ChatGPT vs Professional Translation: Similar Quality, Thousand Times Efficiency, and One-hundred-thousandth Cost

About a year ago, I chatted with Xu Mingen at Muka Cafe in Wanhua District, Taipei, and he asked for his opinion on the further development of the block potential . My opinion is that the block potential is dominated by personal brand and value. It is not appropriate to add writers or accept business matching. It is recommended that Ming En translate the article into English and develop overseas markets. That is a "low-hanging fruit" that can not only maintain the purity of self-media, but also add a new dimension of development.

Of course it is wrong to think too much of yourself, but it is not necessarily good to belittle yourself. Taiwan and Hong Kong are always satisfied with imported products, watch Hollywood movies, read translated books, and sometimes underestimate the quality and value of their own works. In fact, many excellent works in the two places are limited to the local market only due to factors such as cultural and language barriers, market inertia and sales channels. Block potential is an example. The quality of its articles is already at the world level, and it is not inferior to the English market. It has been living in the Tower of Babel just to focus.

Although I am not as good as Ming En, I am still a hard-working writer after all. I have long wanted to translate my own works, but unfortunately I have not been able to do so for many years. However, recently ChatGPT finally let me get my heart out and realize my long-term wish. From now on, I have achieved close to simultaneous publication of the English version of the weekly article:

ChatGPT vs Professional Translation: Comparable quality, a thousand times more efficient, and one-hundred-thousandth the cost

Comparing Google, DeepL and ChatGPT

AI translation is nothing new. Google Translate, which has been in operation for many years, is already quite good, and DeepL, which focuses on translation, is even better. But I have not used AI to translate works, except because of laziness, the quality is always a little bit worse, which is the main reason for my deterrence.

The attention of ChatGPT is already high enough, and I don’t want to deify it further. In fact, the development of AI and other technologies is slowly accumulating. Even if there is a breakthrough point, it is not achieved overnight by ChatGPT, but the continuous improvement of AI technology to just surpass critical point results.

The quality of ordinary sentences translated by Google Translate, DeepL and ChatGPT is similar, and the comparison is of little significance. Therefore, I selected some passages with Cantonese and Teochew, which are particularly difficult, to illustrate the differences between the three.

Zuojiao’s introspection: NFT books are difficult to understand because they failed to break the law of "fish ball theory" <br class="smart">Emailed New Year greetings with writer friend Alice, both congratulated and chatted, and naturally talked about NFT books. Alice said, "A weak question, so what you get when you buy an NFT book is a cover of a different style..."
This sentence is no small matter. First of all, "weak and weak question" means that there is no money here. When a question is really weak, the questioner will not add this preface.
ckxpress.com

Zuojiao's introspection: NFT books are difficult to understand because they failed to break the law of “fish egg theory”
Emailed New Year's greetings with writer friend Alice, both congratulated and chatted, and naturally talked about NFT books. Alice said, “A weak question, so what you get when you buy an NFT book is a cover of a different style…”
This sentence is no small matter. First of all, “weak and weak question” means that there is no money here. When a question is really weak, the questioner will not add this preface.
Google Translate

Left glue's introspection: NFT books are difficult to understand because they fail to break the law of “fish egg theory“
When I emailed Alice, a writer friend, to congratulate her and chat, I naturally talked about NFT books. Alice said, “I'm sorry, but the NFT book I bought is a different cover…”.
This is a very important statement. First of all, “weak question” is the same as having no money here. When a question is really weak, the questioner will not add this preamble.
DeepL

Zuojiao: Google simply surrendered to Pinyin; DeepL translated it into "left glue"🫠

Weak question: Google's "a weak question" is too Chinglish, and DeepL's "I'm sorry" is much more natural. But when "weak and weak question" appeared for the second time, both of them used different translations, the context was wrong, and Google's "weak and weak question" was even more ridiculous.

There is no silver here: Google's "there is no money here" and DeepL's "having no money here" are theoretically correct, but foreigners must not understand it.

And look at the performance of ChatGPT:

Self-reflection of the leftists: The difficulty in understanding NFT books is due to the failure to break the “fish-egg theory” law
During my New Year's email exchange with writer friend Alice, we naturally talked about NFT books. Alice asked, “Can I ask if buying an NFT book is like getting a different cover?”
This statement is not insignificant. Firstly, “Can I ask” is equivalent to “I don't have much to contribute here.” This indicates that the question is not trivial.
ChatGPT

"Zuojiao" is translated as leftist. I don't know if the translators have better suggestions, at least I can't think of it. "Weakly ask" has become "Can I ask", which is a bit lighter, but it is also natural. Maybe there is no more corresponding expression in English. After all, "Weakly ask" can be regarded as a trendy language. I don't know whether it is luck or strength, the second time the phrase appears, it is also translated as "Can I ask", expressing the semantics correctly.

