaying this post, I wanted to also allow myself the opportunity to catch-up with the first swathe of “Monthly Recap” posts on LinkedIn in 2025, in addition to “year-end round-up” posts. I’ve come to appreciate that it has been a busy month if I don’t have time to even consider writing one. However, this is where internal time and performance tracking negates a need for such a round-up. In 2024, H2 showed a remarkable up-tick: in May, based on figures until the end of April, translation time was around 73% of productive hours. By year-end it was up above 80%. In addition, my worked hours were higher for translation in 2024 than total hours for 2023.

    How do you feel about the security/future of your role as a human translator, compared with 12 months ago?

    These figures are why I see the security/future of my role more optimistically going into 2025. But this might be due to the short-termism of recent successes masking and negating struggles earlier in the year. Looking back at the reasons behind my pessimism in the last two years, uncertainty weighed strongly on my mind. Transformation and reorganisation bring uncertainty and insecurity. As a digital transformation programme started, I had felt marginalised and sidelined. And I felt that remit creep was also disruptive for my “course” as a translator. So while doubts existed, along with the simmering AI hype, I remained pessimistic. Learning about a suggested roll-out of MT without harnessing our language data probably fed the pessimism. So what changed so much in twelve months for me to enter 2025 with renewed optimism?

    Getting back to business

    In previous years, the non-translation-based tasks I was logging increased. I advocate that 100% efficiency/productivity is an illusion, as is 100% productivity as a translator. However, translators are susceptible to worrying about a dilution of their time spent translating. At year-end 2023, my productivity tracking showed I was translating for less than 80% of my hours. When I started the job, the level was closer to 90%. I felt a need to arrest the drift towards my knowledge-based job becoming a non-translation-based one. So I enlisted the services of a coach, and focused on using my mid-year appraisal to shed some non-core commitments. It was a timely reboot, and boosted my translator’s esteem. Esteem is so important.

    Translating for a predominantly “non-public” domain means that a lot of my work’s impact never reaches the outside world. Internal visibility is therefore very important. Fortunately, the second half of 2024 served up a plethora of demanding, substantial and internally visible jobs. As a translator I still feel happiest translating, although I can use non-translation tasks to draw breath. I’ve learned to fuel my internal visibility. I am most visible where my translation results in the desired supervisory outcome at short notice. Internal visibility also builds momentum, as has been the case going into 2025.

    A public or private persona

    As wonderful as a very private persona sounds for less gregarious translators, I nevertheless need to maintain a public presence. Presentations and publications (e.g. in the ITI Bulletin and Universitas Mitteilungsblatt) also bolster the public impact of my work as a translator. The workshop I gave in Spiez and the contacts gained there were crucial in a lot of self-esteem issues. Three days’ reflection proved a turning point for “getting back to being me” and to steer out of the doldrums of silo-thinking. As I put the final touches to this piece, in 2025 I already have three further presentations confirmed, a conference participation and other irons in the fire.

    In silo-like environments, especially for the “lone rangers”, i.e. SPLSU in-housers like me or freelancers who do not work together with other translators in virtual teams, social media can become an ersatz barometer of success and a way to shout from the rooftops. The problem is that the algorithms can suck you in, but don’t pay the bills. Add the peacocking influencers to the equation and they will tell you to post hourly/daily/weekly to feed the algorithm. However, my work’s confidential nature means that I can’t get sucked in by the siren-like call of the algorithms. I don’t have the fear of missing out that a freelancer has, if they don’t take on a piece of work. And much of the messages are about the successes – after all you project success far more than failure.

    How are others feeling?

    From some of the end-of-year posts I read, some professionals certainly put in the hard yards and enjoyed exceptional years (in terms of acclaim and remuneration) in 2024. To them: congratulations – your messages show that there plenty of life in professional translation. From viewing their profiles and websites, they all specialise in certain language combinations and with some very interesting niches. The common key to their success also seems to have been their efforts in fresh customer acquisition and keeping customers.

    Some found that new areas of specialisation were opening up: either related to their existing areas or fresh new areas. Others pleasingly reported old customers feared lost returning to them after a dalliance with the AI/MT “good enough” world. For every success story, however, there were also stories of people having lost customers and work drying up. In some cases there were cases of agencies folding owing translators money. One such case was the bankruptcy of WCS Group and the agencies it ran (subsequently bought by Powerling). Many freelancers were left out of pocket. As I added to this post in mid-January 2025, there was a new twist to the Powerling story: The Dutch Society of Translators has just expelled Powerling from being a member. (h/t to Loek van Koeten for this information).

