sageata articol sageata articol

Înapoi la Blog

Will AI replace the human translator in 2023?

16/01/2023

From the moment you woke up this morning to the moment you’re reading this piece, you’ve been through tens, maybe hundreds of translations. Not one ever stood out. This is because you are used to them. For example, your phone screen is translated into English from a source language, so you don’t feel the effort when you hit the snooze button every morning. But we know you still do.

You plug in the coffee machine and use it by pressing “on” or “start”, even though the guy who placed the button there is probably from Japan. The same goes for your television, which has an English menu even though it came all the way from South Korea.

Translations make the world go round, they unite people. Yet many voices say translations will be done without humans one day. Whether we’re talking about your 70-inch TV user manual, the brochure guide from the you-know-what furniture company or the sandwich maker menu, translations will supposedly become exclusively AI-generated any day now.

Smart remote – Free image courtesy of Unsplash.com

What is this artificial intelligence?!

For most of us, artificial intelligence is more of a movie subject rather than a real-life thing. Even though we use it every single day, as we do with translations, we don’t really know much about it. To put it simply, AI is that tool which makes our lives easier without us even knowing about it, being actively present in basically all aspects of daily life.

Lensa AI, for example, is a photo editing app that can turn your selfie, or any selfie, into an animation using artificial intelligence. If you ever dreamed about turning yourself into a superhero, you may do just that with a few clicks, creating your own comic book cover. Lensa’s idea that was spread across the globe intrigued us: is this what the future translator will look like?

AI and translations

An automated translation system that uses AI can analyse an English text and then generate a translation into Romanian, restoring the context and even conveying the tone and the author’s intentions. This type of system can be really useful in many areas, such as business, tourism, international communication and so on.

The previous paragraph was entirely generated by ChatGPT, an artificial intelligence robot which uses the internet to generate on-demand written content. You can write anything in the chat box and the software will generate the text exactly as instructed. The more concise you are, the better this “virtual lamp genie” can generate your desired text.

Another example is DALL-E. This software transforms into an image any instruction you write in the search box – literally any instruction. No matter what you write, say ‘translator of the future’, you will get an image based on what AI thinks you visually asked for. In our example, we’ve asked the program to make up the image of the translator of the future.

Translator of the future – Images generated by DALL-E

Surprisingly, the machine delivered interesting results, very close to what we have in mind when brainstorming about the translator of the future and the translation business in general. A combo between the power of computer, through the software we are using and some that will come later, and the human decision-making process, the unique tool we use to craft the translation and make sure that the text is fluent, enjoyable, grammatically correct and helpful.

Is the future of translators uncertain?

The quality of machine learning translations is on the rise. One study made by the people at Translated, a Rome-based company offering language services, shows that based on the current rhythm of development, six years from now, computer-generated or assisted translations will become as good as the ones delivered by top human translators. Experts analyzed the data from 136.000 specialized translators using Matecat, their in-house CAT tool.

„Humans invented the computer, not the other way around, and as far as people stay imperfect, so will their inventions.”

Statistics show that the required time for translations is in constant decline, just like with the computer-assisted type. If the “perfect” translation used to take three seconds, now it takes just two seconds out of the translator’s time. The verdict? The final report says that the betterment of translation times will make the whole process more efficient, not get rid of the human touch required in translations.

Good news for translators

At our agency, we use software on a daily basis which help us translate a large number of words at the same time. Some of our orders sometimes add up to half a million words, which would take weeks, even months to be translated entirely manually. We use both the almighty machine learning, which allows us to translate faster, and also terminology databases and glossaries with millions of words we’ve used for our clients in the past.

Library storage – Free image courtesy of Unsplash.com

No matter how fast a computer can learn to translate, it is human intervention that remains crucial, after all, in delivering a final, imperfect product that is as close to perfection as possible. Humans invented the computer, not the other way around, and as long as people remain imperfect, so are their inventions.

Today’s machines are capable of executing a bunch of things much better than us, humans. They work fast, on-demand, and they never get tired. The key word here is ‘execute’. Computers can’t come up with ideas or products without receiving instructions from a human. Artificial intelligence lacks imagination, so it will never be able to create an ad campaign or write a great movie script. AI has no soul, the element which gives true power to art and creation.

At most, humans and machines can work together to become a hybrid translator, where the machine drives the individual to reach 99% perfection.

The remaining 1% belongs to nature.

click here for the Romanian version —