For thousands of years, translations were carried out exclusively by human beings. But with the development of computers, the concept of machine translation became practical in principle as outlined by Warren Weaver in his memorandum ‘Translation‘ in 1949. Early expectations of machine translation were disappointed despite significant research funding, and it was not until the late seventies and early eighties that machine translations started to be used by commercial organisations. Initially, the quality of such translations was poor, and they were only useful in limited areas (e.g. weather forecasts). As a result, most machine translations still needed to be enhanced by human post-editors, and it was generally believed that machine translations were (and would remain) too poor to be widely useful. As a result, it seemed unlikely that the future of human translators would be impacted.
Statistical machine translation, based on information theory, was introduced in the late 1990’s and improved the quality of machine translations. This made the combination of machine translation with post-editing an attractive option, but whether the productivity and cost-effectiveness of this approach is better than that of using human translation is still unclear (Koponen 2016). However, the quality of machine translation is increasing at a greater rate than many people expected, particularly with the introduction of neural machine translation by companies such as Google and Microsoft. This means that an ever-increasing proportion of routine and technical translations can be carried out by machine translation at a relatively high standard. Google Translate already translates more words in one day than all human translators in the world translate in a year (Massardo & van der Meer, 2017: “The Translation Industry in 2022”).
What are the implications of this ongoing development for the future of human translators? Does this mean that, as discussed by Jaap van der Meer of TAUS, ‘the future does not need translators’? Mr van der Meer points out that “The time is coming when machines will be better at transforming a text from one language into another than a human translator.” He refers to the moment when this will happen as the ‘singularity’, and quotes the title of a book by Ray Kurzweil (now Director of Engineering at Google) entitled ‘The Singularity is Near‘. van der Meer believes that the future of human translation will be in ‘really creative translations’, ‘hyper-localization’ and ‘transcreation’. Kurzweil now predicts that machine translation will be good enough to replace most human translators by 2029. See also “The Story of the Translation Industry in ‘22“
Van der Meer’s ‘really creative translation’ is an interesting concept. For thousands of years there have been discussions about ‘word for word’ (literal) translations as opposed to ‘sense for sense’ (free) translations, but I’m not sure what the relationship between a ‘really creative translation’ and the source text would be.
My own view is that ‘technical’ translation will increasingly be taken over by machine translation, while human translators will gradually be squeezed into post-editing and the dwindling segments of translation that are still too difficult for the machines, such as ‘literary’ translation (I would include texts such as poetry, perhaps magic realism, and texts involving local references, double meanings or a sense of humour, such as advertising material).
Postscript 1 (from Johnson, ‘Translation platforms cannot replace humans‘ The Economist 29 April 2017):
“Although this might put a few post-modern poets out of work, neural-translation systems aren’t ready to replace humans anytime soon. Literature requires far too supple an understanding of the author’s intentions and culture for machines to do the job. And for critical work – technical, financial or legal, say – small mistakes (of which even the best systems still produce plenty) are unacceptable; a human will have to be at the wheel to vet and edit the output of automatic systems.”
Postscript 2 (from Johnson, ‘Why translators have the blues‘ The Economist 27 May 2017):
“Translation can be lonely work, which may well be why most translators choose the career out of interest, not because they crave attention. Until recently, a decent translator could expect a steady, tidy living, too. But the industry is undergoing a wrenching change that will make life hard for the timid.” The Economist points out that inexperienced buyers who not don’t know much about foreign languages or translation quality will purchase primarily on price. At the same time, machine translation has improved dramatically with translations based on neural networks. Johnson predicts that “Translators in the bulk and middle markets will inevitably be doing more editing, or will be squeezed out”, although literary translation is not threatened.
See also: Videofix: Why a debate on machine and human translation at all? from the Terminology Coordination department of the European Parliament, and ‘Where will humans be in 2050?‘
Image from Olof Mogren, ‘History of machine translation systems‘; slide by Christopher D. Manning.
History of machine translation – Christopher Manning