The Human Translator versus the Machine: Algorithms of Interpretation

Machine translation has improved to such an extent that a time can be predicted when most, if not all, translation will be performed by machine. Both translators and translation studies scholars have reacted to this development with questionable thinking about translation as well as human creativity. Definitions of creative translation generally assume an instrumentalist model: translation understood as the reproduction or transfer of an invariant contained in or caused by the source text, whether its form, meaning, or effect. When joined to such other discourses as Gricean pragmatics, the Bourdieusian habitus, or empirical psychology of reception, this model actually limits the translator’s creativity because it constitutes an impoverished conception of what translation is and how it might be appreciated. Technological progress that enables machines to attain a basic literary translational competence can foreground the hermeneutic function of human translators—but only if we assume a hermeneutic model where translation is understood as the inscription of an interpretation which inevitably varies source-text form, meaning, and effect according to intelligibilities and interests in the receiving culture. Harnessing the machine through computational text analysis and large language models to serve the translator’s interpretative act can redefine translation by foregrounding the tasks that make it at once scholarly and creative. The idea is to construct a collaboration with the machine that redefines and enlarges the scope of involvement for the human translator, increasing its learning and sophistication.