It seems every industry right now is talking about machine learning, artificial intelligence, and other technologies replacing humans in the workplace. Language industries are no different. In fact, automatic translation services and machine translation are some of the biggest driving forces behind the development of such technologies.
It’s pretty obvious how serious Google, Microsoft and their rivals are about automatic translation – and machine learning is powering their latest breakthroughs. However, despite all the hype, machine learning will never be enough to replace human translators.
Automatic translation robots
There are two problems with the idea of machine learning replacing humans in just about any industry. Above all, machine learning by nature (if you can call it that) is ultimately confined to rules defined by human programmers.
In theory, it can “learn” by collecting data and use this to improve its functions but it’s always relying on what it already knows. Machine learning isn’t about to make creative decisions, know when to break the rules or create an original argument on any topic.
Essentially, it can only repeat what humans have already done – and there’s nothing wrong with that. This is what it’s designed to do but too many tech companies are trying to sell machine learning as something it isn’t.
Tech companies are overselling this technology
Using terms like artificial intelligence and machine learning is a great way to make these types of automatic translation products sound more impressive but it also creates a false sentiment about the progress of technology.
To borrow the words of Ilia Kolochenko, CEO of High-Tech Bridge, a company that provides web security products powered by machine learning: “We should not over-estimate it [machine learning] – we need to understand that it can’t replace humans, instead it can significantly aid human tasks.” [Source: Infosecurity]
Kolochenko isn’t underselling machine learning there – that wouldn’t be in his interests – and we’re not trying to either (we have a lot to gain from the technology ourselves). However, Kolochenko understands both the limits and opportunities it comes with. Machine learning will do great things for the workflow of human beings, automating a wide range of tasks that take up professionals’ valuable time – and this has always been the game plan.
Imagine never having to write another to-do list or text message. Better yet, imagine an automatic translation platform that can help get human translators 50% of the way to a finished project. That’s twice the workload each translator would be able to take on, or – in other words – half the cost and turnaround time on each of your translation projects.
Of course, machine learning isn’t anywhere near close to achieving this yet, but the technology is making progress every day and capable of “learning” by itself.
What will machine learning do for translation?
Automatic translation services will continue to improve and play a larger role in translation projects, thanks to machine learning. As the technology becomes more accurate it will be capable of translating basic sentences with accuracy and even understand certain levels of context, as Google is already achieving with search queries in multiple languages.
The problem for machine learning and automatic translation is – no matter how good it gets at learning the rules of languages – it will never be able to grasp their finer nuances and how these change through the translation process.
Machine learning relies on a set of rules to function, but human translators constantly have to bend and break the rules in order to capture the same meaning between different languages. This is a process you can’t define by rules or automate, ruling out machine learning from this part of the translation process.
However, machine learning will have a huge impact on the industry.
The biggest impact machine learning for us won’t actually have anything to do with languages. It will be in automating as much of the project workflow as it can handle. In some cases this might be translation but most of the time this will be tasks like writing emails, collecting data, building glossaries and other tasks that don’t need the human touch.
Even on a purely admin level, machine learning will make very translation project faster, more accurate and more cost-effective without needing to translate a single sentence.
Will machines ever replace human translators?
This is another question entirely. Machine learning is not the same as artificial intelligence and the definition of AI has become more complex – mostly because tech companies have lowered the bar on what they consider artificial intelligence. If you listen to the likes of Google, Adobe and the leading names in tech, they’re already using “AI” to replace human beings.
However, we’re nowhere close to achieving true AI – a system that would be capable of processing the mix of creative, cultural and moral through processes translators use in their profession. We don’t even know yet if this kind of AI is really possible but we know for sure the technology giants are intent on seeing how close they can get.
Either way, machine learning doesn’t come close in this regard but it will empower humans to get a lot more done, faster – and that’s fine by us.