AI Translation: Hope or Hype?

We often hear it said that 2018 is the year of AI in translation. Recent debacles in the use of translation machines, however, suggest otherwise.

In April, Tencent’s Mr. Translator, a simultaneous interpreting and transcription technology, produced gibberish, repeated words, and broken sentences in the Boao Forum, also known as Asia’s Davos.

Just as an example, the machine translated the ‘Belt and Road’ Initiative, a major development strategy launched by the Chinese government in 2013, as ‘road and waistband’.

The Chinese tech giant defended itself by blaming the speaker for saying ‘road and belt’, which confused the machine. However, they did admit on social media that the system had ‘made errors’ and ‘answered a few questions wrongly’.

In September, iFlytek, a Chinese listed company specialised in voice recognition, was accused by an interpreter of stealing human output and presenting it as automated translation in the 2018 International Forum on Innovation and Emerging Industries Developments.

iFlytek’s stock price plummeted immediately after the fiasco. As of mid-October, the price has dropped 30%, hitting rock bottom.

In fact, this isn’t the first time iFlytek has been caught red-handed. In 2017, the company hid the onsite interpreters away from the speakers and audience, pretending the interpretation was an AI product.

These incidents make it clear that machines are falling short of the expectations and praise spouted by tech companies and the media. Importing numerous glossaries into a translation system does not inevitably lead to satisfactory output.

As AI researcher Erik Cambria put it: ‘The biggest issue with machine translation today is that we tend to go from the syntactic form of a sentence in the input language to the syntactic form of that sentence in the target language. That’s not what we humans do. We first decode the meaning of the sentence in the input language and then we encode that meaning into the target language.’

It is very unlikely that a trained interpreter would make an error like the ‘road and waistband’ example. A qualified interpreter can effectively process hidden meanings, as well as minor mistakes that occur during speech. An experienced interpreter analyses the context of the speech rather than translating it word-for-word.

It is not fair to say that attempts to apply AI in the translation industry are fruitless. Online translation platforms, such as Google Translate, no doubt make traveling and basic communication much easier. However, over-reliance on AI translation can be problematic. Companies that use machines to cut costs are at risk of humiliation or even financial loss.

AI’s effectiveness should be neither underestimated nor put on pedestal. Machines, at the stage, are still assistive tools. Human interpreters – experienced, trained professionals – are the key to the success of events, conferences, and meetings.