Great article by Welocalize: AI Is Changing Localization — People Still Define It

[Carl Sullivan]

Yakkers,

Sharing a great article by Nicole Sheehan at Welocalize: (23) AI Is Changing Localization — People Still Define It | LinkedIn

Accessible on LinkedIn, or through Welocalize. 

There is often discussion here about the industry going away. I have mentioned that I am still pretty busy--this article deals well with that topic. It is worth a read. 

Today's offering for your reading pleasure...

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

[Tom Gally]

Carl,

Thanks for that link. It’s good to read that perspective.

Out of curiosity: Has the type of the work you do changed at all since the arrival of GPT-4 three years ago? Are there any kinds of work that you used to do but that have since dried up? Are you getting more of some types of work than you used to?

Tom Gally

--
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/CA%2BQjO-mGhkAYKTR%3Dnf-L%3Dc3sCtczf8QSPwgWUSQT378r9A4iTA%40mail.gmail.com.

[Carl Sullivan]

Tom,

Great question. Wonder about you and others? Yes; a lot has. I do some patents straight up still, but largely I am hired to do post-machine editing of them (although I pretty much retranslate them because the post-machine edits I consider to be pretty unreliable for patent work, so the jobs are a bit less cost-effective for me). I do a lot of pharma that is AI driven; either translation or editing of someone else's edits. Legal/criminal case types of docs--fewer but still getting them. Medical reports, etc.  Business docs--contracts, etc.; slow, but they come. Just took a big medical report job today. A few companies have dropped off of my radar screen, but I still have a core group of companies that continue to work with me. 

Where I used to be overwhelmed with work, now it seems that jobs appear just about the time when I need one to appear. I seem to make my target each month; this may just because God is taking pity on me and helping me out here.  25 years now of full-time translation; it has been a good ride. We will see what the future brings. I wake up just trying to do my best and grateful for each breath I take. 

Carl

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CACnS3NeP3M6XAyAKkeXhg0wB6%2B7zOAi9jECF7mB_wbFcER7s3A%40mail.gmail.com.

[Tom Gally]

Carl,

It's good to hear that you're keeping busy. Probably the variety of work you do is one reason.

I’m not a good datum, as all of my translation work is for the university where I have been a faculty member for more than twenty years. My insider status and knowledge are the main reasons why people continue to ask for my help with translation and editing: I can make judgment calls about how to word high-stakes documents in a way that would not be possible for outside agencies, freelancers, or AI. Also, I’m being paid a salary, so there’s no extra expense in asking for my help.

That said, the quality of the English produced by Japanese staff on their own has increased markedly over the past year or so, clearly because they are learning how to use AI. A couple of weeks ago, for the first time, the English in a document I was asked to check—a letter on a sensitive subject—was perfect. There were no changes I felt justified to make.

Sometimes the staff wouldn’t need my help if they had been given a few tips about how to prompt LLMs. Just today I checked a script for a spoken presentation that the author had had translated by AI. As a text translation, the AI version was completely acceptable. But I could see that it would be difficult to understand when read aloud to an international audience by a native Japanese speaker, so I had Claude rewrite it with shorter sentences and somewhat simpler vocabulary and syntax. I made a few further improvements, but Claude’s version would have been fine by itself.

Tom Gally

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CA%2BQjO-%3DxGtMrSQkGyOpCDQV4UDD-Zw18DEWDcdROibfmUW14FQ%40mail.gmail.com.

[Carl Sullivan]

Tom,

Thank you for that. Very interesting. I haven’t been in Japan for a while; it would be interesting to see how people are doing using AI-assisted tools just to communicate. 

Well, we’re gonna ride this pony to the end. All we can do is hold on and see where it goes.

Best to you.

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/CACnS3NdLZQdn7ZRsu5QjWp%2BCGXMrGEQeATks6LriQrFqB2%3Da2w%40mail.gmail.com.

[Tom Gally]

Carl wrote:

it would be interesting to see how people are doing using AI-assisted tools just to communicate.

I’ve been trying to pay attention to that since later 2016, when Google Translate got sort of good enough to use in some practical contexts. Within a year or two, many Japanese people I met who worked in multilingual or international contexts said that they used machine translation in their jobs. The most common use was to translate incoming emails and other documents from English into Japanese to get their gist. People used MT for outgoing documents as well, though the results wouldn’t have been very good. The quality of LLM-based MT in both directions is much better now, of course, especially if people know to use it well.

As I was involved in research on English education policy then, I was also interested in whether people could use interpreting apps to converse across language barriers. Around 2017, I did some experiments with pairs of graduate students from Japan, Poland, China, and one or two other countries, in which they would try to have a conversation, each speaking a language the other didn’t understand, using the speech-interpreting function of the Google Translate app. The conversations inevitably broke down after three or four turns, with one or both people unable to understand what the MT output was supposed to mean.

When I took a taxi in those days, I would ask the driver whether foreign passengers tried to communicate with them with language apps and, if so, how well it worked. I got mixed responses. Some drivers said that the apps were useful, while others said they often had trouble understanding what the app was saying. One problem seemed to be that most of the customers were Chinese tourists, they would say their destination using the Chinese readings of the kanji, and the app would be unable to convert those to the Japanese readings.

I haven’t tested speech interpretation apps lately, but they should be much better than in 2017. Last year, my two sisters and their husbands—all in their seventies—spent three weeks in Japan. None of them have ever studied Japanese. I spent about a third of that time with them, but the rest of their stay they were on their own. They had very few language problems. Once, I went with them to a drugstore to buy cold medicine. I was expecting to have to read the labels on the medicine, but they just pointed their cameras at the packages and were able to find what they wanted without my help. One day, they went into a café in Osaka and had a long, informative conversation with the elderly proprietor using the Google Translate app.

When I talk to Japanese people about AI, one of the most common ways they report using it is for translation. Just a couple of days ago, I led a discussion with some teachers at a university about AI, and one mentioned how some students in her English classes, instead of trying to read textbook passages themselves, take pictures of the texts and have ChatGPT etc. translate them into Japanese.

I did something similar myself today. I noticed a sign on a railway platform in Japanese, English, Korean, and Chinese. The English was fine, but I was curious how good the Korean and Chinese might be. So I took a picture of the sign, gave just the Korean and Chinese parts to Gemini, and asked Gemini how good the language was. I don’t know any Korean and I don’t know Chinese well enough to evaluate its naturalness, so I can’t assess Gemini’s assessment. But at least the difference between its ratings of the two languages is interesting:

https://gemini.google.com/share/f7de2caa8a1a

Tom Gally

To view this discussion visit https://groups.google.com/d/msgid/honyaku/CA%2BQjO-kQsYumtQyUx0M8Hv7vAODhdOmR6Fx07XCauB9%2BT3dcBg%40mail.gmail.com.