“Generative algorithms produce not language, but a simulated language”

AT he frontrunners in the debate on the uses of artificial intelligence (AI) are translators. Not only because they are sensitive to language and feel concerned in this regard, but because their profession, particularly for those who translate literature, has long been designated as an unattainable horizon for automation, while the astonishing spectacle of the results produced by algorithms is today weakening their economic base.

Some academic voices have recently spoken out to demonstrate their optimism about the future of the translation professions, with some adaptations to the irruption of AI in the sector: a pro domo plea to fight against the announced desertion of training academics in translation, in a context where the future of the profession worries students and families. With the key word adaptation, a categorical imperative here tinged with social Darwinism.

As a translation professional also involved in the field of training (professional and continuing), it seems useful to me to re-examine the purposes of translation training. Rather than preparing, in the name of so-called pragmatism, for these professions as they are redefined and precarious by the market, it seems important to me to reaffirm:

firstly, that literary translation is a factor of emancipation, in that it teaches those who carry it out how to use the language; that it is a powerful instrument for training the mind, from which all learners, whatever profession they pursue, can benefit;

secondly, that this benefit is largely reduced when the activity of translation is deprived of its creative dimension, when it is assigned to the task of monitoring a norm decided by the criterion of the most probable;

thirdly, that the statistics which underpin the calculations of algorithms tend to reduce the possible to the probable, and that this is in contradiction with the singularity of language, which is a condition of all true thought.

Reality and its simulacrum

Literary translators handle languages. This gives them an important social responsibility, because the ethics of language use cannot be limited to correcting the sexist and racist biases of algorithms. In the vast majority, they are opposed to the use of their texts as fuel for a large language model. Far from being opposed in principle to technological innovation, they nevertheless know the ontological difference between a language and a semblance of a language, between a subjectivity nourished by human experience and a writing, however correct it may be, but devoid of any responsibility. Sharing this knowledge allows readers to maintain a distinction between reality and its simulacrum.

You have 53.48% of this article left to read. The rest is reserved for subscribers.