In his short story ‘The Great Automatic Grammatizator‘ (1953), Roald Dahl imagined a computer that can write stories better than humans can.
The computer ends up writing “over half of all the novels and stories published in the English language.” The remaining human writers struggle to make a living…
It’s estimated that the machine translation market size will be worth USD $7.5 billion by 2030.
Does this growth represent the impending arrival of ‘great automatic grammatizors’ in the translation industry? Or will human translation continue to dominate the industry?
Let’s take a look at this question in more depth…
Who are human translators?
Translators have worked on governmental, religious, and literary projects since ancient times.
Traditionally, documents were translated from scratch or based on earlier versions by individuals or teams.
For example, 6 groups making up a total of 47 academics and clergymen translated the King James Bible, published in 1611.
In recent history, commerce has been the field where most translators have found work. Commercial and other types of globalization have in turn led to an increasing number of capable linguists, as well as a large and varied market for their skills, too.
Education has increased translators’ abilities further. Universities around the world now have translation studies courses and other professional bodies offer linguistic training and certification.
How does human translation work?
Nowadays, the translation process has been improved by general quality assurance processes and best practices, as well as computers and specialist translation technology.
Many translators work as freelancers or employees of translation companies or language service providers (translation companies that also provide other language services).
They generally work in an office or remotely on projects overseen by translation project managers and aided by translation software, proofreaders, desktop publishing (DTP) specialists, and other language service workers.
They are also increasingly engaged in localization (adapting translation for different cultures) and transcreation (creative translating/copywriting also aimed at adapting to different markets) work, too.
What is machine translation?
Machine translation (sometimes referred to as MT) is a name for software systems used to translate text between languages.
It is a complex and fast-evolving field with a history dating back to the 1950s. At present, there are many different models for it. These include:
- Example-based machine translation uses machine learning on parallel texts in two languages to build its capabilities
- Hybrid machine translation uses two or more MT systems within an overriding translation process framework
- Neural machine translation (NMT) uses neural networks(computer systems designed to mimic brains) to predict sentence outcomes
- Rule-based machine translation uses the input of linguistic rulesets
- Statistical machine translation uses large bodies of bilingual texts (such as law books) to build models about the most likely translation
In the translation industry today, the consensus is that neural machine translation has the most potential for growth.
What’s the difference between machine translation and AI translation?
These two terms are often used interchangeably, but the difference is that all artificial intelligence (AI) translation is also machine translation. But not all machine translation is also AI translation.
AI machine translation would be able to judge the quality of its own output, whereas non-AI machine translation simply produces an output based on the input it has been given.
More generally, AI is the name of the field for creating machines that can replicate human intelligence. These machines use machine learning to improve their own intelligence. Whereas machine translation is simply the name for any kind of automated translation.
Examples of machine translation today
Machine translation has already replaced some translation that was previously done by humans. Let’s take a look at some of the popular examples.
Google Translate is Google’s own neural machine translation tool.
There are other free online tools such as Bing Translate, Baidu Translate, etc., but with a dominance of nearly 90% of the search engine market, Google’s translation tool is the most popular globally by far.
And its usage is set to grow: it will soon launch a local news translation service.
At present, it’s difficult to find recent data on its usage. But we do know that back in 2016 it had over 500 million users and translated over 100 billion words per day.
It’s almost impossible to estimate what proportion of these users would otherwise have turned to human translators. However, even 0.001% would equal 5 million more potential customers for the translation industry each day.
Computer-assisted translation (CAT) tools
CAT tools have been popular in the translation industry for a while now. They divide text into strings in order to automate the translation of repeated phrases elsewhere in the document(/s).
One of the biggest benefits of CAT tools is their translation memory. The more they are used, the more comprehensive they become.
They often work in conjunction with translation management systems (TMS), which help manage translation workflows. This combination of these tools further ensures consistency on and between projects.
There are a number of ways in which CAT tools can be used. For example, post-edit machine translation (PEMT) is when machine translation completely translates text before human translators go through and edit it.
Arguments for machine translation replacing human translators
1. Technological progress is inevitable
Mathematician and physicist John von Neumann once said:
“Computers are like humans – they do everything except think. If you tell me precisely what it is a machine cannot do, then I can always make a machine which will do just that. The sciences do not try to explain, they hardly even try to interpret, they mainly make models.”
This line of thinking is common in discussions about the future of translation. Some thinkers predict that machine translation is bound to eventually replace human translators.
They cite technical progress in other areas of our life. After all, it wasn’t long ago that the idea of smartphones in everyone’s pockets would have seemed strange.
