Google Translate doubles language support thanks to AI

Google has added a panoply of new languages to its translation app as it works towards creating a service that can translate 1,000 languages.

The company announced that it will be adding 110 new languages to Google Translate, which it describes as its “largest expansion ever”.

According to Google, the newly added languages account for over 600 million speakers across Asia, Africa and Europe.

Widely spoken languages, such as Cantonese, with over 80 million speakers, and Punjabi with over 110 million, will join the Google Translate roster alongside less common languages. For example, the update will feature Manx, a language spoken on the Isle of Man, thought to have fewer than 2,000 speakers.

Heading towards 1,000 languages

The company is hoping to substantially increase the number of languages spoken on the app. In late 2022, Google announced its 1,000 Languages initiative, which aims to take the number of languages offered into the four-figure category. Although this still only accounts for one in seven of the languages spoken worldwide.

While Google hasn’t put a date on when it expects to reach its 1,000-language target, it’s speeding up the pace with which new languages are added. Prior to the June expansion, Google Translate offered 133 languages in total –  a figure that was almost with last month’s update.

Progress in adding new languages has been, perhaps unsurprisingly, driven by a greater use of artificial intelligence and machine learning. The languages added in 2022 used what Google calls ‘Zero-Shot machine translation’, where a language model is trained on a single language dataset rather than a dual-language one.

Many translation models will learn from datasets that look at examples of text in two languages to identify how a word or phrase is used in both – think of comparing how a famous speech or novel might appear in English and French, for example. For less popular languages, where there might not be enough data to build a bilingual dataset, the models are trained just in one language.

Disadvantages of Zero Shot AI translation

The Zero Shot translation can have its disadvantages, however.

“This advance is an exciting first step toward supporting more language technologies in under-resourced languages,” Google said in a blog when it debuted its first Zero-Shot languages back in 2022. “Most importantly, we want to stress that the quality of translations produced by these models still lags far behind that of the higher-resource languages supported by Google Translate.

“These models are certainly a useful first tool for understanding content in under-resourced languages, but they will make mistakes and exhibit their own biases. As with any ML-driven tool, one should consider the output carefully.”

More recently, Google has added PaLM2 to its Translate arsenal. The large language model helps Google Translate learn languages that are related to each other – such as variants of Creole – more efficiently. Or so the company claims.

Will AI replace translators?

Language-learning app Duolingo recently made headlines by announcing that it had replaced some of its human translators with AI.

In January, the company said it had laid off around 10% of its contract translators, and instead was using AI to create its lessons and enable users to talk with AI-powered chatbots to practice their language skills in real-time.

However, the market for translators appears to be holding steady in the face of increasing competition from artificial intelligence. In the US, for example, the number of people employed as translators appears to have increased by over 10% between 2020 and 2023.  

More on Google

• Google upgrades NotebookLM with more sources and faster fact-checking
• Google Gemini health check: how its AI services aim to help the medical industry (and make us fitter)
• Google Gemini is testing with 10 million tokens