The Rise of Automated Translation: Could Algorithms Replace Translators?

In her science fiction novel, The Word for World is Forest, Ursula K. Le Guin writes that in the Athshe people in the fictional planet she describes use the same word for 'translator' and 'god'.

For Le Guin, it seems both gods and translators share a common unique quality: they build bridges between worlds. As machine translation keeps growing, could we soon be invited to share this god-like activity with intelligent robots?

The rise of machine translation

A wide range of human activities directly depends on the provision of high-quality translation and interpretation services. From international trade and business contracts to medicine, legal work, and global politics, translators are truly connecting different worlds with each other.

Translation services make up a global industry that is valued in billions of US dollars and keeps growing every year. According to research, the global language services industry more than doubled in size in the decade from 2009, when it was valued at $23.5 billion to $49.6 billion in 2019. In 2020, it is projected to reach an unprecedented $56.18 billion.

Picture 1 of The Rise of Automated Translation: Could Algorithms Replace Translators?

Photo Source: Pixabay

This figure includes services like language instruction and localization, but an important segment of the industry is the provision of translation services. Unsurprisingly, Europe with its diverse landscape of local languages makes up roughly half of the market, while there is high demand for Spanish content across the US.

Interpretation services are also incredibly sought-after in an array of settings, including scientific conferences and professional meetings. In recent years, tech giants have realized the potential of the global translation market and have begun developing and improving automated translation software.

Google Translate was an undeniable pioneer in the sector. Released back in 2006, it originally used EU and UN documents to mine data and provide machine translation based on statistical models. It remains free and while it is often mocked for its inability to translate subtly nuanced phrases, it has improved vastly over time.

In 2016, Google announced that the service would switch to neural machine translation and has since been much better at gauging context and providing more accurate translation across the 109 languages it supports. Babel Fish, launched in 2012, is also widely regarded as a trailblazer in automated multilingual translation.

Could AI translation eclipse humans?

We have long relied on automation and robots to build things and provide services for us, and as the sophistication of the machines we create has grown so has the range of the sectors they occupy. The automobile manufacturing industry was among the first to become fully automated, with robots replacing human workers on virtually every manual task.

As our ability to build better software expanded, we now use AI for more elaborate tasks. The forex industry now provides the option of auto trading with forex robots, which can imitate traders, manage positions and carry out actions such as setting the lot size and stopping loss. But it is not only in professional settings that robots are there to help us: smart assistants like Apple's Siri or Amazon's Echo are here to make our homes easier to navigate.

Translation seems but the next frontier for AI. Viber has already announced a new feature that allows users to translate conversations in real-time when chatting with their foreign friends, while Skype has also announced the launch of a similar service. Through trial and data mining, the accuracy of such software could gradually reach a level comparable to that of a human translator.

But it seems we are not there yet, as algorithms still cannot grasp the delicate nuances, local slang, or humorous use of language as a human can. Throughout our long history on earth, humans have long regarded language and their ability to communicate as a distinctive feature of what it means to be human, and translators were the ones connecting these different human worlds.

While AI and machine translation is certainly improving, it just seems that we still have a long way to go to reach the point where AI can translate as well as a human does.

Perhaps the way forward, much like in other industry sectors, will be a combination of human translators harnessing the potential of specialized AI-powered software. And while the currently applicable AI translation standards could greatly benefit tourists or a chat with friends, we still need the human touch when it comes to medical or legal translation.