Interface News Reporter | Zhao Yifan
When ChatGPT first appeared, the world was amazed by its super intelligence. However, now, AI is like a huge wave hitting the workers, taking away their jobs.
Recently, according to media reports, the global language learning platform Duolingo cut about 10% of its external contract workers at the end of last year and instead used large models such as OpenAI's GPT-4 to complete translation work and content production. This company is also another one that uses AI as a reason for layoffs.
This news first spread on the overseas social media platform Reddit, where a dismissed translator claimed to have been fired in December of last year. The reason is that AI can complete the work it does in a very short period of time, and the company only needs to retain a small number of employees responsible for inspecting the content produced by AI.
Layoffs have helped many neighboring countries save a lot of labor costs, but the main impact is on external contract workers rather than internal formal employees. At present, the company has around 700 full-time employees.
Duolingo, founded in 2011, mainly helps users learn multiple languages through gamified listening, speaking, reading, and writing methods, covering 40 languages including English, French, and Japanese. It is similar to the familiar applications such as Fluent Speaking and Baicizhan for domestic users, both emphasizing the use of technological means to develop learning content.
Even before the ChatGPT craze swept through the tech industry, this company based in Pittsburgh, Pennsylvania, USA, had already partnered with OpenAI. Currently, multiple neighboring countries have launched three AI features, among which GPT-4 has been applied in one of the most expensive features to cover the cost of calling OpenAI interfaces.
Many companies often use the term "under performance pressure" when laying off employees, but the revenue and stock prices of many neighboring countries are at historical highs.
Previously, Duolingo had suffered losses for 8 consecutive quarters, but after two months of GPT-4 implantation last year, the net loss for that quarter narrowed by 79% compared to the previous year and turned losses into profits in the following quarter.
As of the time of publication, Duolingo's stock price was $214, marking the second time since its listing that it has exceeded $200, far exceeding the top domestic education companies New Oriental's $76 and TAL Education's $11. Before integrating into AI, the stock prices of Duolingo were hovering around $90.
Like many neighboring countries, many technology companies, including NetEase and iFlytek, are introducing AI technology into their translation business. According to data from Global Market Insights, the AI translation market was only $650 million in 2020 and is expected to reach $3 billion by 2027, a nearly fourfold increase.
Compared to traditional manual translation, the biggest advantage of AI translation is its fast speed and low cost. Even in some scenarios with high requirements for translation quality, AI can perform similarly to humans.
A Youdao translation expert revealed to Interface News that the current public quotes from Youdao Zhiyun, Microsoft, and Google range from $6 to $10 per million characters, equivalent to $0.043 to $0.072 per thousand characters. The common quote for manual translation is around 150 per thousand words, with a huge cost difference between the two.
But this does not mean that AI will completely replace human translation in the short term. A person who works as a translator on CNKI said that AI translation of articles related to politics, economics, and history is usually not too bad, and only requires slight manual polishing. However, for literary content, especially ancient Chinese literature and poetry, AI is not very effective.
To completely replace human labor with AI translation, it also requires proficiency in two skills: rich vocabulary data and accurate emotional understanding.
Due to the low frequency of some specialized vocabulary appearing in the corpus, it is difficult for algorithm models to learn specialized vocabulary or expressions from limited samples.
"We need to develop more advanced algorithms or design more intelligent data augmentation technologies to help AI translation systems more accurately grasp proprietary vocabulary and expressions," said a researcher in the field of AI algorithms.
Moreover, AI still lacks sufficient flexibility and precision in handling unique languages such as puns and metaphors. The aforementioned researchers stated that current algorithms still struggle to accurately capture emotions and context in text. For example, when the same word has different meanings in different emotional or cultural contexts, it remains a significant challenge for AI translation.
But with the explosive popularity of ChatGPT, more and more people are indeed worried about being taken away by AI. The latest report released by the International Monetary Fund shows that nearly 40% of global employment is affected by AI, especially for translators, writers, media professionals, etc.
This also gives many companies that actively use AI to reduce labor costs new reasons for layoffs.
Last year, IBM announced that it would replace 7800 job positions with AI and suspend recruitment for positions that AI can handle. Google also stated that it will use more personalized AI tools to replace its advertising sales department. In addition, educational technology company Chegg and cloud storage service provider Dropbox have also laid off employees citing "AI can create more value.".
In the near future, it may be really difficult for migrant workers to maintain their jobs.
(Interface News reporter Wu Yangyu also contributed to this article)