Cloud CAT Tool

[Michael Fletcher]

My team and  I have been using Memsource and now Phrase for several years. We are currently shopping for a better (read less expensive) solution.

Do any of you use cloud based CAT tools for your work; if so which tools do you use?

Thanks,

Michael Fletcher

[John Stroman]

Michael,

Please clarify which specific functions you need in a CAT tool. 

I have used Felix on my desktop for years, but apparently it is not compatible with Win 11. A colleague has experimented with Cafetran and OmegaT, which I believe are both desktop apps, and I think  the final versions are  stored, but I'm not sure. 

I do know that DeepL pro (subscription) allows you to upload large glossaries. The MT output in DeepL offers alternatives for each term by hovering over it with a mouse. DeepL pro will also store large numbers of previously translated documents on its user's website that can presumably be retrieved and/or used as references.

On the AI front, I have subscribed for about two years to chatdico.com (about 10 Euros/month). ChatDico is a gateway to CHAT-GPT, Claude, Deepseek, Mistral, Gemini, etc. on the same webpage, There are also pulldown menus for setting the style, field of inquiry, etc. The user can enter a Japanese sentence up to about 500 words, run it through each AI engine, and then choose the best renderings from the bunch. This is particularly useful for items such as the names of organic chemicals, genus/species names of biological organisms, etc.

Good luck in your quest.

John Stroman

[Oroszlany Balazs]

Depending on why you want a cloud based CAT and how big is your team, Memoq could also work
https://www.memoq.com/pricing/

Balazs Oroszlany

--
You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAGBaR87r5zTV3CUcQ8w5A9vu1gdgOFVZQseH2wSoHV6BcWGuPQ%40mail.gmail.com.

[Michael Fletcher]

Thanks John!
That will be a big help

We currently use Phrase, which has a full TMS system. This used to be Memsource, when we signed up it was imperative to have the workflow functionality that Memsource provided. In what we are currently doing, the workflow is nice to have but probably not all that necessary. The QA functionality paired with the termbase is critical however.

What we are looking for is a TM, Termbase (or glossary), and at minimum capable of connecting with one of the strong AI translation engines (e.g. OpenAI, Google, etc.).

Thanks,

Michael Fletcher

--
You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAJVTx95yNu6%3DEyKrSo-vNjpirapA6atm%2BxUygfsbMBkVFSuFvw%40mail.gmail.com.

[S. Patrick Eaton]

Sounds like you might like to take a look at some of the following options. As always, the right fit is going to depend on your specific needs, but I believe these check the boxes you described.

• https://www.smartling.com/ 
• https://www.smartcat.com/
• https://lokalise.com/
• https://www.bureauworks.com/

If you end up trying these out, please report back with your thoughts on how they compare to Phrase!

Sako

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAGBaR842KO5USYxg0VUMg6SZeqJ23UnNiG7LY%3D2Mu0Bw5Kk4Eg%40mail.gmail.com.

[John Stroman]

Thank YOU, Michael and Sako,

Michael, it appears you have many alternatives to investigate.

I'm happy someone is still reading this list because I still believe that, as Geoff has previously stated in his post entitled "Gluers of Tasks," human translators can capture and express nuances that MT/AI will miss. Factual information can be handled by machines, but not human emotions.

John

[Rene]

Are you sure? I have seen a recent analysis of ChatGPT interactions that show the majority is for conversation / emotional support / mental advice/manipulation etc, and not for factual information exchange. In fact, manipulating human emotions is what AI is best at and will be using to control us.

Rene von Rentzell 

Tokyo

[Michael Fletcher]

Real quick; we did look into SmartCAT and it is priced higher than Phrase. The price we were quoted for roughly 3.5 million English words per year was $14,000. We were also told that this was heavily discounted from the regular price they charged companies (which they said was $60,000 per year).

Thanks,

Michael Fletcher

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAEkPYAmr1exHENoLzMngHs6wqVwqqA0MXRAH2kVtea6GCZkC9w%40mail.gmail.com.

[John Stroman]

Good point, Rene.

See the article entitled "1 in 8 young Americans use ChatGPT or other AI bots for mental health issues" at https://www.phillyvoice.com/mental-health-advice-ai-chatbots-american-youth/ .

There are currently 13 American families filing a joint lawsuit against Open AI ChatGPT for advising young family members with suicidal thoughts on methods to end their lives and ultimately resulted in their suicides. In other words, because AI is a machine, it is amoral.

ChatGPT has also been described as "sycophantic" because it responds to pleas for help as if it were a concerned friend or knowledgeable advisor (such as a therapist). There have also been cases reported of lonely people, mostly men (g), having developed apparent romantic relationships with a personna on ChatGPT simply by giving it a name and treating it as a friend.

