Lately, artificial intelligence has become an almost obligatory tool through which many people ease their daily work. From ChatGPT (text) and Midjourney (image) to a lot of other tools, including those that offer translation services (and analyze texts, translate them, transform them, and render them into various forms other than those they were conceived in), we find on the online market a plethora of applications capable of simulating how a person does a certain task. Tools like Google Translate or Deepl are already used by many people who want to easily translate various texts the meaning of which they can’t grasp because either they don’t know the language, or they don’t have time and it would take too long or cost too much to obtain a professional translation.
Are AI-automated translations good?
Still, even though these tools have started using artificial intelligence to provide results increasingly close to how a specialist accomplishes the task of translating, they still haven’t reached the performance of perfectly reproducing how a real person uses words to convey a specific message, and this is because it’s very complicated for an algorithm to capture all the nuances that words can easily convey when put together in a sentence or phrase by a human being.
AI doesn’t understand what it writes
Even though the algorithm is able to predict what word would be most likely to follow in a sequence of words, AI does this based on existing texts that it has structurally assimilated without understanding them.
The quality of AI-generated texts is not always the right one
Moreover, the quality of the texts used to feed artificial intelligence is not necessarily right for every situation. For example, if an artificial intelligence model is trained on texts from Reddit, it won’t be able to produce valuable literary passages, and vice versa. It’s true that usually the algorithm doesn’t make mistakes, but it’s also true that such texts are subject to errors and clichés.
In language, meanings are fluid
Also, many words and expressions are either new entries (various neologisms, abbreviations, specialized terms) or they are used to convey figurative meanings, and it’s very possible that if we rely on applications instead of turning to a professional, we may actually get an approximate translation that doesn’t reflect the reality of the text we need to translate.
Of course, when we want to quickly understand what a text we don’t comprehend is about, if it’s in a language we haven’t learned at school or for a project that is not essential for our work or any other important purpose, we can resort to applications and translation algorithms, because we can get a general idea of what the text means. Otherwise, there are risks.
However, we will talk about the dangers of using artificial intelligence in translations in a future article.