    Upskilling and job crafting for survival?

    Before I was able to actually narrow my remit, I had had to consider upskilling (i.e. obtaining alternative skills to complement my skills as a translator) and even put my foot in the water in actively pursuing courses to be fit for the new world of human-machine translation. However, obtaining new and possibly diametrically opposed skills to those I already possess as a translator proved counterproductive. Instead, with new areas of supervision coming online, my focus has now reverted to deepening my breadth of knowledge in the subject areas I cover. Some translation professionals have echoed this: those who will survive already possess all the skills and specialisations to survive.

    Teaching old dogs new tricks?

    Regarding the prospects of who will survive the AI deluge, I’ve read numerous estimates about the proportion of translators who will “survive” the AI revolution, with many stating between 10 and 25% percent, although the range is far wider. Part of the issue also relates to the stage of their career that translators are at. As William Lise identified in a blog post of his, some are close enough to retirement, and others young enough to change position. However, there is a substantial group of translators, particularly mid-career ones, trapped by the roots they have put down.

    Whether people who have retrained from other professions are any safer is hard to tell. They may bring expertise from a past career, but may lack the translation experience. Possibly being newer in the “trade” might work both ways: be more firmly tied to making it work as the cost of retraining hasn’t been recouped yet, or in contrast, not so firmly embedded in the profession that they can’t “get out”. From a number of contacts who always viewed translation as a “safe Plan B”, they’ve changed their minds about wanting to commit to it.

    Expertise counters AI hype

    Nonetheless, the reality after the tidal wave of AI hype has proven that expertise remains essential – accountability and credibility of translations are areas where human translators still have an advantage. AI and NMT flushes out generalists working for agencies and pseudo-specialists. In this case, broad fields of specialisation (e.g. financial/legal) for agencies maybe stops people from standing out from the crowd. Others say they experience agency work decided upon purely by means of “fastest finger first” – an issue I mentioned when I blogged about the profession/industry schism in autumn 2023. In that case, expertise is unlikely to be given a chance to shine through.

    In contrast, genuine specialists in narrow fields remain an elusively rare commodity. Regarding AI, there is a healthy scepticism about how it can really be a substitute for expertise and experience. Simply throwing more scraped data at the problem isn’t the solution, particularly as synthetic language data now swamps the originally lush large language pastures trained on human generated language. In this regard there is a counter revolution of some boutique LSPs looking for high-end translators whose personal service commands premium rates. In a couple of cases, some freelancers have even reported that they have profited from customers turning to them due to unsatisfactory agency experiences, viewing them as a “perfect fit” after lacklustre past experience.

    And when the boot is on the other foot?

    Occasionally, I outsource work to freelancers. The objective remains to ensure the desired supervisory outcome. This also sheds a lot of light on the “black box of translation”, market practices and how solid briefs helps so much. I have come to get a good feeling whether translators 1) want the job and 2) feel they can do justice to the job in hand. Genuine experts seem less fazed in not being able to take a job on. I also admire their honesty. Such a situation might be vastly different than dealing with an agency, where selling and margins are everything. The requirement of a satisfactory outcome, allows me to use a best bidder approach, rather than a cheapest bidder one.

    Capitalising on AI’s vulnerabilities

    Amidst the OpenAI/DeepSeek saga, I used the opportunity to highlight the accountability, control and expertise that expert human translation offers that AI and MT cannot. When “data scraping” allegations surfaced, I chose to capitalise on highlighting data confidentiality. My approach for the aficionados who brazenly claim how much time their ChatGPT Pro subscription saves, is to ask how they feel prompting techniques have changed, robustness of sources, and their views about the size of the context window.

    The disarming tactic is to speak the fanboy’s language rather than coming across as too protectionist. Only then do you highlight the issues that impact your translation work, and therefore confirm why your expertise is required (e.g. in a zero/low-resource language combination, with high demands on confidentiality, and the necessary to avoid hallucinations).

    Changing job remits

    In terms of job creation, I’ve observed a tendency towards not replacing departing staff, or at best retaining existing headcount. New translator jobs are seldom. Looking at job descriptions, may advertised positions have been for maternity cover positions, often initially limited to a year. It can easily take a year to get to grips with new procedures, practices and subject areas. Other vacancies have more of a project manager/coordinator role emerging rather than a “translator” remit.