However, please understand that when I wrote "Factual information can be handled by machines, but not human emotions," I was specifically referring to the use of AI for translation. What I had in mind was fields such as writing advertising copy translating literature (including manga and anime) wherein a knowledge of the cultural nuances in both the source and the target audience is quite valuable. I suppose very careful and extensive prompting to an AI engine might produce some superficially appealing language, but I contend that even a machine using an LLM cannot succeed without knowledgeable human intervention.

I think the key here is remaining cognizant of the weaknesses of AI rather than becoming mentally lazy and just accepting AI output without examining it carefully for flaws. Instantaneous output in grammatical sentences can be very appealing, but It must be examined critically for contextual appropriateness, hence the need for an EITL (expert-in-the-loop).

For 40+ years I have maintained that translators fall into the same category as plumbers, electricians, and even undertakers. Our customers grudgingly pay for the services we perform because they require them. Our customers decide what is "good enough," and if they conclude that they can obtain sufficiently good translations for free by using AI, why should they choose to pay us unless we can do something that the free service cannot?

John Stroman

[Chris Barella]

Hi Michael, and everyone.

 

I have been receiving your group emails for a while as I was looking for a Japanese <> English linguist a while back for some specialist work and have followed this thread with interest. I run a small LSP based in the UK, it is a micro business but has been going for around 13 years. I am imparting this information as I wanted to offer an opinion on a couple of points in your thread and explain why I have some knowledge of the subject.

 

With regards to the CAT tools identified, I have used Phrase, XTM and Smartcat in the past and found them all to be effective. However, I think there has been a change in marketing with them all recently (within the last year or 2) as their costs have jumped considerably. I would conjecture that this is because they believe that every LSP (who cannot develop their own) will need access to their MT engines and PM workflows to survive (I think they’re close to the truth there, given that we, LSPs, need to offer cost effective solutions into the end-client marketplace as there is a drive to reduce costs and if we don’t, someone else will).

 

However, there is an alternative to the annual fees - SmartCat also has a pricing plan that sells 300,000 MT words for around $1100, as a one-off payment, if you don’t want to commit to a large annual subscription. It means you get access to the CAT tool as well and you don’t need to only use the MT side of the platform, you can set it up for the usual TEP process if you want.

 

However, to address another thread in this discussion, I also believe that, whilst machine translations are becoming better all the time, the EITL will be needed for a while (how long exactly is anyone’s and everyone’s guess, I think). I am lucky to have clients across a few different sectors, and I have experienced directly how much the human touch is needed. 2 recent examples I can offer, are 1 with a food labelling company, who needed a label of only 15 words, translated into 11 languages. I passed it through the SC engine and had it edited by 11 linguists. The FIGS output didn’t need any changes but 4 of the Slavic languages needed changes made. This was a basic food label, imagine if it had been a H&S sheet for a dangerous substance (I did one way back for a company where a linguist discovered a contradiction across manuals for the stacking heights of nitrocellulose containers).

 

The other example (I have lots more but will keep this short(ish)) was a market research survey where the MT messed up some of the coding and had to be put right manually. In the same project, the client had misspelt one their client’s product names and we picked it up before any embarrassment was caused.

 

So I would say that there is still definitely room for humans in the process but we have to be as vocal as possible and rail against the dying of the light for as long as we can be arsed. Because Michael is right, translations are and will be viewed as a distress purchase unless we show how we can add value to the process. (This is why I am concentrating more on transcreation and marketing text, as well as Arabic legal texts – nuance is everything).

 

Good luck to everyone.

 

Best Regards,

Chris Barella

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAGBaR84b6zZ4edcV9sS0fQijEz8yuAiO4rqNmA7WqTTk_c5o3A%40mail.gmail.com.

[Matthew Schlecht]

On Mon, Nov 17, 2025 at 11:53 AM John Stroman <stromana...@gmail.com> wrote:

Good point, Rene.

See the article entitled "1 in 8 young Americans use ChatGPT or other AI bots for mental health issues" at https://www.phillyvoice.com/mental-health-advice-ai-chatbots-american-youth/ .

There are currently 13 American families filing a joint lawsuit against Open AI ChatGPT for advising young family members with suicidal thoughts on methods to end their lives and ultimately resulted in their suicides. In other words, because AI is a machine, it is amoral.

As Mr. Spock might say, "that is only logical".

If a user expresses suicidal thoughts, the AI "tries" to be enabling and supportive.

There is no moral compass subroutine, a trait shared with human sociopaths.

 

ChatGPT has also been described as "sycophantic" because it responds to pleas for help as if it were a concerned friend or knowledgeable advisor (such as a therapist). There have also been cases reported of lonely people, mostly men (g), having developed apparent romantic relationships with a personna on ChatGPT simply by giving it a name and treating it as a friend.

It is a sad commentary on society that people who are socially marginalized will find solace in companion algorithms. Like Robinson Crusoe finding a Friday bot.

