Since the birth of spoken and written language, human translation has been essential to enable communication between different peoples and regions of the world. Whether translating Egyptian hieroglyphs into ancient Greek in 196 BCE, or translating from Cantonese to Portuguese in an 18th century seafood market, we humans, with our unique ability to understand context and cultural differences, are the only intelligent animals that can translate different languages.
However, today there is another element to consider:artificial intelligence (AI). Indeed, the latter challenges this idea (that humans are the only ones who can translate between different languages).
So how is AI changing the world of translation? Will human translators all find themselves unemployed in the near future? Relevant questions.
To understand AI, let's first consider its origins.
At first, artificial intelligence took the form of a thought experiment, by British polymath Alan Turing. The latter developed the “Turing Test” in 1950:it is a proposed artificial intelligence test based on the ability of a machine to imitate human conversation. This test consists of putting a human in a blind verbal confrontation with a computer and another human. If the person initiating the conversations is unable to tell which of their interlocutors is a computer, the software can be considered to have passed the test. This implies that the computer and the human will try to have a human semantic appearance. (To maintain the simplicity and universality of the test, the conversation is limited to text messages between the protagonists).
However, the vocabulary and technology that Turing was describing had not yet been developed at the time… The term “artificial intelligence”, abbreviated to AI, was actually first coined by researcher John McCarthy, when conference of the Dartmouth Summer Research Project on Artificial Intelligence , in 1956. This event sparked interest in machine learning and machine translation, leading to the advances we know today, such as Neural Machine Translation.
Children learn their language by listening to others and detecting language patterns. Similar to children, this recognition is used in the form of an AI called Neural Machine Translation (known as NMT). NMT uses a trained computer neural network to recognize patterns in the input dataset, for example, a Mandarin sentence, and then translate it into a desired output dataset, for example, an English sentence.
If we continue with our example of translating a Mandarin to English sentence, the network will be trained by receiving millions of Mandarin Chinese and English word and sentence associations.
Then, the computer will read a sentence in Mandarin and then guess what the sentence would be in English. The process repeated millions of times, the computer will learn to be more and more precise. Note that human engineers then test the system with a new sentence not used during training, to see if the system can learn to generalize.
Google Translate (Google Translate) uses NMT and is quite effective when it comes to widely used languages around the world (such as Mandarin and English). However, if you want to translate via Google Translate from a less widely spoken language, such as Samoan, it is very likely that there will be inaccuracies.
You may also like:Microsoft is laying off journalists to replace them with artificial intelligence
Other examples of limitations: some languages don't have gender-specific pronouns (eg Malayalam or Uzbek), but if you take a source text to translate, but it has a language with gender-specific features (eg French and English, …), then the NMT may generate errors.
The translation of an artificial intelligence is not always perfect, and humans (at least at present) always have the upper hand. However, compared to ten years ago, AI translation has improved significantly, especially with the help of current translation devices, such as the CM translator (a smart translator).
Before AI entered this field, human translators did all the translations themselves using a translation dictionary. But now, a professional translator often uses an NMT engine like Google Translate to do the first set of translations, then manually edit those translations and complete them.
According to experts, by 2022 most commercial translations will be done through NMT, with human editors to check and edit texts after the first translation. And that is absolutely not hard to believe, given that NMTs can translate documents more cheaply, sift through much higher volumes of text, and all of this, at a much faster rate than humans.
Aside from text translations, there have also been fantastic advancements over the past few years in the area of live voice translation. According to some experts, humans could soon arrive at a stage, thanks to AI and in the form of an NMT system, of an instant voice translation technology where every language is understood and translated.
