Statements by ATA and The European Council of Literary Translators’ Associations on AI

[John Stroman]

Fellow 'Yakkers,

To those of us living in the US, I hope you enjoyed a wonderful Thanksgiving holiday.

Today's Slator newsletter features summaries of statements issued by ATA and CEATL regarding AI. Here are the links:

ATA: American Translators Association Issues Statement on Artificial Intelligence

https://slator.com/american-translators-association-issues-statement-on-artificial-intelligence/

CEATL:  "Machines are Not Translators"

https://slator.com/machines-are-not-translators-europes-literary-translators-push-back-on-ai/

The CEATL statement is understandably stronger than the ATA statement considering it focuses on literary translations.

John Stroman

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

[Bill Lise]

Interesting that they quoted me. 

>>>

“I think there are two things at work, looking out for corporate members and not wanting to bring bad news to paying individual members,” wrote another observer. He added. “[I]t does not serve the interests of translators to dance around the issue and just talk about how good or bad AI is, including a claim that translators are leading the way.”<<<

At least they gave links to my statements on LinkedIn; the links didn’t survive the copying I just did.  

What they didn’t do was note my suggestion, made elsewhere, that the organizations stop ignoring the issue of not having customers to purchase your translations after you adopt the technology, because the agencies are using the technology themselves and obviating the step of human translation (at least for them).

The translation organizations seem very reluctant to mention acquisition of direct clients as at least a temporary survival strategy. It is a terrifying prospect for some and an impossible goal for many. The life expectancy of translators with direct clients is greater than that of those working for agencies, but it is also going to be shortened. 

--
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 on the web visit https://groups.google.com/d/msgid/honyaku/CAJVTx94pv%3DzWgDWJi5UKMBVdQMMAScUiMWGB3sWtxDgSkE9kAw%40mail.gmail.com.

[Bill Lise]

The full post from me that they quoted from:

>>>

Another post by ATA about AI in which they totally ignore the reality that translators are losing translation work because agencies have adopted AI. I realize that ATA has corporate members and I realize that this bad news is painful, but it does not serve the interests of translators to dance around the issue and just talk about how good or bad AI is, including a claim that translators are leading the way. Unless translators find sources of work other than AI-using agencies, the outlook is extremely grim. Translation work from agencies is going to end. In fact, that has already started. Wake up, guys.<<<<

[John Stroman]

Bill,

Thanks for letting us know that "another observer" is a highly authoritative source of accurate information. 

I seldom work with customers based in Japan anymore. The pandemic and the dismal exchange rate were the primary factors in my decision, so I may change my mind in the future. 

Based on my experience, I think it is safe to say that both agencies and direct customers in the US mainly require translations into English from Japanese merely as one of many foreign languages. Obviously, their goal is to obtain the best possible rendering of a document into English at the lowest reasonable cost. In other words, the English-speaking customers in the US decide whether a J>E translation is "good enough" for their particular needs. Consequently, human J>E translations only for gathering information have essentially been replaced by MT/AI.  Many agencies have turned to post-editing by human J>E translators for more fluent renderings, and it is only a matter of time before direct customers do the same. 

Even so, highly context-sensitive work still appears to be available in the US at reasonable rates for human J>E translators with experience, expertise, and good writing skills in technical fields that demand factual accuracy, in the advertising, instruction manuals, etc., of products where a  translation error may expose the customer to future legal liability, and in literary translations. The tasks that many experienced translators in the US now face involve incorporating technological advances into their workflow to increase productivity and expanding their fields of expertise to obtain new customers. I think the range of opportunities for human J>E translation as a career has definitely narrowed in the US, but I do not think we are doomed to extinction.

John Stroman

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

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 on the web visit https://groups.google.com/d/msgid/honyaku/CAJVTx94pv%3DzWgDWJi5UKMBVdQMMAScUiMWGB3sWtxDgSkE9kAw%40mail.gmail.com.

--
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 on the web visit https://groups.google.com/d/msgid/honyaku/CAD0Wv_KHsLDtcJGxxxCqpDuWaV-cd7gqWtbhU3y_%2BYz1mpT7aw%40mail.gmail.com.

[Bill Lise]

As a point on the graph, I do a considerable amount of reader-driven translation work for US law firms, and none has balked at paying 30 yen or more per English word. But that work is apparently slowly going away because of MTPE. 