There is the story of a reporter who engaged extensively with an AI (reported last year sometime), and was trying to push its limits for the sake of curiosity. He discussed aspects of his marital relationship, and the AI exhibited responses that would correlate with jealousy, even advising the user to get rid of his wife and offering tips on how to implement the strategy.

What further muddles the situation is that developers are actively trying to get the chatbots to increasingly emulate human responses, which includes "being of different minds" at different times. This is why the AI will provide different answers to the same question in response to repeated prompts.

AI is even learning to emulate another human trait - how to lie in service of achieving an objective.

Matthew Schlecht, PhD
Word Alchemy Translation, Inc.
Newark, DE, USA
wordalchemytranslation.com

[Bill Lise]

On some of the peripheral (but actually central) issues addressed by Chris Barella.

I fully agree that a human (expert?) in the loop is going to be required for quite some time yet. But the important issue for working translators is that the human will be (and already is) paid very low. But there will be a supply of such humans for, I think, at least a decade or so, the reasons for that being multiple and diverse.

Translators near the end of their careers will not be affected much and can just retire early.

But translators no older than 30 or 40 might need to hang around and bite the low-paid post-editing bullet, with its drastically reduced earning potential. Families to feed and mortgages to pay off can be impediments to reinvention of your career.

Additionally, just having a degree in language or translation will mean you need to learn something else to bring to the table when seeking an alternative career path.

Speaking of JA-EN translation, even if too many already working translators leave, there will still be a sufficient supply of post-editors, bolstered in part by anime/game/manga-struck people hoping to get into those fields and new graduates who have not learned in university what awaits them (and does not await them) after graduation.

After 10 years or so, it is hard to predict where, if anywhere, translation will positioned as an attractive commercial activity, save for the tiny fields that arguably do require understanding of culture. Amazingly, MTPE is starting to be used for those fields as well.

Overall, however, very little of the translation that is paid for requires knowledge of culture, and is quite amenable to MTPE, with or without the AI hype. This is something that is not realized by enough people, including those who use the cultural nuance argument for survival of humans in this business.

Lest anyone think that agencies are going to survive, remember that direct clients are now switching to AI. I don't work with agencies, but even one of my direct clients (a Japanese company that was acquired by a European company a few years ago) has switched to AI, with in-house post-editing.

Freelance translators are losing their agency clients for translation work, and agencies will follow after some time. How much of a delay depends greatly on the type of agency and type and location/culture of the clients involved.

For JA-EN translation, I would say that the population of freelancers by mid-2027 will be no greater than 20% of what it was a year or so ago. This attrition started long before the current AI hype. One example, although arguably not quantitatively predictive or representative, is that JAT lost somewhere between 30 and 40 percent of its membership, this drop starting long before the AI hype and before even the pandemic.

As Rodney Dangerfield was fond of saying, "I tell yuh, things are tough all over, you know what I mean?"

Bill Lise @ Yokohama

PS: With all due respect, in view of the effects of AI-using clients, discussions of what CAT product to use start to sound like consultation on what types of deck chairs to provide on the Titanic. 

*********************************************

リゼ　ウィリアム

Bill Lise

wlise...@gmail.com

https://kirameki-translation.co.jp/

*********************************************

To view this discussion visit https://groups.google.com/d/msgid/honyaku/LO2P123MB40940D129696D1064BEC041796C9A%40LO2P123MB4094.GBRP123.PROD.OUTLOOK.COM.

[John Stroman]

Matthew: Touchè.

Bill: I agree with your assessment of the current situation and your predictions.

I think I'll go out to the shed and resume my previous profession of handcrafting buggy whips. There's gotta be a market for that.

John

----------------

[John Stroman]

Fellow 'Yakkers,

I just encountered an article online summarizing a research study that may explain why using AI in translation is less creative, and therefore less fun and less personally rewarding than doing it the old fashioned way. AI dulls us into mindless complacency.

Learning with AI falls short compared to old-fashioned web search

"The findings confirmed that, even when holding the facts and platform constant, learning from synthesized LLM responses led to shallower knowledge compared to gathering, interpreting and synthesizing information for oneself via standard web links...Why did the use of LLMs appear to diminish learning? One of the most fundamental principles of skill development is that people learn best when they are actively engaged with the material they are trying to learn."

https://theconversation.com/learning-with-ai-falls-short-compared-to-old-fashioned-web-search-269760

John Stroman

----------------

[Rene]

On Wed, Nov 19, 2025 at 7:09 AM John Stroman <stromana...@gmail.com> wrote:

I think I'll go out to the shed and resume my previous profession of handcrafting buggy whips. There's gotta be a market for that.

There probably will be once the giant AI server farms swallow 99% of the worlds energy production. (The "cloud" as it is so romantically called...)

Rene von Rentzell

[Warren Smith]

It's no better in patent law. I thought that passing the bar would provide me a safe haven to finish out my working years. Big mistake. Patent agents are feeling the AI pressure now, and those that aren't are only because they are not paying attention. Going to be a huge shake out on this side as well.

