Category: Translation

Translation topics, mostly directed at language professionals.

Shameful and non-shameful use of AI

I don’t use AI when translating documents for clients, for the simply reason that I don’t need AI to translate, I don’t like the translations it produces, and I’m the translator I present to my client as the person who cares about and will execute the translations of their documents. Using AI would be a betrayal.

Translation-brokering agencies, which are well along in their replacement of human professionals with AI—and that leaves former translators with only extremely low-paid and mind-numbing post-editing work—are in a different situation.

The reason is simple. Almost all translation that is paid for by translation consumers is done by entities that are not themselves involved in executing translations, beyond purchasing the translations and then, if necessary, purchasing editing thereof before selling them. That is the case now, and it has been the case for many years, from long before humans found themselves being replaced by AI. Since entities that sell translation are only very rarely involved in doing translations, it makes sense for them to move away from expensive human professionals, and they are succeeding in that move.

Top Page of my company website

That said, for a process or task that I do not purport to do myself or sell to clients as a practitioner, I am more willing to use AI. If you look in the upper-right corner of my company website pages, you will see a hamburger menu icon. I have manually written the HTML for my considerable number of webpages for many years, resisting adding such a feature, not understanding how to do it, either with or without JavaScript. I gave in the other day and used Claude.ai to build that feature into my website. It took less than three minutes to obtain the required patch of html markup and the associated JavaScript, required no cash outlay, and it gave me what I needed with just a plain-language prompt of about three lines describing what I wanted.

There is a putative environmental impact—it is much smaller, of course, than the impact resulting from building a fake video of a deceased celebrity or a dancing cat—but I am not at all ashamed I did that. That would not be the case if I were in the business of selling webpage designs.

Returning to translation, the translations I do are mine and will continue that way, as I continue to resist the mindless rush into a world where translators surrender to AI-using brokers and professionals of all sorts outsource not only their writing but also their thinking to a collection of software commands. That’s not my style, and I cannot see that changing anytime soon.

Restoration of Content Probably Lost Forever Elsewhere

Having reached the reasonable conclusion that countless IJET Conference presentations (not just mine, but of many other volunteers to who spent the time and effort to present) before JAT abandoned publishing of IJET Proceedings and switched to videorecording of presentations have been lost or deleted from the JAT website, I have restored not only the full scan of the Proceedings of IJET-1 (Hakone 1990, where I did not present), but also my presentations at the following.

  • IJET-5 (Urayasu 1994)
  • IJET-7 (Yokohama 1996)
  • IJET-8 (Sheffield 1997)
  • IJET-9 (Yokohama 1998; not a lone presentation, but rather from comments I made in a panel discussion)
  • IJET-16 (Chicago 2005, kindly presented on my behalf by Manako Ihaya while was tied up interpreting here in Japan)
  • IJET-19 (Okinawa 2008)
  • IJET-27 (Sendai 2016)
  • IJET-30 (Cairns 2019)

I have placed links to all of the above at https://kirameki-translation.co.jp/forpros.html.

I will undertake to preserve these, as they are probably gone forever from the JAT website, not that they have much relevance at this stage (save for IJET-30, which turned out to be unexpectedly prescient, given the currently progressing demise of tier-two translators.

It’s time for agency-dependent freelance translators to face reality.

At the risk of accusations of having gone over to the dark side, I need to say that translators should accept the reality that the Japanese-to-English translation work that has supplied livelihoods to the great majority of freelance translators is now being handled quite satisfactorily (as determined by the paying clients, of course, not translators) by AI translation combined with editing done by former translators at very low rates.

The resulting cost is just a fraction of what it used to be, and this business model will succeed for at least a decade or two as is, with the help of currently idle translators, because a sufficient number thereof have no other career paths, but still have bills to pay.

Even after that, of course, there will be a continuing influx of Japanophilic language learners who are not aware of the real world of translation before they reach that real world and might not accept the reality of it even after they arrive there. That labor supply will continue to enable the above-noted business model. There won’t be a return to the old paradigm.

But even before the Japanophilic beginners take over post-editing from the dwindling supply of professional translators formerly working in mainstream demand fields, many translation consumers will themselves be doing AI translation and editing in-house.

That next stage of the transformation is already underway, as numerous translators continue to believe that things will work out, claiming that humans will always be needed to provide “cultural nuances.” The reality, however, is that only a tiny portion of translation that is paid for has anything to do with culture, and the parts that do can be fixed by very low-paid post-editors, at this stage meaning former translators who have no choice but to bite the low-paid post-editing bullet.

Denials, delusions, and occasional counter-indicative success stories aside, an evidence-based evaluation of where things actually already are points to exactly the above picture. The evidence?

  • Specific known translators forced to leave;
  • laughable rates paid to be the “human in the loop;”
  • the major translation sellers that control most clients and countless new translation-selling startups offering “human-in-the-loop” AI translation as their main service;
  • a paucity of translation work offered on even sleazy click-work reverse-auction platforms, which are now filled with very low-paid post-editing work;
  • no evidence of either a return to human professionals or dire AI translation-caused consequences, other than unverifiable anecdotes and aspirational, wishful-thinking predictions; and
  • a significant reduction and demographic transformation of the membership of the formerly most active group of JA/EN translation practitioners in Japan.

And there’s much more evidence from the real world, divorced from the aspirational predictions of survival made by people who hang out at the Denial & Delusion Café.

The above is not a pretty picture, but it is real, and you don’t have to wait for it to come to your neighborhood; it’s at your doorstep, if it’s not already in your kitchen eating your lunch.