Will ChatGPT Really Kill the Chinese Translation Industry?Roman Kierst (小罗)Sam DaviesFour translators and interpreters talk about the impact of new technologies on their work and career prospectsSince ChatGPT burst onto the scene last year, the world has been inundated with stories about how AI will take the jobs of millions. In China’s creative industries, that process is already well underway. Generative AI models can, in a matter of clicks, produce intricate images, succinct copy, and even error-free code.One industry that has operated under the shadow of technology since before the idea of ChatGPT was even conceived, however, is translation. Some thought Google and other software would spell the end of human translators—but they survived. Now, the industry is changing again in the face of large language models (LLMs) like ChatGPT. Here, four translation industry insiders working in English and Chinese describe how AI has changed their work. Dylan Levi King , translator of various Chinese authors including Jia Pingwa , and writer As I started to establish myself as a translator, I had an image in my head of what their practice looked like: They sat in front of a word processor, with the original work cracked open beside them, and pounded out their interpretation. This is still how I usually work, but I’ve come to see I’m an anachronism.When I met the Turkish translator Giray Fidan at an event in Beijing, he was taken aback that I did not employ computer-assisted translation tools. He pointed me in the direction of software that could maintain a personalized lexicon, plugging in automatically the translations of terms and phrases. If I wanted to go further, there were tools that incorporated machine translation more completely.In fact, machines already do most translations. Translation agencies are more likely to engage polishers—people checking through machine translations—than translators.Fewer are willing to admit their use of LLMs than will talk about computer-assisted translation tools. This is down to a popular belief that LLMs possess an intelligence not claimed by other software. These translators are using the tools that they are told will replace them. That’s frightening to them.They admit privately that ChatGPT is useful, however. They appreciate that instructions on tone and register, and word choice can be prepended, and then modified along the way to produce a more appropriate translation.The choice by translators to throw their lot in with the machines owes more to economic conditions than philosophical considerations. As pay rates went down in an age of global competition, very few translators could make a living without these tools. Augmenting outflow with machine translation was the only way to turn out the massive amounts of text required to keep afloat. The human translator got stupider while the machines got smarter. The replacement of most human translators with machine translation and LLMs has already happened.My own practice is still anachronistic. This is partly a philosophical choice. It is deeply unfashionable, but I do believe that the promotion of LLMs and machine intelligence is helping to make us—individually and as a global society—stupider. But my choice is made easier by the fact that I am privileged to work mostly with literary fiction, whose translation rewards a degree of artistry that machines cannot yet be compelled to attempt.I often work with texts, like Jia Pingwa’s The Shaanxi Opera, which, because of their use of dialect phrases, classical allusions, and poetic language, are still unable to be fully parsed by most human readers, let alone LLMs.I am lucky and I am cursed: The technology barons in charge of generative AI development don’t give a damn about literary fiction, which means they’re just like most people in the world. Wei Lincha, translator, former director of the US office of the China Publishing Group When I worked in the United States in 2018, I was asked to review a translated book about the Ningxia Hui Autonomous Region. The phrase 塞上江南, a nickname for Yinchuan, capital of Ningxia, came up frequently. The phrase itself means that Yinchuan, a city north of the Great Wall, is so beautiful and cultured that it looks like one south of the Yangtze River—an area known as Jiangnan (江南)—where China’s most advanced cities have been for much of the last few hundred years. I looked it up on a translation app, and got a nonsensical answer: “plug in Jiangnan.”Frankly, I have often been frustrated by translation technology. One of my friends works with a firm that produces translation devices. On a visit to him at his company, he introduced me to one of their machines, which was quite cute and no larger than a human hand. I turned it on and gave it a go. Typing in the Chinese phrase 五大三粗 (an idiom meaning “sturdy and towering”), I asked the machine to render it into English. The result was “five big and three thick,” much to my amusement and my friend’s embarrassment.It dawned on me that the machine may work well with sentences on a scientific subject. I made it translate a second sentence: “水是由氢气和氧气组成的.” This time the machine worked perfectly: “Water is composed of hydrogen and oxygen.” My friend was relieved. Preview Mode - Subscribe to unlock full content
READ MORE LIKE THIS