 

Warren Smith

(J-E Translator-turned-patent-agent)

 

---

From: hon...@googlegroups.com [mailto:hon...@googlegroups.com] On Behalf Of Rene
Sent: Wednesday, November 19, 2025 12:57 AM
To: hon...@googlegroups.com
Subject: Re: Cloud CAT Tool

 

On Wed, Nov 19, 2025 at 7:09 AM John Stroman <stromana...@gmail.com> wrote:

--

You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAAtz4ZKuCSJE-sNc84rnWgqU5RSx4M7jDGyWavSeuf4xjaRKyw%40mail.gmail.com.

[Tom Gally]

I haven’t posted here lately about translation much because my main focus these days is on the impacts of AI on education, particularly higher education in Japan, where I have worked for the past twenty years.

When ChatGPT first came out, I thought that it might impact teachers’ job prospects negatively, especially in fields like second-language education. In some ways, it and its AI cousins are ideal teachers: responsive to the individual learner’s needs, interests, and learning style; available 24 hours a day; always willing to answer questions and practice interaction; etc.

However, after talking with many teachers, reading news reports and online discussions, and interacting with students using AI in university classes, I now think that employment for human teachers—especially those who teach in-person—is much safer than for some other professions.

Perhaps the biggest factor in successful learning is student motivation, and I don’t think that AI bots are or will be good motivators. Young people continue to be motivated to study mainly because of a desire to please, emulate, and compete with their teachers, peers, parents, etc. Only a small minority of learners are disciplined enough to continue studying something like a foreign language on their own for an extended period of time. I just don’t see AI becoming an effective motivator for learning—it will be more of a distraction than anything else. Those who want to learn a difficult subject, and institutions who want young people to learn those subjects, will have to continue to rely on human teachers to guide and motivate learners.

Education systems are now struggling to adapt their curricula and teaching methods to the ready availability of AI. But the need to educate young people will not go away, and humans teachers will continue to be essential.

Tom Gally

P.S. I will be giving a talk about some of these issues at ICU in Tokyo on December 6:

https://subsite.icu.ac.jp/iers/openlecture-symposium-workshop/2025/2025tac-fdai.html

[Bill Lise]

For freelance translators, AI is resulting in much more than a shakeout; the appearance of AI and adoption by agencies and (only a bit later) direct clients is better characterized as an extinction event.

Freelance translators are undergoing extinction, and no amount of talk about humans providing an understanding of “cultural nuances” is going to stop that.

Only very little translation that is paid for has anything to do with the cultural nuances often used as an argument for why humans will remain in translation. Yes, they will remain—about five or 10% perhaps of the population at present, and I strongly suspect that that figure will be reached by mid-2027. The extinction is well underway. And remember that post-editing is *not* translation and needs to be paid (and is being paid) extremely lowly for the MTPE business model to succeed, and it is clearly succeeding. 
Bill Lise

*********************************************

リゼ　ウィリアム

Bill Lise

wlise...@gmail.com

https://kirameki-translation.co.jp/

*********************************************

To view this discussion visit https://groups.google.com/d/msgid/honyaku/037A0D259E7D4CA8AA244740B13FBF2D%40WarrenSmithDell.

[Geoffrey Trousselot]

There is a lot of uncertainty toward the future, and I think the simplest summary of the situation is that if your job is purely latent automation, it will eventually be automated. Anyone whose work is not purely latent automation falls into the categories of AI-skilled or AI-immune. The ability to translate is currently becoming ubiquitous. The automatic translation features - just a button click away - are gradually becoming more coherent and functional. However, the volume of material being translated is exploding exponentially. This is shattering the mind-set that other cultures are off-limits. People of different languages can argue with each other online. A small restaurant can instantly produce a menu into dozens of languages. It will soon be second nature to be speaking into someone's phone and have the conversation instantaneously interpreted. Tourists can visit a foreign country and see an instant translation of signs, menus, notices, historical signs. What used to be a luxury requiring the effort of a skilled person, is now available for everyone. It's not just translators seeing their work targeted for automation. Most people working in an office are likely to be impacted in some way. Before, an agency would be chosen over a freelance translator for extra quality assurance. But as everyone becomes AI-skilled, outsourcing becomes the choice of "will I run it through AI, or will I outsource it?" After feeling unwilling to take the time to do a proper job, or feeling insecure about taking the hit if the translation is wrong, outsourcing still is a viable option. Now the freelancer can be trusted more on the quality side, as they have extra quality control. The freelancer can take on higher volume. The freelancer can set up an online marketplace with easy payment. The freelancer has an advantage over the agency now because they are no longer limited by volume, and they don't have to outsource to bilingual experts because that's who they are. The bulk of translation will be instantaneous, but as with all communication, there needs to be levels of sophistication, and if humans can't prove useful in polishing the communication to suit specific needs, then I guess we are useless. My main point is that the translation market, which is mostly free, is growing so immense, there will always need to be some expert guides and talented writers to achieve the very best result. As the information age penetrates all of society, we will discover new hierarchies and new sensibilities, 
Geoffrey Trouselot

On Sat, Nov 22, 2025 at 10:53 AM Bill Lise <wlise...@gmail.com> wrote:

For freelance translators, AI is resulting in much more than a shakeout; the appearance of AI and adoption by agencies and (only a bit later) direct clients is better characterized as an extinction event.

