People stuff and computer stuff

Come on in and take a look, welcome to the future! We’re in 2023 and if anything has marked the last few weeks - apart from Shakira’s latest diss track - it’s the startling amount of news stories featuring artificial intelligence that have presented it as a miracle worker. Image creation, bursting into the world of music, the explosive appearance of ChatGPT… The headlines keep pouring in, with each feat bigger than the last. From the first block, we could highlight the endearing collection of images that have personified Spanish regions, from the second we could collect enough material to teach a semester of audiovisual translation and from the third… Well, we’ll say that professors have been the first ones to face the consequences. The famous OpenAI chatbot walks through university corridors, it’s achieved pass marks in Law exams and it’s become the tool that will apparently dethrone Google as our preferred search engine.

The possibilities become greater in the same way as the dangers, so the web has quickly become full of perspectives that speak of both heaven on Earth and the terrifying future predicted by The Terminator. And anywhere between the two, of course. The most recent news stories that hint at 180 Buzzfeed employees being fired to be replaced by AI don’t seem very promising. We’re starting to hear the cyber rumour of the list of jobs that are in danger of extinction because of the arrival of artificial brains that are capable of beating the possibilities of human beings with flying colours.


There is some truth in these fears: computers are computers. This makes them the ideal candidates to do computer stuff, but not people stuff. Obviously, this sounds like a truism or something ridiculous Joey Essex would have said on TV; but this key principle is also the starting point for the topic at hand, seeing as translation is clearly a human thing. Language is a human thing. The description of reality is a human thing, as is the description of emotions. Machines can imitate them, there’s a reason why the Turing test was created and there’s a reason why Blade Runner has become a science fiction gem, but there’s a long way to go before machines can experience them. Replicants (Blade Runner) and Jane (Speaker for the Dead) aside, it’ll be a while before the conscience of flesh is passed on to metal. In order to speak like a human, you must live life and think like a human.

This isn’t the first time we’re seeing this linguistic debate. We all remember the scandal that occurred when it was revealed that the Squid Game subtitles had been created through the use of machine translation and then post-edited later on by a real person. We now find ourselves in the same position, but with another turn of the screw. The original problem is still the same, the only thing that changes is the power of the tools that carry out the computerized work.


Every machine translation tool has been previously fuelled with information provided by human beings – us. Whether that’s through a huge amount of text or through the insertion of a series of linguistic rules that allow the message to be transferred from one language to another, the machine collates the information that we’ve given it and provides a result that can be impeccable in many cases. Being human is funny, as when we do the same job, our mind doesn’t think about the texts that we store in our brain or universal rules, but it puts all its neuronal effort into understanding the source text and expressing it in the target languages. We can see it clearly here: computer things and human things.

Even if these computerized tools have our own language as a starting point, with its corpus and rules, the result isn’t the creation of language, but an analysis of the verbal language that has been given so as to make an imitation of what it has been taught. I’ll say it again: being human is a funny thing. In translation classes, automation is the first thing that professors try to get rid of in students. Critical thinking and deep reading are encouraged, with the aim to identify alliterations, puns, degrees of verisimilitude, registers, sociolects, idiolects, irony, double meanings, rhymes, arguments, voluntary ambiguities, inferences and several resources in order to be sure that, when we adapt complex elements, the receiver can understand them easy thanks to our way of expressing them. None of these aspects have anything to do with the words, syntaxis and semantics strictly speaking, but with the way that it adopts the message once it reaches the receiver’s brain. In the same way that sounds are just waves until our ear decodes them, texts are just lines on paper until our brain interprets them. And when we say brain, we mean ‘human brain’.

Language is ambiguous by nature and everyone expresses themselves in a way that suits them. Reading J. R. R. Tolkien isn’t the same as reading George Orwell, C.S. Lewis, or Agatha Christie. Anyone interested in the world of words will agree, I suppose, that this is where the magic lies. Jorge Luis Borges, the well-known Spanish author and translator, translated, with better or worse luck, Orlando by Virginia Woolf. Julio Cortázar, also Spanish author and translator, did the same with Edgar Allan Poe’s tales. If these works have been exchanged, or if we’d given them to a class of translation and interpreting students and we compared the results, we’d discover a voice behind each proposal. A voice that’s unrepeatable and one-of-a-kind. Both in its perception of the text and its expression, thanks to the nuances that we’re able to consider beyond grammar.

That’s precisely the obstacle that machine translation, at least in a near future, won’t be able to overcome. From 0 to 10, human translation enhances the amount of options to read, of interpretations, of proposals, of voices and options; while the machine translation that we have, as correct and pragmatic as it may be, will always produce the same text given the same conditions and will never change its initial 5. If someone hasn’t been convinced by this explanation, they should go to a class of Spanish teenagers and give them an insult in English to translate however they want: ‘jerk’. After seeing the colourful range of responses to this four-letter word - and let’s not talk about a novel! - I think the verbal creativity of the human brain will clearly beat the ‘idiota’ that machine translation tools produce as their only solution.

P.S. We’ve asked Chat GPT and it agrees with this line of argument.

Writer: Maite Madinabeitia
Translator: Katie L. Wright
Proofreader: Elisabet Pina