    Monitoring open opportunities (I receive them through mailing lists from professional associations) is useful for gauging remit shift/creep. Job descriptions have clearly changed. Jobs creation rather than replenishment occurs in the area of LangTech. New LangTech units in larger language services are in-housing expertise. From conversations with people fitting the new profile, many highlight prominent “sponsors” within the organisation and strong links to IT being behind the creation of the new position.

    Managing language data has definitely become more than a “rainy day” activity – as has terminology work. In a small language services unit, terminologists were traditionally considered a luxury. With the advent of Machine Translation, robust terminology has gained in importance. Machine translation-generated texts into German have demonstrated why I need terminology for all locales of German. My recent work has really brought home the differences between Swiss/German/Liechtenstein/Austrian banking terminology.

    Driven loopy – the expert/machine/human in the loop/lead.

    As previously mentioned, the very strong industry-led approach to human-machine translation is of “machine in the loop” and “human in the loop”. The industry’s financial and PR clout dictates the way translation (both as an industry and a profession) moves forward. However, industry-led perspectives focus on leveraging technology to an extent where human involvement is negligible or a poorly-paid afterthought.

    This is quite apparent from the shift in the industry from humans predominantly “translating” to “post-editing”. In some cases the actual level of human expertise in the post-editing stage is questionable. Pitiful rates fail to motivate a professional: low per word rates for MTPE require unrealistic output levels to earn enough. It would take raw output pretty close to publishable in the first place that you can simply sign off. However, this realistically only works where translation is only required to be “good enough”. And the long-term job satisfaction of this approach is also negligible.

    The HITL narrative is pushed so far that the MITL approach barely gets a look in. Rebranding translators as “language experts” is a mere sop. In much the way that the electorate in the UK may/may not have “had enough of experts”. “Language experts” is just another weaselly term: genuine expertise may often be found in far narrower areas or a single source-target language combination. Imagine the (justified) outrage if we were to rebrand microbiologists or astrophysicists as “science experts”.

    Throw more language data at it?

    The fact is that amid Messrs. Altman et al. scraping the Internet for content to build their LLMs, human generated language data has been exhausted. Tech bros continue to recite their “more data = better results” mantra. The synthetic data has already flooded the Internet, creating new “reheated” synthetic language data. All that changes here is the consistency of the turgid porridge.

    The “more data = better results” approach is like a juggernaut or steamroller, or raging waters trying to pass through a pipe of a certain diameter. Upgrading pipes might permit a greater volume of waters to flow, but unless done end-to-end the flood risk still exists.

    Many AI companies are still a long way from break-even let alone posting profits. This raises ethical questions. Why should we allow tech companies to break human knowledge-based industries, accelerate climate change, only to line the pockets of the super rich, if they ever turn a profit? Industry dictates the terms: amid skewed arguments of increased efficiency, knowledge-based work is still fraught with “hallucinations”. Why should translators tolerate such hallucinations?

    Resistance is (not) futile?

    My view about the Expert in the Lead results from my conviction that the role of the human in human-machine translation remains essential. I do concede that the days of “human translation” from the formative days of my career are gone. Instead, rather than resist the use of technology, the emphasis has shifted to ensuring human expertise remains in control. For me, this involves making the smart choice about the use of technology, rather than rejecting it. Experts in human-machine translation can resist by refusing to have their workflows dictated to them. Refusing to be a cog in the process keeps them in the lead rather than in the loop.

    My bespoke service revolves around my correctly blending multiple translation memories (setting those penalties in relation to age of TUs, subject matter, incorrect locale/language variation) and really knowing what the translation is about. At the same time I also can make a sound decision about the sources of reference material to access. This has far better chances for meaningful and fruitful success, than the drudge of cleaning out the stochastic parrot’s sodden cage from an LLM prone to hallucination.

  • What are the values of an Expert in the Lead?

    What are the values of an Expert in the Lead?

    In my recent article in the latest edition of the Universitas Mitteilungsblatt, one section covers the “Expert in the Lead” (XITL). XITL is a concept that has attracted a lot of my thoughts in recent months. It is my approach for considering the future role of human translators in the era of human-machine translation.