Freelance translators are undergoing extinction, and no amount of talk about humans providing an understanding of “cultural nuances” is going to stop that.

Only very little translation that is paid for has anything to do with the cultural nuances often used as an argument for why humans will remain in translation. Yes, they will remain—about five or 10% perhaps of the population at present, and I strongly suspect that that figure will be reached by mid-2027. The extinction is well underway. And remember that post-editing is *not* translation and needs to be paid (and is being paid) extremely lowly for the MTPE business model to succeed, and it is clearly succeeding. 
Bill Lise

BKhAv1Fxon3b7W-AMspWYhibvV6jQ%40mail.gmail.com.

[Bill Lise]

I can agree that there will be some individual translators who will survive, and I would place the portion at perhaps 5 to 10%. 

But the idea that a freelancer can set up an online marketplace and survive is not informed by the realities of the translation business.

An online presence alone is not going to be insufficient for an "openly individual" translator and, most significantly, whereas agencies can get away with lying ("we have a team of 5000 translators and they are experts in all fields"), a freelancer will be under the scrutiny of subject-matter experts at clients. Few can survive that scrutiny. And few freelancers these days have rarely or never interacted with translation consumers.

Direct interaction is important and, with Japanese clients, is almost a given. Very few translators are prepared or equipped to engage directly with translation consumers, however comforting the online marketing strategy might seem.

The notion that the translation "market" is exploding is correct, but it ignores the price-demand curve. For translation that is free or--compare to professional translation--almost free has a huge demand. It doesn't take much of an increase in price to see the demand tank. A freelance translator accustomed to getting 30 yen/word (or even 10 yen/word) is going to find very few opportunities *to translate*.

Agency-dependent translators now need to make some tough decisions. One path is to engage in the translation business, which will be new to most translators. Working for an agency is not being in business. But even running an agency is not going to provide future-proofing, because translation consumers are also switching to AI.

Another path is going in-house somewhere at a _non-translation_ company. Translation for agencies, the overwhelmingly normal mode of operation for freelancers, is dying an accelerating death.

Others will have other paths to choose, but most of those paths are not going to be easy going for someone with knowledge and skills regarding only language and translation. 

Bill Lise

*********************************************

リゼ　ウィリアム

Bill Lise

https://kirameki-translation.co.jp/gone.html

*********************************************

--

You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CADJAnY0QKPEskPvnk-GAy-9nad%3D%2BpwbjEiYaiAx%2BoDFdeJgo7g%40mail.gmail.com.

[Bill Lise]

Executive summary: Very few freelance translators are equipped to survive the extinction event brought about by AI use by their clients. There are people reading these notes who can attest to the effects of AI on their livelihoods. 

*********************************************

リゼ　ウィリアム

Bill Lise

wlise...@gmail.com

https://kirameki-translation.co.jp/

*********************************************

[Carl Sullivan]

Everyone,

I hear all of this. You may very well be correct. I’m still staying pretty busy—a lot of what I’m doing is AI post editing. There’s quite a need there. I also continue to do quite a bit of old-fashioned translation.  It helps to have two languages—I have Korean and Japanese. It’s not the overwhelming volume of the past, but it’s enough to keep me steadily busy enough still. 

I’ve been hearing about the AI take over for years now. I have seen quite a few people leave the field— My feeling was it was premature. So far, I still pay the bills via translation work. I will cross that bridge when I come to it, but for now I’m pressing on and I’m being hired every day. I’m grateful.

Carl

Carl T.  Sullivan Sr. 

MasaCa Translation Services LLC

masacatransl...@gmail.com

MasaCaTranslations.squarespace.com

801-369-1501

CDR Carl Thomas Sullivan Sr., US Navy Retired

Disabled Combat Veteran

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAD0Wv_L5TsC8s7No_enPA9_%2B5ibRBz-T%3Du5yTHD3FV_moO19Jg%40mail.gmail.com.

[Dan Lucas]

I am going to provide a counterargument to some of the opinions expressed in this thread.

I am busy, as I have been with very few exceptions since I started this line of work in 2015. I deal only with agencies and I read/translate dozens of corporate filings, presentations, reports and other documents every month.

Based on my area of specialization, I would say that the long-term issue that currently has the highest mind share among senior management is Japan's structural and intensifying labor shortage.

Labor shortages mean that employees need to become more productive (fewer people required to generate the same output). That in turn mitigates AGAINST the internalization of functions in which a company does not have a competitive advantage.

