Gizmodo fires Spanish staff amid switch to AI translator
arstechnica.com
arstechnica.com
also current MT benchmarks are completely broken for whatever reason (benchmark itself and/or evaluation model ). They don't reflect reality at all.
Yeah, I have no idea how the likes of DeepL and other paid-for translators are going to survive long term when "GPT-4 and friends" are that good. What's their business plan?
DeppL tried to compensate by making their front-end more annoying and hijacking clicks and highlights which invalidate the browser spell checker, making it impossible to correct a word via click, which I guess is one way of expediting their demise.
The sad part is, in my small town there's the office of another small ML driven translation start-up that was way better than Google when it launched a few years ago, but now when I check their careers page, they're not hiring anymore which I guess says enough about their business and future growth plans. Dead man walking.
Less so than hiring a human but not cheap enough that you’d want it to automatically translate long websites as a browser extension or anything like that
A friend who translates Korean has pointed out that it's kind of special, in that the symbols used are exactly the same for many different things. So the only way to know the actual meaning is from context, which is what then needs translating.
Apparently online translators from Korean to other languages are truly, horribly bad due to this.
Hangul(Korean writing system) is not a logograph like Chinese/Japanese but an alphabet. Many words do have different meanings but all languages have this to an extent. I don't think that's the reason Korean is context dependent.
The bigger reason is that Korean is very pro drop.
Strictly speaking 친구 있어요 ? means Friend Have ?
And you would say exactly that for any of these English equivalents
"Do they have friends ?"
"Do you have a friend ?"
"Do you have friends ?"
"Does he have a friend ?"
And so on.
Plurality as a special marker doesn't really exist. Possessive and Article Determiners don't really exist either. Articles don't exist at all. Possessive markers are few and only used when surrounding context might not be enough to convey the needed information.
Subject/Object/Topic markers exist but are regularly omitted.
You might ask, well how would you know which of those equivalents 친구 있어요 means ?
There is no way to tell before reading or hearing the surrounding context.
Traditional models are unable to decipher this context and translations are only good with close language pairs. Google translate can be very good for English to French and Japanese to Korean but fail on Korean to English.
---
"Hard by a great forest dwelt a poor wood-cutter with his wife and his two children. The boy was called Hansel and the girl Gretel. He had little to bite and to break, and once when great dearth fell on the land, he could no longer procure even daily bread. Now when he thought over this by night in his bed, and tossed about in his anxiety, he groaned and said to his wife: 'What is to become of us? How are we to feed our poor children, when we no longer have anything even for ourselves?' 'I'll tell you what, husband,' answered the woman, 'early tomorrow morning we will take the children out into the forest to where it is the thickest; there we will light a fire for them, and give each of them one more piece of bread, and then we will go to our work and leave them alone. They will not find the way home again, and we shall be rid of them.' 'No, wife,' said the man, 'I will not do that; how can I bear to leave my children alone in the forest?--the wild animals would soon come and tear them to pieces.' 'O, you fool!' said she, 'then we must all four die of hunger, you may as well plane the planks for our coffins,' and she left him no peace until he consented. 'But I feel very sorry for the poor children, all the same,' said the man."
---
And here is GPT 3.5's final rendition of it, after converting it to Japanese and back:
---
"Near a large forest, a poor woodcutter lived with his wife and two children. The boy was named Hansel, and the girl was Gretel. They had little to eat, and during a severe famine in the country, they could no longer even provide daily meals. One night, as he lay in bed pondering this, the woodcutter, filled with anxiety, said to his wife, "What will become of us? How will we feed our poor children when we no longer have anything for ourselves?" His wife replied, "Husband, tomorrow morning, let's take the children into the forest. There, we will light a fire and give each of them one more piece of bread. Then, we will go to work and leave the children on their own. They won't be able to find their way back, and we will be freed from them." The man said, "No, wife, I cannot do that. How can I leave my own children alone in the forest? Wild animals will come and tear them apart." She retorted, "You fool! In that case, we will all four perish of hunger, and you can just as well plane the boards for our coffins," and she didn't leave him in peace until he agreed. "But the poor children are so pitiful," the man said."
---
It's a generally terrible translation with a simple (though heavily stylized) text. Not only is absolutely all style, imagery, and character eliminated from the writing, but it's also overtly nonsensical in parts. Lighting a fire for the children ends up translated as simply lighting a fire, as if the goal is to start a forest fire. Similarly the man's sorrow for the children ends up being completely lost, and is translated as him simply calling his children pitiful. I'd be curious to see GPT4's output on this, or ideally even more challenging paragraphs.
[1] - http://www.literaturepage.com/read/grimms-fairy-tales-72.htm...
The fun part of chatgpt is you can ask it to do non-standard translations like
"can you give me the direct non idiomatic translation of the following Japanese into English, to help me remember these phrases in English words but the Japanese way of saying it"
Sure, I can provide a direct, non-idiomatic translation of the Japanese text into English. This kind of task is straightforward for me, so no worries.
"Big forest's right next to, poor wood-cutter and his wife, and two children were living. Boy was Hansel, girl was Gretel called. He had almost no food, once big famine attacked the country, daily bread even could no longer obtain. Night, in bed this thing thinking, anxious tossing and turning, he groaned, to wife said: 'We what become? Poor children to how food provide, ourselves even nothing when.' 'Husband, tomorrow morning early, children to forest's thickest place take let's go. There for them fire light, each one more piece of bread fragment give, then we to work go, them alone leave. They home return way find not become, we them let go can.' 'No, wife,' man said, 'that can't do, why I children in forest alone leave can? Wild animals soon come, them tear apart will.' 'Oh, fool!' she said, 'then we four people all die of hunger, our coffins' plank plane might as well.' And, she him agree until him peace not did. 'But, still poor children very sorry,' man said."
It doesn't give us a rough first draft. It gives us the equivalent of a second or third draft. The equivalent to the final result we'd get back from translators after a lot more back and forth.
Honestly, it's better or equivalent to any human service I've used. Usually when it translates badly, a human would also translate badly and it means that section needs rewriting.
It handles all kinds of things that I wouldn't expect, idioms, jokes, technical writing. There are very few mistakes.
I can only imagine it's even better at Spanish.
We don't pay for human translation anymore. The workload is reduced enough that we do the final editing ourselves.
We still use human translation for legal stuff because that has to be done by an official translator. So I guess that's safe.
Or do you mean official verified and stamped translations?
PS: Not to mention, Gizmodo also cut ORIGINAL spanish language journalism, when at least this part still would add value alongside the translated articles. Maybe they really don't care about the spanish speakers.
Some “AI artists” even specify the name of a real life artist in their prompts.