    I am currently also running a poll on LinkedIn (still open at time of publication of this post!). It asks people to assess their personal security/future as a human translator compared with 12 months ago. Why the comparison with 12 months ago? This relates to my December 2023 blogpost “Who’s in/on the lead as we head into 2024“. I want to follow-up on this later this month based on poll responses. I am also asking some respondents what might lie behind their response to the poll.

    In the Universitas article, I highlighted some values of an expert in the lead, which I have expanded upon here. The list is not exhaustive – I really welcome your comments!

    Being in command of technology and rejecting over-commoditisation

    1. Being technologically agnostic/neutral: The expert in the lead knows when and how to make use of technology. And similarly when not to. Consider the useful tools, but prioritise the human expertise aspect. Stay open to new ideas and innovative approaches: e.g. penalties for TU age, or using QA checks to reduce cognitive load burden. However, you call the shots when, where and how technology is used, rather than being in thrall to it. They decide which tools are used, not just the one that is the flavour of the month among LSPs. By all means use technology, but also know when not to.
    2. Rejecting the concept of translation as a commodity: in the race to the bottom, translation has become (excessively) commoditised. Boiled down to a number of words, characters, lines or pages. Then discounts squeezed for use of CAT tools, repetitions, or event the reduction of the (not-necessarily expert) translator to an MT post-editor. In contrast, the expert in the lead nurtures the customer relationship to understand what the customer needs. Pricing reflects the need for feedback rounds, terminology work, fine-tuning the brief and delivering what the customer wants and needs. (For example check out my thoughts on Chris Durban’s talk in Spiez this year – and the need to visit Clientland).

    Know your customers and audiences

    1. Convincing decision-makers about the value of human translation: the expert in the lead is on an equal-footing. Their professionalism commands respect. When I outsource a translation, I actively look for the best fit for the job. I take the blend of specialisation, experience and their passion for the subject matter into account. I do not try to beat them into a corner over pricing.
    2. Understanding your target audience: the expert in the lead takes the time to clarify with the customer in advance who the audience (e.g. the readership) is. Taking the time to settle on a strict brief in advance leads to a more satisfactory outcome for both sides and helps you to engage with your customer.
    3. Knowing when/how/when you should be used: Sometimes customers might have multiple translation needs. On occasions, a gist translation might suffice, or editing and revision. Get them on board for where they really need your full premium service – e.g. for handling their public-facing translations. Sometimes, you need to learn when to say “no!”

    Expertise and specialism instead of narcissism

    1. Convincing by expertise rather than social media presence: No-one “has to post on LinkedIn”. And a decent translator will not need to dedicate considerable office hours cultivating a social media presence. I am active on social media, but prefer to engage on other posts rather than post myself. Social media doesn’t pay my salary. And besides I struggle with its narcissism: where it is all about the “upside”, and never the downside. I’ve now settled on an approach of applied concerted laziness on LinkedIn. Know how and when to reach the people you have to reach, and how to use indirect visibility. Sometime you just need to “know how not to use LinkedIn incorrectly”.
    2. Being passionate about your expertise: sometimes your customer may not be sure that you really know what their request is about. I convey my expertise – and passion by engaging with a legal reference (e.g. the law or a provision in it) as an ice-breaker. Invariably, it shows we’re speaking the same language (even if I am translating it into another target language). Demonstrate your specific expertise within a broader field of expertise.
    3. Placing value on expertise-related training and education: conference programmes frequently strike me as too broad or general. To attend a conference, I need to convince my employer why I need to participate. Otherwise, I attend privately (at my own expense, conditional on being allowed to include participation on my CPD log). I struggle with the esoteric sessions – and prefer 1:1 online coaching for that purpose. Instead, I champion relevant expertise-based training. I focus on specialist training to increase my expertise – and realise the gaps in my knowledge from others’ questions. And I ensure that takeaways from conferences apply to my actual daily work.

    I’ve not touched on the area of the role of translator accountability, but this is an area I intend to look into further in the future. I see it as an increasingly important area for the professional translator.




  • Who’s in/on the lead as we head into 2024?

    Who’s in/on the lead as we head into 2024?

    The debate about the future of (human) translation and changing role of translators is the biggest topic in translator circles. 2023 has been the year of the (unstoppable?) march of machine translation. Within a year of bursting onto the scene as an unknown, OpenAI’s chatbot, ChatGPT, can apparently also translate. Human translators increasingly face tighter, more competitive markets. Many are not even consulted about their replacement by MT solutions, but maybe grudgingly offered PEMT work. And there are talks of tightened budgets and gloomy outlooks of recession. So are the days of out-and-out translators numbered?