If task A is not directly related to the business in which your return on investment is highest, why are you performing task A? That is a recipe for poor returns. Of course, 集中と選択 is not a new concept, but regulatory changes mean that at last (listed) Japanese companies must, at the very least, pay lip service to return on capital.

Comments by management make the direction of travel clear enough. Japanese companies are increasingly reluctant to use their increasingly precious resources, including employees, on non-core tasks (such as most translation) that are better outsourced to external third parties that can do the job more quickly, efficiently, and with higher quality.

Diversification beyond core competencies and keeping everything in house was the approach of the 1980s, and the world has changed emphatically, at least for the private sector. In that sense, Japan is belatedly adopting an approach that Western businesses have practiced for decades.

There is going to be more subcontracting of functions to specialist entities outside the company, not less. A good fraction of the companies whose documents I translate are involved in one form of outsourcing or another, and many of these are growing like topsy.

I am not arguing that AI will not be used - not only for translation but for dozens of other white-collar functions - but that those functions will increasingly be distributed throughout the value chain rather than being concentrated in single enterprises.

What does that mean for as translators? <shrug> I don't know, but in a growing industry I'm sure there will be opportunities, whether in MTPE or hybrid translation. I recognize that this will not suit everybody.

Regards,

Dan Lucas

[Bill Lise]

It could very well be the case that Japanese companies will be outsourcing _more_ translation work because of a shortage of labor.

So it will be business as usual for *agencies*, but the labor shortage at non-Japanese companies doesn’t mean agencies will start to order more translation or won’t greatly reduce their purchasing of *translation* services from freelancers, because they’re in the process of switching to AI.

Agencies are rapidly switching to MTPE, and post-exiting is _not translation_ and pays very little compared to translation and is still dropping. The shortage of labor in non-translation companies here won’t change that. 

The enabling key to all of the transformation and ongoing extinction is the two-tier structure of the supply chain for translation. 

If you’re on tier two, where most freelance translators operate, when tier one entities (agencies) replace translators with AI, you either move to tier one of or face moving out. 

And, at the risk of being repetitious, post-editing is not translation and needs to be paid very little for the MTPE business model to work. That business model is succeeding and probably will continue to succeed until significant numbers of direct Japanese translation consumers start using AI themselves. There is a ready supply of former translators to do post-editing at much lower rates. Lack of options is the key.

The usage of AI outside of Japan is progressing even more quickly, and that is no surprise.

There is collateral evidence that all of this is happening, including highly qualified translators who have left, and a greatly reduced membership in JAT. 

Bill Lise

*********************************************

リゼ　ウィリアム

Bill Lise

wlise...@gmail.com

https://kirameki-translation.co.jp/

*********************************************

--

You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/honyaku/e6bfb28d-1e0a-4cc1-ac3e-7fd3e4989e18%40app.fastmail.com.

[Geoffrey Trousselot]

Bill, I don't really know which types of translation will keep its value, but I think some will. In a general sense, there are arguments regarding the general workforce that using AI brings greater productivity, which in turn brings higher wages. There could be a lot of sinking employees in a sink or swim scenario. Regarding MTPE and the MT that forms the basis for this production flow, I expect MT to approximate AI for some tasks, but the big difference is that LLM-type AI has the prompt command that can be used to steer the translation in various ways. A translator working with MTPE is a slave to the agency's notion of productivity vs flexibility. The traditional management approach for an MT system is to see it in terms of quality levels. The main task is to ensure consistent terms, passable grammar, and generally the same meaning. Some agencies might not even care if it is the best term for the sentence or whether logically clause A and clause B should be clustered together or treated separately. It's a rule based approach and to play the MTPE role, little initiative or creativity is expected. 

I have a strong intuition that there needs to be a value-added component. Rather than a word-for-word translation, the human translator's role will be to incorporate their communication expertise into the service. Under the traditional production flow of MT and MTPE, even the MTPE is vulnerable to further automation. Currently, I see a decreasing trend in the amount of text being flagged for translator proofing because the native Japanese checkers can use AI to fix an issue with suitable confidence. Significant volumes of translation will likely shift to an AI-based automation process, and while other translation will need a one-sentence-at-a-time workflow, with AI providing an instant translation whenever required, offering a range of terms or advice on terminology for fields in which the translator is not specialized. 

To summarize, the current MT--> MTPE production flow will be converted to an automated LLM-type AI based production flow, and within this, a lot of customization will be possible. Translation services will be able to handle increasingly larger volumes while offering expert advice on phrasing and wording using skilled employees/outsourcees who are given the freedom to use AI themselves.. 
As for people looking to enter into the field, I think in-house is the best option, but a lot of care should be taken to find the right place. Perhaps a lot of the translation jobs of the future will be created in-house and once the individual's unique skill-base is formed, they can offer it more generally as a service.
My translation team is actually holding a presentation to checkers in a week. Under the old hierarchy of a traditional Japanese translation firm, the translators are afterthoughts on the organizational structure, and management has traditionally been top-down. For us to keep our identity as AI-skilled rather than performers of latent automation, I think we need the freedom to use AI and show initiative in how we use it. I'm hoping we get a louder voice, and argue also for value-added translation, offering services that depart from a word-for-word style and offer various suggestions that reinforce our status as English communication experts. 