But in general, I think that it is safe to say that the degree of disturbance by the MTPE business model differs greatly, not so much by location, but rather between reader-driven and writer-driven translation, the latter being less likely to rely on MTPE. 

Regarding adopting technology, I certainly think that is a good idea.

The problem is that simply adopting technology will not necessarily solve the problem if agencies don’t order *translation* work from humans.  And now there is significant talk of having AI do the post editing as well, as amazing as that might seem.

Direct clients (certainly in Japan, anyway) are later adopters of that technology and offer at least somewhat of a delay in the demise of human translation, but even many direct clients will adopt MTPE. At a Japanese patent attorney association event about six years ago, I was told by one benrishi I know that their firm was already using MT with post-editing.  That said, Japanese patent firms are putatively acting as translation agencies. 

I don’t think there is an easy solution to the problem for translators. There will likely remain agencies that will continue to use humans, but they will be able to supply translation work to only a tiny portion of the human translators. The large players are well on their way to a MTPE business model approach. 

There are vast numbers of translators in the US that have done discovery document translations for litigation. I have seen their work over the years. Almost none of their names are familiar to me. They appear to be a large silent core of workers.

Their work was largely taken over by translators in China and India. I’ve seen their work also in depositions when documents are presented to deponents. 

Even those translators are now being replaced by MT. 

The fat lady has not sung yet, but she’s waiting in the wings. 

To view this discussion on the web visit https://groups.google.com/d/msgid/honyaku/CAJVTx97ZJRHmAucdtZ7XXqp9pibfN57W0wtRpHePgY010kEipg%40mail.gmail.com.

[John Stroman]

Bill,

Points well taken.  A few weeks ago Tom Galley introduced to Honyaku a video interview with an agency owner in the Czech Republic, and one takeaway is the question of legal liability for translation errors that result in damages. 

It appears that there is no clear precedent for a lawsuit filed against a machine. For example, if a self-driving vehicle causes an accident, does the liability lie with the human (non)operator of the vehicle or the manufacturer under the umbrella of mechanical failure or false advertising? Does liability for a costly translation error lie with an MT/AI engine, a post-editor, or the company that outsources the job to said post-editor?

In my own case, I have used both MT and AI as glorified typists to guard against omissions, my greatest source of errors, but I still have to rewrite every output sentence to match the context and "voice" of the writer, as well as the extensive glossary of boilerplate terminology provided by my customer. I still consider my work to be human translation that is aided not by the bookcases full of dictionaries and reference books of the past, but by translation memory software on my desktop and online resources that have to be read carefully and checked for accuracy. I'm still performing the same work but with updated tools.

John

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

On Fri, Nov 24, 2023 at 2:41 PM Bill Lise <wlise...@gmail.com> wrote:

As a point on the graph, I do a considerable amount of reader-driven translation work for US law firms, and none has balked at paying 30 yen or more per English word. But that work is apparently slowly going away because of MTPE. 

But in general, I think that it is safe to say that the degree of disturbance by the MTPE business model differs greatly, not so much by location, but rather between reader-driven and writer-driven translation, the latter being less likely to rely on MTPE. 

Regarding adopting technology, I certainly think that is a good idea.

The problem is that simply adopting technology will not necessarily solve the problem if agencies don’t order *translation* work from humans.  And now there is significant talk of having AI do the post editing as well, as amazing as that might seem.

Direct clients (certainly in Japan, anyway) are later adopters of that technology and offer at least somewhat of a delay in the demise of human translation, but even many direct clients will adopt MTPE. At a Japanese patent attorney association event about six years ago, I was told by one benrishi I know that their firm was already using MT with post-editing.  That said, Japanese patent firms are putatively acting as translation agencies. 

I don’t think there is an easy solution to the problem for translators. There will likely remain agencies that will continue to use humans, but they will be able to supply translation work to only a tiny portion of the human translators. The large players are well on their way to a MTPE business model approach. 

There are vast numbers of translators in the US that have done discovery document translations for litigation. I have seen their work over the years. Almost none of their names are familiar to me. They appear to be a large silent core of workers.

Their work was largely taken over by translators in China and India. I’ve seen their work also in depositions when documents are presented to deponents. 

Even those translators are now being replaced by MT. 

The fat lady has not sung yet, but she’s waiting in the wings. 

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 on the web visit https://groups.google.com/d/msgid/honyaku/CAD0Wv_LpPzdCn6%2BuEMpesd%3DabAqERZuFXB%3DZfMXxUEDu2vJETA%40mail.gmail.com.