    The Chartered Institute of Linguists, which I recently joined, has released a white paper: CIOL Voices on AI and Translation. It addresses some initial reflection and major concerns. The White Paper points to a shift in professions: today’s professional translators will be the future’s language experts and consultants. Sometime new job titles are dismissed as a case of “old grapes in new bottles”?

    The introduction to the White Paper concludes:

    […} we can ensure that linguists remain at the forefront of AI integration in our field – the essential expert ‘humans in the loop’.

    Steve Doswell, Linguist, consultant and Chair of CIOL Council in CIOL Voices on AI and Translation

    The use of “expert ‘humans in the loop’” is telling here. Without attaching the “expert”, it implies that an involved human may not be an linguistic expert. This ties in with concerns about the need for human judgement in using MT and LLMs for translation. It remains essential that users clearly understand their responsibility, as well as the pitfalls of using unsupervised MT. In-house language units must have an active role in training and onboarding users. Their involvement in the decision-making regarding the adoption of such approaches remains essential. It is not an out-and-out IT decision – even if the technological nature of the solution, means IT must be on board. There is some very sensitive messaging in moving from a “human translation” approach to “human in the loop” if bypassing the intermediate “machine in the loop” stage.

    Potential for upskilling and job crafting

    This presents possibilities for upskilling and job crafting – both useful tools for in-house staff retention. New remits might help retain senior staff members wishing to have a change from day-in-day-out translation. Any in-house solution will need dedicated language technologists. Language technologists are the new translators in terms of language services recruitments. Central banks and financial market supervision authorities have been hiring people with this profile for several years.

    It is also important to remember that for any solution to work to its full potential, will need dedicated staff. The quick and dirty approach might be to outsource, but such solutions, although quicker to implement, may not allow the desired level of control. An attractive interface is one thing, but there might not be the possibility to tweak the temperature of the underlying model, or to train it to your specifications – which is beneficial to extract the maximum benefit for your use case. However, this training isn’t possible on the fly – it needs a long-term training concept and commitment. And naturally potential succession management issues need handling too. These issues may be due to sabbaticals, secondments, retirements and maternity leave. Entrusting an entire solution on a single set of shoulders is also an operational risk.

    In this case, human involvement is still in more of an expert capacity – training and refining the engine, and ironing out the wrinkles. (Rinse and repeat as required!) Other tasks include managing new versions of software and interfaces, or plugins to CAT environments and maintenance. With an outsourced solution the situation is not so clear cut. This brings us back to the issue of the position of the human expert in the loop – and whether human or machine is subordinate – in the translation process as a whole, and the problems with the terms used.

    Driven loopy – the expert/machine/human in the loop/lead.

    I first heard of “human in the loop” mentioned at the 2021 edition of the Translating Europe Forum (TEF). TEF is the European Commission’s annual translation *industry* event. Over the last two years, I have lost count of the amount of discussions I have had with other people, about it. The problem it throws up lies in the interpretation of the role of the human.

    Moving further back, human-in-the-loop in 2012 was a classification for autonomous weapons systems. In that context, a human must instigate the action of the weapon. Human-on-the-loop is a classification whereby a human may abort an action. Lastly, and most terrifyingly, human-out-of-the-loop is the classification for no human action is involved. In this case human-in-the-loop does not imply that the human is subordinate to the machine.

    An intermediate stage exists between human translation and human in the loop: “machine in the loop“. In that case the machine is subordinate to the human, or more likely an expert. Both “machine in the loop” and “human in the loop” are weasly terms. Both fail to mention the role of human expertise – which is why some prefer “human at the core” or “human in the lead“. Additionally, one experienced colleague recently pointed out on LinkedIn that anything “human” omits to say anything about their expertise. This is why I actively try to opt for “expert in the lead” (should that maybe be EITL or XITL?).

    There can be a lot of difficulties in explaining the delicacy of the situation to lay colleagues – they see a binary situation: human translation or machine translation.

    After all, If you are not in the lead, but only in the loop, then you are effectively “on the lead”. And naturally there is the issue of the subsequent drift from human in the loop to human on/out of the loop. In that situation, we’re in the territory of fully autonomous self-driving vehicles.