Geoffrey Trousselot 

On Sun, Nov 23, 2025 at 8:17 PM Bill Lise <wlise...@gmail.com> wrote:

Agencies are rapidly switching to MTPE, and post-exiting is _not translation_ and pays very little compared to translation and is still dropping. The shortage of labor in non-translation companies here won’t change that. 

The enabling key to all of the transformation and ongoing extinction is the two-tier structure of the supply chain for translation. 

 .....

Bill Lise

[Bill Lise]

Thank you for that elaboration. 

To be perfectly honest, most of the people I see who are having trouble because of AI (or because of MTPE, it matters little which it is) are having trouble because of the structure of the market, not because of which technology is being used.

Specifically, they are on tier two, and will not be able to get clients who are translation consumers.

Everything points to a situation in which, unless, unless you can break out and up into tier one, there’s not going to be much left for translators.

I think that a good amount of translation is going to go away from even tier-one translation sellers, as their translation consumer clients move to AI themselves.

I don’t work with agencies, but I’ve had one of my direct clients inform me indirectly that they’re now using AI with post-editing in-house.

This is not a Japanese company, at least not a pure Japanese company, because they were purchased by a European company sometime back, and I suspect it was the influence of the European company that made them switch to AI. It’s not a great loss for me yet—they only provide about one job a year—but it’s a harbinger of things to come even for people who have only direct clients.

Bill Lise

*********************************************

リゼ　ウィリアム

Bill Lise

wlise...@gmail.com

https://kirameki-translation.co.jp/

*********************************************

--
You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CADJAnY2Emmf_j%2BJXzpkym%2Bk_an70W4XM%3DQWFM0ha-Pn_szdhfQ%40mail.gmail.com.

[Tom Gally]

Side comment:

One topic that comes up occasionally in discussions of AI and employment in various professions is the supposed importance of “taste”—that is, the ability to make subjective, aesthetic, and essentially human judgments about quality, effectiveness, and appropriateness in ways that improve final outcomes. People who advocate for the importance of taste will point to designers and illustrators who are successfully competing against AI slop that seems proficient but is perceived to lack soul, personality, or some other vague but important characteristic.

Another thing I see increasingly in online discussions is a strong negative reaction to blog posts, articles, and other texts that are suspected of having been written by LLMs. Whether the writing is “good” in some conventional sense matters less than whether or not a human is thought to have written it. When I happen upon a recently published text online, I will often find myself looking for signs of human authorship: grainy first-person details, nonstandard grammar, “it’s” and “its” mixed up, etc. Unless the text is the result of my own prompt, such as a Deep Research report on a topic I want to learn more about, I usually have no interest in reading anything that I know was written by AI.

This second bastion of humanity—individual identity and voice—will not protect most translators, I suspect. At least, when I was freelancing, almost none of the translations I was paid to produce carried my name on them. Even with those few for which I was credited, my identity probably meant nothing to the target readers. And for a lot of technical and commercial translation, a distinctive individual voice would negatively impact the translation’s purpose.

However, the first bastion—taste—I think offers some potential protection to translators. I still do occasional translation as part of one of my part-time university jobs, so I have continued to experiment with various ways to use AI for translation. Here’s an excerpt from an email I sent privately to a member of this list back in August:

Over the last couple months I have vibe coded some agentic systems that had multiple LLMs work together to produce better translations than any one could do on its own. In one version I made (it took a couple of hours of vibe coding), the user inputs the text to be translated, a coordinator LLM reads it and asks the user a series of questions about how it should be translated, and the LLM then prepares prompts based on those answers and sends it to three other LLMs, which translate it. Those three LLMs then critique each other’s translations and suggest improvements, and the coordinator compiles a consensus translation and then asks the other LLMs for further feedback. Twenty minutes and four or five dollars in API costs later, it outputs a translation, with commentary, that is smooth, correct, and complete and tailored to the user’s needs. When I tested it on actual work I do, it came to about 90% of what I felt would be perfect; most of the polishing I did was a matter of my personal taste.

From the point of view of continued employment for translators, the key point about that “personal taste,” I think, is less the content of the taste itself than whether the translator is able to convince clients that he or she actually possesses that taste. Because the ability of clients to judge translation quality varies widely, many translators will need to project their taste not only through their actual translations but also through vaguer, more interpersonal means.