The most difficult part "There is no silver here" was cleverly avoided, and it was rewritten according to the context as "I don't have much to contribute here", which is considered very good.

The only thing that disappoints me is that I have never been able to see the correct translation of "fish ball" into "fish ball", which ignores Hong Kong culture.

Overall, ChatGPT gives me the feeling that it is like a person who can digest the meaning of the entire text and write it out in English, while the original AI translation tools are more "determining the meaning of words", especially Google.

Last week, I used ChatGPT to translate the article <Solution looking for a problem as a path to entrepreneurship> . One of the paragraphs in particular can illustrate the above observation. The article talks about McDull and Mai Tai. When Google translated Mai Tai into "Mai Tai" horribly, DeepL translated it into the correct "Mrs. Mak", and ChatGPT knew how to translate it into "McDull's mother", which is really interesting.

If you are interested in learning more, you may wish to put the following short paragraph into various translation tools, and leave a message to share your observations:

McMug Comics has such a short story. Mai Tai asked McDull why he always eats a lot. The kind-hearted McDull said that if you eat a lot, you will have strength, and you can open bottles for others.

A cost gap of one hundred thousand times

After a shot in the arm, I took another step and asked ChatGPT to help me translate two books.

Thanks to the open source community, I cloned bilingual_book_maker from github, set up the environment, and threw the entire "Blockchain Sociology: Reimagining Money, Media, and Democracy" to gpt-3.5-turbo API for translation.

Because I don’t know how long it will take to run, I deliberately executed it before going to bed at 3 o’clock in the middle of the night, thinking that I would let it run slowly for several hours, and then check it after waking up, but the result is that I have been staring at the screen, observing the quality of the translation, and still I didn't fall asleep. After 1 hour, 4 minutes and 55 seconds, the translation of 100,000 words was completed, and all the sentences were fluent. The availability rate was 99%. There were only a few mistakes, most of which were names. Then I look at the cost and it's a whopping $0.33.

I used the same method to translate about 80,000 words of "The so-called "I don't invest", that is, all in legal tender" . The quality is similar. This time it took 1 hour, 36 minutes and 43 seconds, and the cost was 0.36 US dollars.

In comparison, the traditional human translation takes a month, which is fast, not counting the communication between the two parties and clarifying the meaning of the original text. As for the cost, the median price of traditional translation is about 1.2 Hong Kong dollars (4.5 Taiwan dollars) per word. In other words, it costs more than 200,000 Hong Kong dollars to translate "District", which is 100,000 times the cost of AI translation. I don’t mean professional translators to bargain at all. They do need a lot of time and energy. However, the income I get from a book is only tens of thousands of Hong Kong dollars, and the average income per word is less than 0.4 Hong Kong dollars (1.5 Taiwan dollars). Don't use it all for translation, and it's far from being able to afford the market price. The objective result is that most of the works in the world are not blessed with translation, and AI is about to change this.

The cost and efficiency gap between AI and human translation is too great, and the theory of AI manufacturing unemployment will definitely appear again. It is terrifying that half-life cultivation bases are easily replaced by machines overnight, and the society must pay attention. But optimistically, there is another possibility. When the level of AI translation is high enough, the translated text will be a thousand times or even ten thousand times that of the past. Although 99% of the process is changed to AI, the remaining 1% is for "workers". The demand for "intelligence" may be much greater than the current translation market.

To put it bluntly, no matter what kind of profession, facing AI, either make good use of it, or have the strength to reach the level where even it can't reach the 1% level, so as to have a chance to avoid being eliminated.

AI translation tipping point has arrived

I'm not sure whether the saying "AI singularity has come" is exaggerated, but what is certain is that as an author who uses ChatGPT to translate popular science texts, its level has surpassed the qualification point I required (at least translated into English but not language), the speed is enough for me to publish simultaneously, and the cost is so low that I have nothing to say.

It was too late to say that, when I was writing, I just received an email from OpenAI titled "GPT-4 is here", and immediately switched to the GPT-4 model, and then translated the above paragraph again, and the resulting title was translated as:

Self-reflection of the Leftist: NFT books are hard to understand because they fail to break the “Fishball Theory” law
GPT-4

The picture I was most looking forward to appeared.

further reading

1. ChatGPT, Midjourney are hot, we are all AI proletariat
2. Artificial intelligence is not necessarily good, and people-oriented is not necessarily right
3. Zuojiao’s introspection: NFT books are difficult to understand because they failed to break the law of “fish egg theory”

My articles are not limited for profit, all are open for reading. If you like this article, please bookmark its Writing NFT, support writing, and preserve news.

Original link: chungkin Express