    Resistance is futile?

    AI technology is clearly here to stay. While there is a certain hype cycle, it is not just a passing fad. The truth is that its limitations are well recognised: AI/MT cannot be used unsupervised in many settings. There are possibilities that the enhanced use of technological assistance might also open up new seams for translation (MT is a suitable use case for e.g., translating Airbnb and travel site reviews where a gist translation is what is needed). Humans will remain an integral part for training the underlying systems. Otherwise, at some point there will only be synthetic data to train systems that require high quality human data for training. Increased efficiency needs to be offset against the lack of job satisfaction that some will experience from being relegated to post-editing.

    Resistance to the advancing AI/MT tide is futile – both in-house and as freelancers. The battle to fight is in educating and countering assumptions that the lay public holds of machines being better, faster and cheaper. People need to understand the real risks and costs. However, part of this battle will also be to ensure that the current cohort of translators/language consultants/language technologists in the making learn the skills they will need for the career of the future. Many university courses adapt to the changing times at a pace observed in glacial creep. This is where professional associations come in – both in upskilling existing linguists, but also in supporting the next generation as it begins its journey.

  • It’s good to talk about LangTech – and to attend physical meetings

    It’s good to talk about LangTech – and to attend physical meetings

    I’ve just returned from a work trip to Athens, my first in-person away day since before the pandemic. I’d last attended a conference in person in 2019, and an away day of my working group back in 2017. A big takeaway has been how good it is to talk face-to-face.

    Hybrid and virtual meetings really cut out the opportunity for one-to-one conversation. You log on with a minute to spare, there is no chatting during breaks often only the length of a brief comfort stop. I felt very fortunate that my hosts also arranged dinners for the participants on both evenings. The hospitality extended at the meeting itself was also conducive to being able to talk to colleagues from other institutions. The breaks for coffee and lunch allowing me to speak to several other attendees. Even quick chats to the colleagues sat either side at the beginning of the meeting had a novelty value. Compare that with the awkward silence of hybrid meetings, where at best there is a round of salutations before silence. With presence meetings you are consciously “in the room” the entire time.

    Hybrid and virtual meetings have killed off one-to-one conversation

    Many hybrid and virtual meetings have an unwritten rule not to use the chat function during the meeting. This makes options for conversations with other colleagues very limited. Online meeting fatigue has also meant that meetings have been pared back in length massively. People attend for the bare minimum time and with cameras off (to conserve bandwidth rather than not wanting to be seen).

    By travelling for the meeting, I felt like I was attending the meeting with a renewed purpose. I was eager to talk to as many of the other participants as possible. I knew many for a number of years from presence meetings, but we had sparse contact during the pandemic years. Even sessions with less direct relevance for me than others provided interesting comments. Some are already due to flow into my internal Language Services Handbook (LaSH). The LaSH is a living document compiled since 2017 and that covers various aspects of language services. It addresses issues like best practices, lessons learned from past experiences, technical issues resolved, and handling procurement processes.

    And break!

    Scheduled breaks in meetings allowed me to talk to other participants. Their situations range from those encountered by fellow SPLSUs (single person language service units) who I ally with, and where we discuss how we manage without a team, and the challenges of smaller language units. With larger language service units, I frequently talk about how job remits change, We talk about team members upskilling, diversification of activities, changing trends in job types, and the changing profiles of linguists. After all, larger teams have more options to look into new areas, and for individuals to “personalise” their position. Most importantly, it is a really important chance to talk about LangTech. Within the group, practically all the language services represented use Computer Assisted Translation (CAT) / Translation Memory software, as well as terminology software. This is our starting point

    We need to talk about LangTech

    The real hot potato is about the increasing use of technology in language services. LangTech is ultimately already an essential part of language services. During my time as an in-house translator, I have witnessed the neural machine translation (NMT) revolution, the use of natural language processing (NLP) and AI in language services, and now the rise of large language models (LLMs) including ChatGPT.

    Some of the colleagues I met with also use additional LangTech solutions and are further along with deployment of machine translation. In some cases, they even have in-house language technologists who work on finding the system(s) that suit(s) their needs. More often than not, there is more openness to talk about what hasn’t quite worked out, or has perhaps not proved as successful as hoped.