When could Machine translation replace human translators?
For people confident in the technological takeover, there are simply two questions left to consider:
- When will human translators be replaced?
- Will it happen gradually or suddenly?
There is a lot of the same kind of speculation about many other industries, too. Drivers, lawyers, doctors, and soldiers, for example, are all careers that some predict will sooner or later be outsourced to technology.
Yuval Noah Harari is one such advocate of this view. He believes that because of the advantages of “connectivity and updatability” in some lines of work, machine replacement of humans is certain:
The automation revolution will not consist of a single watershed event, after which the job market will settle into some new equilibrium. Rather, it will be a cascade of ever-bigger disruptions. Old jobs will disappear and new jobs will emerge, but the new jobs will also rapidly change and vanish. People will need to retrain and reinvent themselves not just once, but many times.
- ‘WHY TECHNOLOGY FAVORS TYRANNY‘, The Atlantic (Oct, 2018)
2. Machine translation better suits digital environments
There are already popular plugins and web browser extensions that automate part of the website translation process.
Fields such as news, hospitality, and eCommerce all need to be constantly updated in order to optimize their accessibility.
These fields are improved further by the application of developments like document-level context. This is when machine translation applies an understanding of the document as a whole (as opposed to individual words and sentences) to select terms for translation.
These machine translations may contain errors, but then so too does some work done by the average human translator.
A similar argument is often made about self-driving cars: they may accidents, but these will be at a much lower rate and easier to correct afterwards than accidents by human drivers.
Arguments against machine translation replacing human translators
1. Machine translation will never understand context and nuance as well as humans
A modified version of this position is that it is only true in some areas. These include literary and marketing translation, for example.
Areas like these often require more creative localization than translation of manufacturing documents, contracts, etc.
Furthermore, the authenticity of the translation might be considered by the target audience. Audiences may simply prefer to engage with translations they know were done by a human.
2. Machine translation is simply a tool for human translators
An optimistic and increasingly common view is that machine translation is an opportunity for translators to improve their output’s quality and quantity.
From this point of view, translation is simply a disruptive technology that will increase human translators’ productivity. In some areas, it already is.
And with the growing body of content proliferated by the growth of internet users around the world, translators’ earning potential could grow a lot.
“I think machine translation is good enough for post-editors – or translators that become post-editors, to gain productivity and pass on some of their savings to their clients. What I think is that at some point in the future, maybe what is now called ‘a translator’ or ‘a post-editor’ is going to be some sort of a specialist who is going to be an amalgamation of both and maybe more. Maybe someone who can train an engine, be creative with a headline, etc.”
3. Machine translation won’t have enough data from all languages
Languages are complex. Learning one well enough to be able to translate into or from it takes a lot of time for a human and a lot of data for a machine.
The data for all languages is not the same. Widely spoken languages such as English and Spanish have huge data sets to work with, but they are only two of the world’s estimated 7,000 languages.
Will interpreters become obsolete?
Interpreters are people who give spoken translations between languages. There is a lot of crossover between their services and those of translators.
They might also be at risk of being replaced by machines. Meta (formerly Facebook) is developing an AI-based spoken translation tool. And it’s not the only company working on this.
Speaking about this rising technology, Meta founder and CEO Mark Zuckerberg said at a recent presentation:
“The ability to communicate with anyone in any language — that’s a superpower people have dreamed of forever, and AI is going to deliver that within our lifetimes.”
However, it’s worth noting that even if these kinds of tools do become commonplace, they won’t necessarily replace human interpreters in every sphere.
It can be argued that only human interpreters, with their cultural knowledge and understanding of social and psychological cues, can truly be trusted for certain projects.
Fear of machines replacing humans’ work has been around for a long time.
Some dismiss these fears as sci-fi-inspired fantasies that ignore the fact machines usually end up being co-opted by humans.
Others argue that the nature of AI means that the coming technological revolution is going to be fundamentally different from what came before it.
Technology has transformed how translation is done and how its quality is maintained and reviewed.
It has lowered turnaround times and translation rates. Some believe that it could change the translation industry even more radically in the near future.
But knowing which translation tools to use in the first place can be difficult. Some are more or less suited to different tasks. Each translation company will likely find their own solution in this area (or take outside consultation on the issue).
The translation industry has already seen machine translation adopted in some areas. Even Google Translate likely takes some work away from human translators.
There are good arguments both for and against the answer to this question. The only certainty we can have is that both sides will likely still be debating it up until there is a definitive answer.
That is, of course, presuming that there will be…