Discover more from Following the yuan
Can you replace language professionals with ChatGPT? | Following the yuan
Chinese language translators and interpreters share their experiences with AI and thoughts on potentially being replaced by it.
My very first full-time job was a competition with machines. Back in 2015, I was hired as a contractor by an international online travel agency as an English content editor. I didn’t know until later that it was an internal A/B test. And humans lost.
I dreaded the work — checking out boxes and writing descriptions of 30 Chinese hotels every day — I felt like a robot but was worse than one because I wasn’t as productive, and I struggled every day. I quit after four months.
The recent ChatGPT hype that started last Nov. in the United States and took over the public discourse in China in recent months called to mind that painful experience. As every industry on earth, especially the ones related to dealing with information, begins to have an existential crisis, I start wondering if this could be the one that hits hard on translating and interpreting, which is something I studied before losing out to the machine (actually I don’t know if it was a software, or AI, or Xenomorph from Alien who looks like it can type ridiculously fast).
If other industries only started rethinking whether they would be replaced now, the interpreters and translators are the Original Gangster in this conversation. Violette Liu, a friend and the co-founder of language services provider InterSpect, said this is what her clients use as an ice breaker since she started working around a decade ago.
Like many interpreters, she had a bad experience with now Shenzhen-listed intelligent speech and artificial intelligence company iFlytek. Around four years ago, she was working at a conference as a simultaneous interpreter where iFlytek was also providing its services: by having a staff listening to her interpretation in real-time and making adjustments on the copies that AI translated, and when the real-time subtitles were projected onto the screen, it looked terrible.
Big Techs and Telcos don’t exactly see eye-to-eye with language professionals. In 2018, Tencent’s AI-driven simultaneous interpreting service did a terrible job at the Bo'ao Forum Asia Annual Conference, although the press release boasted that it could replace humans. In 2013, China Mobile’s former President Wang Jianzhou predicted that the adoption of 4G may mean that simultaneous interpreters are no longer needed.
Simultaneous interpretation, born in the 1945 Nuremberg trials, is considered the highest standard of interpreting. And the humans who are capable of doing it have been deified for their superhuman-like short-term memory and excellence in vocabulary, pronunciation and delivery. “For that reason, it’s convenient [for tech companies] to use us to attract media attention,” Violette said. “We are collateral damage.”
Similarly, William White, a Berlin-based interpreter of the prestigious International Association of Conference Interpreters (AIIC), who works in Chinese, English and French, said that his work hasn’t been affected at all: “So far I’ve not been aware of any clients attempting to replace human interpreters with AI, any more than you’d replace human conference participants with machines.”
While the language professionals I talk to are firm about how they won’t be replaced, they refer to some tools that have been assisting their work. Instead of coming from Big Techs like Baidu, Google or iFlytek, these are made by much smaller firms and are dedicated to the target audience. Computer-assisted translation softwares have long been in use, and DeepL, a machine translation system launched in 2017, has been applauded for providing references in highly specialized industries such as medical and legal areas. Meanwhile, remote interpreting tools such as Kudo, Interprefy, Green Terp, to name a few, have been swiftly adopted by the industry during the pandemic when offline events were canceled.
William said that what has been floated a lot in his field is Computer-Aided Interpretation that equips interpreters with a live machine-generated transcript of the original speaker. “But that can’t substitute for your actual output as an interpreter, it’s more like having all the right things Googled for you in advance,” he added.
For translation work, the professionals’ reaction is a bit more ambiguous: they believe that AI currently can’t do that well in getting the cultural contexts right, but with time, it would improve.
Andrea Deng, a young translator who’s focused on game localization and movie and drama subtitling, believes that the industry should embrace emerging tech. Because if it doesn’t, it would be forced to at some point.
She recently tested a paragraph of creative content in Chinese in ChatGPT and asked it to translate it into English with context and use cases provided, but the results were “rather unnatural”. However, she found it worked well if she asked ChatGPT to improve fluency on her own translation.
“I believe the human-AI cooperation mode will persist for a certain period of time before AI becomes highly intelligent enough to handle complex translations," she said.
So what happens if language professionals need to change their roles with the evolution of AI?
“The skills we learn when we are training to be translators and interpreters are highly transferable,” Miguel Fialho, a senior lecturer at the University of Bath, my alma mater, who taught me back in the days and Andrea recently, assured me.
Take consecutive interpreting for example, “We are analyzing information, putting bits and pieces together like a puzzle, using our memory and visualization to figure out the message or gist of the speaker. At the same time, we are code-switching and we are controlling our nerves while speaking in front of a large group of people,” he said. “I can think of dozens of jobs in which those skills would be invaluable.”
“At the core of our education system is the idea of critical thinking,” he continued. “Translation and interpreting are no exceptions to this.”
That I believe, as I hope that I applied those skills well in journalism. I realized that instead of ChatGPT per se, we are talking about a symbolic movement of whether humans can be replaced by machines. It's a deep insecurity that boggles humans for ages and language is just a small part of it.
Now that I have more perspectives, I’m glad that the machine replaced me in 2015 – maybe that job should never exist. The machine can’t quit and can’t find the work boring, so why don’t we let it do the job we don’t want to do and spare us for the more valuable things?
Thanks for reading Following the yuan! Subscribe, like and comment to support my work.