    The LangTech debate is an interesting one, in that I could be an ostrich and stick my head in the sand and cite (over-)busyness as a reason not to look into it further. However, at the same time, if there are some tasks that it is able to perform and help with that increase my productivity, or alternatively free up translation capacity, then this is a churlish approach. I also cannot afford to rest on my laurels, and be late to the party. After all, certain tasks are currently very time-consuming for relatively little return. For this reason, some alignments and terminology work often only getting fleeting consideration, due to the time needed for them to make a positive impact.

    Why I am not LangTech averse…

    Good LangTech solutions might reduce a two-hour task to one of twenty minutes. A couple of hours invested can go a long way. It makes investing a couple of hours into alignment or terminology extraction far more conducive. I did a lot of alignment to try to establish a larger translation memory when starting out. Some of the benefits from those early alignments were only realised a number of years later.

    Another benefit of talking about LangTech isn’t just about what works, but also about what doesn’t. Understanding the cost-benefit analysis for a large language services unit and economies of scale also is useful to rule ineffective solutions. Granted, I only get to see how the public sector is considering LangTech,

    And how about LangTech in the private sector and for freelancers?

    The private sector is savvy to MT and having the human expert in the loop doing the PEMT task. I have started to see some freelancers offering “supervised MT” as a premium service. Here light post-editted machine translations are fit for purpose (e.g. gisting) although not a polished human translation. And this is a premium rate service. The potential new charging models are interesting in this regard. Otherwise, there is a race to the bottom with dumping rates abounding for PEMT work through agencies. Where I am particularly interested in their approach is about how they offer such a service for sensitive material. I am also interested in the mitigations in place and the workflows involved.

    Another interesting consideration is also what the PEMT margins are like such services, particularly if context matching is used in the MT process (or 102% (double context) and 101% (single context) matches in MemoQ). There is a very thorough examination of how MT character charging *really* works in this piece in the Multifarious blog. If sudden a text of 700-800 characters is using 2,500 characters from your character allowance, it might run through your character allowance quicker than you think.

    Preparing for presence meetings – (Re)learning to listen, observe, process and reflect

    I find that my preparations for presence meetings are far more thorough than hybrid/online meetings. By nature, I am never a passive participant in any meeting I attend, whether personally or virtually. However, with a presence meeting, I find I take more time over preparation, e.g. finding out who else will attend and what I can talk to whom about. I also try to prepare a set of potential questions in relation to the presentations at the meeting. Presentations of technical solutions also work better in person, and I usually have a think about the possibilities of how to apply what I see.

    In contrast, in a hybrid setting due to Webex-weariness, I probably only really tune into the meeting a few minutes before it begins, and seldom look at presentations beforehand if available. There is also the distinct temptation to relegate a hybrid meeting to a second screen and not to fully listen. Having had three years of exclusively hybrid meetings, it was definitely a case of having to almost re-learn meeting skills.

    After three years of speaking into webcams, with either a headset, conference spider or podcasting mic, and not having my own window displayed on my own screen, I found the return to full presence mode meant readjusting to simultaneously listening and observing gestures and facial expressions of speakers. The increased effort in terms of concentration, if unused to a presence meeting setting for a while, can be quite tiring.

    The value of presence meetings for freelancers

    Having freelanced for 14 years, I am aware of the somewhat lonely nature of day-to-day work. My advice would therefore be to try to find yourself a sparring partner. They could be someone who either works in your field or in another, or maybe in a different language combination. They should be in a similar situation to you and you should make time to meet in person, rather than online. Attending presence-based CPD events and conferences can be a great way to network to find your sparring partner. I’ve seen numerous freelancers have done this recently and get a lot out of such meet-ups.

    I realise that financial constraints for freelancers might make a two night trip with flights an unjustifiable expense. You might well be able to fine a more affordable way of meeting up with someone. How about roughly halfway from your respective offices, to have a concentrated presence meeting? Also remember the value of having time away from the screen. Such meetings remain essential as a time for reflection about what works or not in your job. They also provide a chance to consider changes to make to the way you tackle certain tasks. This can increase productivity or weed out distractions.

    If that doesn’t sound like your cup of tea, you could always consider having a presence meeting-cum-Christmas party – I used to have an “unofficial Christmas party” with freelancing friends – so that we all had a night out and were able to round off the year of freelancing in a convivial atmosphere!

    Part of this blog post also appears on my personal website at www.michaelbailey.at, That version has more of a focus on presence meetings, while this version focuses more on LangTech.