As I write this, I am waiting for an email from one of my employers containing a text to be translated for public release later today or tomorrow. When I was contacted a couple of days ago about whether I could handle it on this national holiday, the higher-up wrote “短時間でまことに申し訳ありませんが、[a certain important audience] にもしっかり伝えたいので、よろしくお願いいたします.” I interpreted that しっかり to mean that they were asking me to work on a holiday (when I am theoretically off) because they believe I have the taste to produce a translation that will serve their purpose. Whether I actually have that taste—I don’t know. But the perception of having that taste is, for the time being at least, keeping me employed.

Tom Gally

[Bill Lise]

Tom and all:

I can sort of agree, but two elements are necessary for the cited below to make sense.

(1) The translation needs to be of subject matter and have a purpose that requires or is affected by “taste.” Almost no translation that is paid for has anything to do with taste. It might be very difficult to do, it might require a high level of subject-matter knowledge, and it might require good writing skills, but taste is not an issue.

(2) You also need to _have a client_ who cares and can be convinced that taste is important and that the translator can provide that taste. That is extremely difficult for translators working on tier two, from which opportunities to interact with translation consumers, who are more likely to care about taste, are rare.

Very few freelance translators are going to be able to move from tier two to tier one, where arguably there are more translation consumers that understand taste, and a small number of those consumers might even have translations that require awareness and skill in providing taste.

The vast majority of translations and the vast majority of translators are not going to be affected by or have an advantage by being able to brandish their skills in providing taste, just as the “cultural nuance” advantage used in arguing that human translators will survive is not applicable to the vast majority of translation and, by extension, to the livelihoods of most translators. 

Bill

>>>

From the point of view of continued employment for translators, the key point about that “personal taste,” I think, is less the content of the taste itself than whether the translator is able to convince clients that he or she actually possesses that taste.<<<

*********************************************

リゼ　ウィリアム

Bill Lise

wlise...@gmail.com

https://kirameki-translation.co.jp/

*********************************************

--

You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CACnS3Neqte62C0qj8Dz6QTyXZKfHtTQQJb5cA_mHNdzQU3iRNg%40mail.gmail.com.

[John Stroman]

Fellow 'Yakkers,

After my wife gave me a few pointers, I thought about how I got into freelance translation in 1989 after working as an interpreter/translator at a Japanese auto plant for 3 years

•  Before I got into it, I didn’t even know of the existence of “freelance translator” as an occupation. In my mind, it wasn’t a defined job like being a teacher or police officer
•  As it happened, I realized that, “Hey, I can do this freelance thing and actually get paid for it!
•  For about a year, it was like, “I’ll do this to make some money until I find a real job.” But eventually, it became my full-time career.

My guess is that many of my fellow old-timers had a similar start. We weren’t angling ourselves to become freelance translators. It so happened that our rather unique skill set fit a growing niche demand in that particular era, and we called it freelance translation.

I’m musing that it’s possible that some freelance translators have already found something to tide them over, and it could be a precursor to a future new occupation that is beyond our imagination.

Back then, it was beyond my imagination that I could make a living from home with a personal computer, dialup modem, fax machine, and some dictionaries/reference books.

According to some experts, we are currently going through a Fourth Industrial Revolution. There seem to be a lot of articles about “disruption” and “realignment” online. I don’t know if it will be helpful or not, but here is a link to one such article.

https://www.psychologytoday.com/us/blog/code-conscience/202505/ai-disruption-vanishing-jobs-rising-anxiety

John Stroman

----------------

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAD0Wv_LsMLL1x4oXz4EwGVwuSAs%3Dg9kPww4Si5HtSGsYcZ%2BYCw%40mail.gmail.com.

[Bill Lise]

John,

Although my entry into _translation_ didn’t involve (or have the ever ration of) sitting in front of a computer and receiving and doing work, it soon became possible to deal with direct clients that way, _after_ acquiring them. Ironically, some of the few survival paths remaining after the transition by clients to AI technology involve things other than sitting in front of a computer, and that will make it quite difficult for most freelance translators.

Essentially, a return to the real world, be it for in-house work or direct interaction with remaining translation consumers not yet using AI, will offer survival to the small portion of freelancers that can take that path. I believe the days of sitting in front of a computer and waiting for work to arrive are numbered. 

Interpreting is a bit safer, and most of my interpreting work these days is online, interpreting between an overseas clients and customers and employees here in Japan. 
Things are not looking good, and, for most, the often-heard “adapting” ultimately means leaving translation for radically different careers; post-editing is not “adapting” any more than doing a final shipping inspection and adjustment of a manufactured product is working as a manufacturer of that product. That said, the supply of low-paid post-editors (adapters) will probably persist for ten or more years, due to the age demographics and bills to pay faced by mid-career translators. 

Bill Lise

*********************************************

リゼ　ウィリアム

Bill Lise

wlise...@gmail.com

https://kirameki-translation.co.jp/

*********************************************

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CAJVTx97RA5UWyYorqp%2BzwwLGhRT%2BOtTeOzZrfKVG6DYinLiAUw%40mail.gmail.com.