Dangers of machine translation

[Warren Smith]

https://restofworld.org/2023/ai-translation-errors-afghan-refugees-asylum/

 

 

“That’s why you need human attentiveness. The machine, it can be your friend that you use as a helper, but if you’re using that as the ultimate [solution], if that’s where it starts and ends, you’re going to fail this person.”

 

I worry about this sometimes... ChatGPT, can lull somebody into "inattentiveness."

 

I also worry greatly about the "anchoring effect" with machine translations. The machine reads something incorrectly. Once the human translator reads the suggested translation, s/he cannot "unread" it. It has to bias our perception, even if we are diligent. But I don't trust translators to be diligent... AI makes it too easy to "cheat" (and too easy for unqualified translators to pretend to be qualified...)  

 

I worry about our industry. How do we use these new tools and still maintain quality?

[Kai McClary]

Hello Smith-san,

I, too, am worried about the translation industry as a whole. If machines such as ChatGPT are seemingly improving every day, I worry how soon they might be the dominating force in the industry. It seems difficult--nigh impossible--to get a foothold starting out when there are machines who are more attractive and less of a liability for consumers. Of course, it's not all "doom and gloom". The benefits of human translators still hold more sway than its AI counterparts. Still, I can't help but worry if(and seemingly when) it overcedes human translation capabilities.

Kind regards,

Kai McClary

novice translator

--
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/FA9CCD339D2948809ADC302E36512914%40WarrenSmithDell.

[Matthew Schlecht]

On Wed, Apr 19, 2023 at 7:23 PM Warren Smith <warren...@comcast.net> wrote:

 

“That’s why you need human attentiveness. The machine, it can be your friend that you use as a helper, but if you’re using that as the ultimate [solution], if that’s where it starts and ends, you’re going to fail this person.”

A good analogy is the use of AI-driven cars, which still come with strict instructions that a human being's hands be on the wheel and foot be alternately on the accelerator and brake.

Without a human in the loop, things might go OK for much or most of the time until that white triple-bottom semi-tractor trailer changing lanes in front of you is read by the AI as the clear horizon.

 

I worry about this sometimes... ChatGPT, can lull somebody into "inattentiveness."

Mmmmm, yeah.

 

I also worry greatly about the "anchoring effect" with machine translations. The machine reads something incorrectly. Once the human translator reads the suggested translation, s/he cannot "unread" it. It has to bias our perception, even if we are diligent. But I don't trust translators to be diligent... AI makes it too easy to "cheat" (and too easy for unqualified translators to pretend to be qualified...)  

 

I worry about our industry. How do we use these new tools and still maintain quality?

To me, the answer is fairly simple.

You the human still insert yourself into the loop at one or more points, and at a minimum your final once-over (and whatever previous steps) before delivery should satisfy your own established quality criteria.

If your AI has evolved enough to produce equivalent quality to that which you once produced using your own wetware, you still need to do a once-over to ensure that is a valid assessment.

The catch will be convincing clients that your paid input is still crucial.

Or, maybe they will choose to cut costs and only worry about that white triple-bottom semi-tractor trailer changing lanes in front of you once it happens.

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

[Warren Smith]

Matthew Schlecht, PhD, wrote:

A good analogy is the use of AI-driven cars, which still come with strict instructions that a human being's hands be on the wheel and foot be alternately on the accelerator and brake.

Without a human in the loop, things might go OK for much or most of the time until that white triple-bottom semi-tractor trailer changing lanes in front of you is read by the AI as the clear horizon.

 ------

Exactly.

I worry about an overly automated car for the same reason.

I took a test drive in a car with a lot of fancy self-driving features (mostly in terms of being able to stay in the proper lane and keep a proper distance from the car ahead). The next day I left on a long drive from New Hampshire to Utah. During the drive I grew sleepy a few times, and pulled off the road for a nap, and took a hotel room for one night.

 

I wondered -- if I had had the car with the self-driving functions, would I have been tempted to push things a bit more? Would I blink my eyes slightly longer.... be more likely to fall asleep at the wheel? Would the net result have been that a long drive would be MORE dangerous? With AI cars, will more inebriated people say, "I've got this," and climb behind the wheel when otherwise they would know better?

 

Yeah -- it's the same question of discipline.

 

What I DO know is that some of the big translation houses are NOT exercising the discipline to use skilled translators, where "bilingual editors" and AI can do a convincing job.

 

W